Datasets:
wanshenl
/
pgsql-hackers

Dataset card Data Studio Files Community
threads
listlengths
1
2.99k
[ { "msg_contents": "> > Do we need *both* database & tablespace to find table file ?!\n> > Imho, database shouldn't be used...\n> \n> That'd work fine for me, but I think Bruce was arguing for paths that\n> included the database name. We'd end up with paths that go something\n> like\n> \t..../data/tablespaces/TABLESPACEOID/RELATIONOID\n> (plus some kind of decoration for segment and version), so you'd have\n> a hard time telling which files in a tablespace belong to which\n> database. Doesn't bother me a whole lot, personally --- if one wants\n\nWe could create /data/databases/DATABASEOID/ and create soft-links to\ntable-files. This way different tables of the same database could be in\ndifferent tablespaces. /data/database path would be used in production\nand /data/tablespace path would be used in recovery.\n\nVadim\n", "msg_date": "Mon, 26 Jun 2000 15:15:39 -0700", "msg_from": "\"Mikheev, Vadim\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Big 7.1 open items " }, { "msg_contents": "\"Mikheev, Vadim\" <[email protected]> writes:\n> We could create /data/databases/DATABASEOID/ and create soft-links to\n> table-files. This way different tables of the same database could be in\n> different tablespaces. /data/database path would be used in production\n> and /data/tablespace path would be used in recovery.\n\nWhy would you want to do it that way? Having a different access path\nfor recovery than for normal operation strikes me as just asking for\ntrouble ;-)\n\nThe symlinks wouldn't do any good for what Bruce had in mind anyway\n(IIRC, he wanted to get useful per-database numbers from \"du\").\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 26 Jun 2000 18:48:22 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items " }, { "msg_contents": "Tom Lane wrote:\n\n>\n> The symlinks wouldn't do any good for what Bruce had in mind anyway\n> (IIRC, he wanted to get useful per-database numbers from \"du\").\n\nOur database design seems to be in the opposite direction\nif it is restricted for the convenience of command calls.\n\nRegards.\n\nHiroshi Inoue\[email protected]\n\n\n", "msg_date": "Tue, 27 Jun 2000 13:07:28 +0900", "msg_from": "Hiroshi Inoue <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" }, { "msg_contents": "> Tom Lane wrote:\n> \n> >\n> > The symlinks wouldn't do any good for what Bruce had in mind anyway\n> > (IIRC, he wanted to get useful per-database numbers from \"du\").\n> \n> Our database design seems to be in the opposite direction\n> if it is restricted for the convenience of command calls.\n\nWell, I don't see any reason not to use tablespace/database rather than\njust tablespace. Seems having fewer files in each directory will be a\nlittle faster, and if we can make administration easier, why not?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 27 Jun 2000 00:16:49 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" }, { "msg_contents": "Bruce Momjian wrote:\n\n> > Tom Lane wrote:\n> >\n> > >\n> > > The symlinks wouldn't do any good for what Bruce had in mind anyway\n> > > (IIRC, he wanted to get useful per-database numbers from \"du\").\n> >\n> > Our database design seems to be in the opposite direction\n> > if it is restricted for the convenience of command calls.\n>\n> Well, I don't see any reason not to use tablespace/database rather than\n> just tablespace. Seems having fewer files in each directory will be a\n> little faster, and if we can make administration easier, why not?\n>\n\nI only objected to emphasize the advantage in getting useful per-database\nnumbers from \"du\". It's only misleading in dbms design.\n\nRegards.\n\nHiroshi Inoue\[email protected]\n\n", "msg_date": "Tue, 27 Jun 2000 17:53:10 +0900", "msg_from": "Hiroshi Inoue <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" } ]
[ { "msg_contents": "Hello,\n\nI just upgraded from 6.5.2 to 7.0.2 on SuSE 6.4 Linux, and now I get this\nmessage when trying to insert records:\n\nERROR: fmgr_info: function 19104: cache lookup failed\n\nThis only happens on tables that use refint.so (under contrib/spi in the source\ndistribution). There is a trigger that calls a function, to check primary or\nforeign key. Has anyone else encountered this? \n\nHere's the schema:\n\ncreate subscriptions table\n-------------------\nCREATE SEQUENCE \"subscriptions_script_id_seq\" start 1 increment 1 maxvalue \n 2147483647 minvalue 1 cache 1 ;\nCREATE TABLE \"subscriptions\" (\n\t\"script_id\" int4 DEFAULT nextval('subscriptions_script_id_seq') NOT NULL,\n\t\"mag_id\" int4 NOT NULL,\n\t\"paid_amount\" money NOT NULL,\n\t\"start_date\" date NOT NULL,\n\t\"end_date\" date NOT NULL,\n\t\"notes\" text);\nCREATE UNIQUE INDEX \"subscriptions_script_id_key\" on \"subscriptions\" using btree \n ( \"script_id\" \"int4_ops\" );\nCREATE TRIGGER \"sub_mag_trigger\" BEFORE INSERT OR UPDATE ON \"subscriptions\" \n FOR EACH ROW EXECUTE PROCEDURE \n check_primary_key ('mag_id', 'magazines', 'mag_id');\n\ncreate function\n------------\nCREATE FUNCTION \"check_primary_key\" ( ) RETURNS opaque AS\n '/usr/local/pgsql/lib/modules/refint.so' LANGUAGE 'C';\n\n\n -- \n\n Michael Fraley\n [email protected]\n http://www.usfca.edu/~fraley\n", "msg_date": "Mon, 26 Jun 2000 15:40:55 -0700", "msg_from": "Michael Fraley <[email protected]>", "msg_from_op": true, "msg_subject": "ERROR: fmgr_info: function 19104: cache lookup failed" }, { "msg_contents": "\n\nThis works, but you may create the function first...\nAnd then the trigger...\n\nOn Tue, 27 Jun 2000, Michael Fraley wrote:\n\n> ERROR: fmgr_info: function 19104: cache lookup failed\n\n> create subscriptions table\n> -------------------\n> CREATE SEQUENCE \"subscriptions_script_id_seq\" start 1 increment 1 maxvalue \n> 2147483647 minvalue 1 cache 1 ;\n> CREATE TABLE \"subscriptions\" (\n> \t\"script_id\" int4 DEFAULT nextval('subscriptions_script_id_seq') NOT NULL,\n> \t\"mag_id\" int4 NOT NULL,\n> \t\"paid_amount\" money NOT NULL,\n> \t\"start_date\" date NOT NULL,\n> \t\"end_date\" date NOT NULL,\n> \t\"notes\" text);\n> CREATE UNIQUE INDEX \"subscriptions_script_id_key\" on \"subscriptions\" using btree \n> ( \"script_id\" \"int4_ops\" );\n> CREATE TRIGGER \"sub_mag_trigger\" BEFORE INSERT OR UPDATE ON \"subscriptions\" \n> FOR EACH ROW EXECUTE PROCEDURE \n> check_primary_key ('mag_id', 'magazines', 'mag_id');\n> \n> create function\n> ------------\n> CREATE FUNCTION \"check_primary_key\" ( ) RETURNS opaque AS\n> '/usr/local/pgsql/lib/modules/refint.so' LANGUAGE 'C';\n> \n> \n> -- \n> \n> Michael Fraley\n> [email protected]\n> http://www.usfca.edu/~fraley\n> \n\n\n\n", "msg_date": "Tue, 27 Jun 2000 23:48:36 +0200 (CEST)", "msg_from": "Peter Vazsonyi <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Please help cache lookup failed" }, { "msg_contents": "Michael Fraley <[email protected]> writes:\n> I just upgraded from 6.5.2 to 7.0.2 on SuSE 6.3 Linux kernel 2.2.13, and\n> now I get this message when trying to insert records:\n\n> ERROR: fmgr_info: function 19104: cache lookup failed\n\nDid you perhaps delete and recreate check_primary_key() without\nalso recreating the triggers that refer to it?\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 27 Jun 2000 17:49:04 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Please help cache lookup failed " } ]
[ { "msg_contents": "> > We could create /data/databases/DATABASEOID/ and create \n> > soft-links to table-files. This way different tables of\n> > the same database could be in different tablespaces.\n> > /data/database path would be used in production\n> > and /data/tablespace path would be used in recovery.\n> \n> Why would you want to do it that way? Having a different access path\n> for recovery than for normal operation strikes me as just asking for\n> trouble ;-)\n\nI just think that *databases* (schemas) must be used for *logical* groupping\nof tables, not for *physical* one. \"Where to store table\" is tablespace'\nrelated kind of things!\n\n> The symlinks wouldn't do any good for what Bruce had in mind anyway\n> (IIRC, he wanted to get useful per-database numbers from \"du\").\n\nImho, ability to put different tables/indices (of the same database)\nto different tablespaces (disks) is much more useful then ability to\nuse du/ls for administration purposes -:)\n\nAlso, I think that we *must* go away from OS' driven disk space\nallocation anyway. Currently, the way we extend table files breaks WAL\nrule (nothing must go to disk untill logged). + we have to move tuples\nfrom end of file to top to shrink relation - not perfect way to reuse\nempty space. +... +... +...\n\nVadim\n", "msg_date": "Mon, 26 Jun 2000 18:42:10 -0700", "msg_from": "\"Mikheev, Vadim\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Big 7.1 open items " } ]
[ { "msg_contents": "I see mention in SQL99 of function definitions which can have IN, OUT,\nand INOUT arguments. Any thoughts on how this could be supported in\nPostgres? I assume that we would need to figure out how to generate\nmulti-element tuples from the usual function calls, but wouldn't this\ndovetail with getting better SET return capabilities too?\n\n - Thomas\n", "msg_date": "Tue, 27 Jun 2000 02:41:18 +0000", "msg_from": "Thomas Lockhart <[email protected]>", "msg_from_op": true, "msg_subject": "SQL99 functions" }, { "msg_contents": "Thomas Lockhart <[email protected]> writes:\n> I see mention in SQL99 of function definitions which can have IN, OUT,\n> and INOUT arguments. Any thoughts on how this could be supported in\n> Postgres?\n\nI noticed that but haven't quite figured out how it's supposed to fit\ninto the SQL worldview at all. Surely\n\tSELECT foo(x) FROM table\nshouldn't silently mutate into an UPDATE depending on how foo() is\ndeclared. Exactly where is a function with OUT args useful in SQL?\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 27 Jun 2000 04:51:15 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: SQL99 functions " }, { "msg_contents": "> > I see mention in SQL99 of function definitions which can have IN, \n> > OUT, and INOUT arguments. Any thoughts on how this could be \n> > supported in Postgres?\n> I noticed that but haven't quite figured out how it's supposed to fit\n> into the SQL worldview at all. Surely\n> SELECT foo(x) FROM table\n> shouldn't silently mutate into an UPDATE depending on how foo() is\n> declared. Exactly where is a function with OUT args useful in SQL?\n\n create table t1 (x int);\n create function foo (out int) returns int as ...;\n select foo(x) from t1;\n\nwill give two columns for the result.\n\n create function foo (inout int) returns int as ...;\n select foo(x) from t1;\n\nwill mutate the result, but not the underlying stored value of t1.x.\n\nBeware, I haven't yet confirmed this by reading ;)\n\n - Thomas\n", "msg_date": "Tue, 27 Jun 2000 13:45:29 +0000", "msg_from": "Thomas Lockhart <[email protected]>", "msg_from_op": true, "msg_subject": "Re: SQL99 functions" }, { "msg_contents": "Thomas Lockhart <[email protected]> writes:\n> create table t1 (x int);\n> create function foo (out int) returns int as ...;\n> select foo(x) from t1;\n\n> will give two columns for the result.\n\nYou've *got* to be kidding.\n\nTo name just one problem with that, where do I put an AS to relabel\nthe extra column?\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 27 Jun 2000 10:14:37 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: SQL99 functions " } ]
[ { "msg_contents": "Not knowing much on this side of things but in JDBC, CallableStatement\nmentions things like IN, OUT & INOUT args - not sure about INOUT, but IN &\nOUT are there. Perhaps OUT are not valid for selects, but are in stored\nprocedures?\n\nPeter\n\n--\nPeter Mount\nEnterprise Support\nMaidstone Borough Council\nAny views stated are my own, and not those of Maidstone Borough Council\n\n\n-----Original Message-----\nFrom: Tom Lane [mailto:[email protected]]\nSent: Tuesday, June 27, 2000 9:51 AM\nTo: Thomas Lockhart\nCc: Hackers List\nSubject: Re: [HACKERS] SQL99 functions \n\n\nThomas Lockhart <[email protected]> writes:\n> I see mention in SQL99 of function definitions which can have IN, OUT,\n> and INOUT arguments. Any thoughts on how this could be supported in\n> Postgres?\n\nI noticed that but haven't quite figured out how it's supposed to fit\ninto the SQL worldview at all. Surely\n\tSELECT foo(x) FROM table\nshouldn't silently mutate into an UPDATE depending on how foo() is\ndeclared. Exactly where is a function with OUT args useful in SQL?\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 27 Jun 2000 10:13:14 +0100", "msg_from": "Peter Mount <[email protected]>", "msg_from_op": true, "msg_subject": "RE: SQL99 functions " }, { "msg_contents": "Peter Mount wrote:\n> \n> Not knowing much on this side of things but in JDBC, CallableStatement\n> mentions things like IN, OUT & INOUT args - not sure about INOUT, but IN &\n> OUT are there. Perhaps OUT are not valid for selects, but are in stored\n> procedures?\n> \n> Peter\n> \n> --\n> Peter Mount\n> Enterprise Support\n> Maidstone Borough Council\n> Any views stated are my own, and not those of Maidstone Borough Council\n> \n> -----Original Message-----\n> From: Tom Lane [mailto:[email protected]]\n> Sent: Tuesday, June 27, 2000 9:51 AM\n> To: Thomas Lockhart\n> Cc: Hackers List\n> Subject: Re: [HACKERS] SQL99 functions\n> \n> Thomas Lockhart <[email protected]> writes:\n> > I see mention in SQL99 of function definitions which can have IN, OUT,\n> > and INOUT arguments. Any thoughts on how this could be supported in\n> > Postgres?\n> \n> I noticed that but haven't quite figured out how it's supposed to fit\n> into the SQL worldview at all. Surely\n> SELECT foo(x) FROM table\n> shouldn't silently mutate into an UPDATE depending on how foo() is\n> declared. Exactly where is a function with OUT args useful in SQL?\n\nThey are used (at least) in PL/SQL nad other PLs. \n\nTo return multiple values OUT args are one possible solution.\n\nAlso one could have session variables possibly like this:\n\nDECLARE VARIABLE MY_VAR INT;\n\nEXECUTE MY_PROC(MY_VAR);\n\nSELECT * FROM SOME_TABLE WHERE ID=MY_VAR;\n\n-----------\nHannu\n", "msg_date": "Tue, 27 Jun 2000 13:06:40 +0300", "msg_from": "Hannu Krosing <[email protected]>", "msg_from_op": false, "msg_subject": "Re: SQL99 functions" } ]
[ { "msg_contents": "comments prefixed with PM...\n\n--\nPeter Mount\nEnterprise Support\nMaidstone Borough Council\nAny views stated are my own, and not those of Maidstone Borough Council\n\n\n-----Original Message-----\nFrom: Philip Warner [mailto:[email protected]]\nSent: Tuesday, June 27, 2000 10:07 AM\nTo: Giles Lean\nCc: Zeugswetter Andreas SB; [email protected]\nSubject: Re: [HACKERS] Proposal: More flexible backup/restore via\npg_dump \n\n\nAt 07:00 27/06/00 +1000, Giles Lean wrote:\n>\n>Are you are also assuming that a backup fits in a single file,\n>i.e. that anyone with >2GB of backup has some sort of large file\n>support?\n\nThat's up to the format used to save the database; in the case of the\n'custom' format, yes. But that is the size after compression. This is not\nsubstantially different to pg_dump's behaviour, except that pg_dump can be\npiped to a tape drive...\n\nPM: So can most other Unix based formats. On the intranet server here, I\npg_dump into /tmp, then include them in a tar piped to the tape drive.\n\nThe objective of the API components are to (a) make it very easy to add new\nmetadata to dump (eg. tablespaces), and (b) make it easy to add new output\nformats (eg. tar archives). Basically the metadata dumping side makes one\ncall to register the thing to be saved, passing an optional function\npointer to dump data (eg. table contents) - this *could* even be used to\nimplement dumping of BLOBs.\n\nPM: The problem with blobs hasn't been with dumping them (I have some Java\ncode that does it into a compressed zip file), but restoring them - you\ncan't create a blob with a specific OID, so any references in existing\ntables will break. I currently get round it by updating the tables after the\nrestore - but it's ugly and easy to break :(\n\nThe 'archiver' format provider must have some basic IO routines:\nRead/WriteBuf and Read/WriteByte and has a number of hook functions which\nit can use to output the data. It needs to provide at least one function\nthat actually writes data somewhere. It also has to provide the associated\nfunction to read the data.\n\nPM: Having a set of api's (either accessible directly into the backend,\nand/or via some fastpath call) would be useful indeed.\n\n>\n>As someone else answered: no. You can't portably assume random access\n>to tape blocks.\n\nThis is probably an issue. One of the motivations for this utility it to\nallow partial restores (eg. table data for one table only), and\narbitrarilly ordered restores. But I may have a solution:\n\nPM: That would be useful. I don't know about CPIO, but tar stores the TOC at\nthe start of each file (so you can actually join two tar files together and\nstill read all the files). In this way, you could put the table name as the\n\"filename\" in the header, so partial restores could be done.\n\nwrite the schema and TOC out at the start of the file/tape, then compressed\ndata with headers for each indicating which TOC item they correspond to.\nThis metadata can be loaded into /tmp, so fseek is possible. The actual\ndata restoration (assuming constraints are not defined [THIS IS A PROBLEM])\ncan be done by scanning the rest of the tape in it's own order since RI\nwill not be an issue. I think I'm happy with this.\n\nPM: How about IOCTL's? I know that ArcServe on both NT & Unixware can seek\nthrough the tape, so there must be a way of doing it.\n\n[snip]\n\n>Using either tar or cpio format as defined for POSIX would allow a lot\n>of us to understand your on-tape format with a very low burden on you\n>for documentation. (If you do go this route you might want to think\n>about cpio format; it is less restrictive about filename length than\n>tar.)\n\nTom Lane was also very favorably disposed to tar format. As I said above,\nthe archive interfaces should be pretty amenable to adding tar support -\nit's just I'd like to get a version working with custom and directory based\nformats to ensure the flexibility is there. As I see it, the 'backup to\ndirectory' format should be easy to use as a basis for the 'backup to tar'\ncode.\n\nPM: I don't see a problem there. The Java stuff I have use the standard\njava.util.zip package wich has a simple api for zip files. Tar or most other\nformats could be implemented in a similar compatible fashion.\n\nThe problem I have with tar is that it does not support random access to\nthe associated data. For reordering large backups, or (ultimately) single\nBLOB extraction, this is a performance problem.\n\nPM: Tar can do this sequentially, which I've had to do many times over the\nyears - restoring just one file from a tape, sequential access is probably\nthe only way.\n\nIf you have a tar spec (or suitably licenced code), please mail it to me,\nand I'll be able to make more informed comments.\n\nPM: The tar spec should be around somewhere - just be careful, the obvious\nsource I was thinking of would be GPL'd, and we don't want to be poluted :-)\n\n[snip]\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 27 Jun 2000 10:55:28 +0100", "msg_from": "Peter Mount <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Proposal: More flexible backup/restore via pg_dump " } ]
[ { "msg_contents": "Please HELP!\nHi. All I'm new with PostgreSQL. Can't find no more docs on PL/pgSQL except\nthose on postgresql.org in \"programmers guide\". Does anyone know any links.\nPlease HELP!\n\ngrishick\[email protected]\n\n", "msg_date": "Tue, 27 Jun 2000 14:41:25 +0400", "msg_from": "\"Grigori Soloviov\" <[email protected]>", "msg_from_op": true, "msg_subject": "Any Documentation on PL/pgSQL?" } ]
[ { "msg_contents": "\n\n--\nPeter Mount\nEnterprise Support\nMaidstone Borough Council\nAny views stated are my own, and not those of Maidstone Borough Council\n\n\n-----Original Message-----\nFrom: Philip Warner [mailto:[email protected]]\nSent: Tuesday, June 27, 2000 11:31 AM\nTo: Peter Mount; Giles Lean\nCc: Zeugswetter Andreas SB; [email protected]\nSubject: RE: [HACKERS] Proposal: More flexible backup/restore via\npg_dump \n\n\nAt 10:55 27/06/00 +0100, Peter Mount wrote:\n>comments prefixed with PM...\n>\n>PM: So can most other Unix based formats. On the intranet server here, I\n>pg_dump into /tmp, then include them in a tar piped to the tape drive.\n\nSince you are using an intermediate file, there would be no change. fseek\nis only an issue on tape drives and pipes, not files on disk. I suspect\nthat most people can afford the overhead of restoring the backup file to\ndisk before restoring the database, but it'd be nice to start out as\nflexible as possible. In the back of my mind is the fact that when the WAL\nand storage manager are going, raw data backups should be possible (and\nfast) - but then again, maybe it's a pipe dream.\n\nPM: Actually, with MS-SQL here, I use it's \"Backup Device\" (M$'s name for a\nfile ;-) ). It then has a scheduled job that backs up into that file (a full\nbackup empties the file first, then the incremental's append to it). Another\nserver then just backsup this single file.\n\nPM: Anyhow, when I need to do a restore, I'm presented with a list of what\nbackup's are stored on that \"device\", and can restore from there.\n\n>PM: The problem with blobs hasn't been with dumping them (I have some Java\n>code that does it into a compressed zip file), but restoring them - you\n>can't create a blob with a specific OID, so any references in existing\n>tables will break. I currently get round it by updating the tables after\nthe\n>restore - but it's ugly and easy to break :(\n\nI assumed this would have to happen - hence why it will not be in the first\nversion. With all the TOAST stuff coming, and talk about the storage\nmanager, I still live in hope of a better BLOB system...\n\nPM: I do as well. I've got a project coming up soon where the database will\nbe storing thumbnails of photographs (with details of where the full images\nare stored, what film/negative# the original is from (for 35mm), notes, etc.\nAnyhow, the current BLOB system may not cope with it, and the kludges I've\nused in the past probably won't help.\n\n>PM: Having a set of api's (either accessible directly into the backend,\n>and/or via some fastpath call) would be useful indeed.\n\nBy this I assume you mean APIs to get to database data, not backup data.\n\nPM: I was thinking of the backup data (in a similar way you can with SQL7).\nA user could (if they have permissions) trigger a backup or a restore.\n\neg, my log database here is backed up using:\n\tBACKUP DATABASE eventlog TO eventdb WITH INIT, NAME=N'Eventlog\nBackup', DESCRIPTION=N'The NT & Exchange Event Logs'\n\nHere, eventlog is the database name, eventdb the \"device\", ie the file it\nwrites to, INIT erases the \"device\" and NAME/DESCRIPTION are written to the\ndevice to help locate the backup when restoring.\n\nAnyhow, this is possible future stuff ;-)\n\n>\n>This is probably an issue. One of the motivations for this utility it to\n>allow partial restores (eg. table data for one table only), and\n>arbitrarilly ordered restores. But I may have a solution:\n>\n>PM: That would be useful. I don't know about CPIO, but tar stores the TOC\nat\n>the start of each file (so you can actually join two tar files together and\n>still read all the files). In this way, you could put the table name as the\n>\"filename\" in the header, so partial restores could be done.\n\nWell, the way my plans work, I'll use either a munged OID, or a arbitrary\nunique ID as the file name. All meaningful access has to go via the TOC.\nBut that's just a detail; the basic idea is what I'm implementing. \n\nIt's very tempting to say tape restores are only possible in the order in\nwhich the backup file was written ('pg_restore --stream' ?), and that\nreordering/selection is only possible if you put the file on disk.\n\nPM: It's an idea for those with large (>=2Gb) databases. Most filesystems\ndon't like an individual file to be larger (or equal) to 2Gb. You may want\nto keep this in mind when writing to a file (ie, break at 1Gb as we do with\ntables).\n\n>\n>PM: How about IOCTL's? I know that ArcServe on both NT & Unixware can seek\n>through the tape, so there must be a way of doing it.\n\nMaybe; I know BackupExec also does some kind of seek to update the TOC at\nend of a backup (which is what I need to do). Then again, maybe that's just\na rewind. I don't want to get into custom tape formats...\n\nPM: Some tape drives support a \"mark\", but I've not seen/used one. When\nwatching ArcServe, it seems to rewind to the beginning, then \"seek\" to the\nbegining of the backup. But once it's done that, it knows where it is, and\nseeks forward from there.\n\nDo we have any tape experts out there?\n\n\n>PM: Tar can do this sequentially, which I've had to do many times over the\n>years - restoring just one file from a tape, sequential access is probably\n>the only way.\n\nIt's just nasty when the user reorders the restoration of tables and\nmetadata. In the worst cast it might be hundreds of scans of the tape. I'd\nhate to have my name associated with something so unfriendly (even if it is\nthe operators fault).\n\nPM: That's true, but sometimes (and it happens here a lot), it's\nunavoidable.\n\n>PM: The tar spec should be around somewhere - just be careful, the obvious\n>source I was thinking of would be GPL'd, and we don't want to be poluted\n:-)\n\nThat was my problem. I've got some references now, and I'll look at them.\nAt least everything I've written so far can be put in PG.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 27 Jun 2000 12:38:40 +0100", "msg_from": "Peter Mount <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Proposal: More flexible backup/restore via pg_dump " } ]
[ { "msg_contents": "\n> That'd work fine for me, but I think Bruce was arguing for paths that\n> included the database name. We'd end up with paths that go something\n> like\n> \t..../data/tablespaces/TABLESPACEOID/RELATIONOID\n> (plus some kind of decoration for segment and version), so you'd have\n> a hard time telling which files in a tablespace belong to which\n> database.\n\nWell ,as long as we have the file per object layout it probably makes sense\nto \nhave \"speaking paths\", But I see no real problem with:\n\n..../data/tablespacename/dbname/RELATIONOID[.dat|.idx]\n\nRELATIONOID standing for whatever the consensus will be.\nI do not really see an argument for using a tablespaceoid instead of\nit's [maybe mangled] name.\n\nAndreas\n", "msg_date": "Tue, 27 Jun 2000 15:27:03 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: Big 7.1 open items " }, { "msg_contents": "Zeugswetter Andreas SB <[email protected]> writes:\n> I do not really see an argument for using a tablespaceoid instead of\n> it's [maybe mangled] name.\n\nEliminating filesystem-based restrictions on names, for one.\nFor example we'd not have to forbid slashes and (probably) backquotes\nin tablespace names if we did this, and we'd not have to worry about\nfilesystem-induced limits on name lengths. Renaming a tablespace\nwould also be trivial instead of nigh impossible.\n\nIt might be that using tablespace names as directory names is worth\nenough from the admin point of view to make the above restrictions \nacceptable. But it's a tradeoff, and not one with an obvious choice\nIMHO.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 27 Jun 2000 10:23:48 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Big 7.1 open items " } ]
[ { "msg_contents": "\n> Thomas Lockhart <[email protected]> writes:\n> > I see mention in SQL99 of function definitions which can \n> have IN, OUT,\n> > and INOUT arguments. Any thoughts on how this could be supported in\n> > Postgres?\n> \n> I noticed that but haven't quite figured out how it's supposed to fit\n> into the SQL worldview at all. Surely\n> \tSELECT foo(x) FROM table\n> shouldn't silently mutate into an UPDATE depending on how foo() is\n> declared. Exactly where is a function with OUT args useful in SQL?\n\nThis is something Oracle pushed through, because that is how they do it.\nI prefer the usual way of doing such things where you have parameters\nand return values to functions (but return values should be multi column \nand multi row capable).\n\nAndreas\n", "msg_date": "Tue, 27 Jun 2000 15:39:13 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: SQL99 functions " } ]
[ { "msg_contents": "\nI have done several searches in the mailing list archive, and all come up\nwith no results (eg. search hackers for vacuum). There seems to be a problem.\n\nA user on the SQL list just reported that they could not find 'backup' in\nin admin, general and sql.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 27 Jun 2000 23:39:41 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Mailing List Archive Problem?" }, { "msg_contents": "On Tue, 27 Jun 2000, Philip Warner wrote:\n\n> \n> I have done several searches in the mailing list archive, and all come up\n> with no results (eg. search hackers for vacuum). There seems to be a problem.\n> \n> A user on the SQL list just reported that they could not find 'backup' in\n> in admin, general and sql.\n\nUpgraded to 3.1.0 of UDMSearch the other day, and it broke ... sent in a\nreport to the developers, got told that 3.1.1 would have it fixed but\ncan't get anything out of them as to when a v3.1.1 will be packaged ...\n\nI just back tracked to 3.0.18 and am rebuilding those indices now ...\n\nSorry guys, just got too used to our development practices where fixes\ngo out faster, figured I'd wait it out a day or two ;(\n\n\n", "msg_date": "Tue, 27 Jun 2000 22:18:54 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Mailing List Archive Problem?" } ]
[ { "msg_contents": "\n> Thomas Lockhart <[email protected]> writes:\n> > create table t1 (x int);\n> > create function foo (out int) returns int as ...;\n> > select foo(x) from t1;\n> \n> > will give two columns for the result.\n> \n> You've *got* to be kidding.\n> \n> To name just one problem with that, where do I put an AS to relabel\n> the extra column?\n\nIt gets the name of the variable ( here \"out\"), no way to rename it,\nbut no problem since you can name the variable however you like.\n\nAndreas \n", "msg_date": "Tue, 27 Jun 2000 16:53:30 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: SQL99 functions " }, { "msg_contents": "> > > create function foo (out int) returns int as ...;\n> > > select foo(x) from t1;\n> > > will give two columns for the result.\n> > You've *got* to be kidding.\n> > To name just one problem with that, where do I put an AS to relabel\n> > the extra column?\n> It gets the name of the variable ( here \"out\"), no way to rename it,\n> but no problem since you can name the variable however you like.\n\nActually, there is a third (optional) field which is the variable name:\n\n create function foo (out pname int) returns int as ...;\n\nNot sure what it should default to if the name is not specified.\n\n - Thomas\n", "msg_date": "Tue, 27 Jun 2000 15:07:31 +0000", "msg_from": "Thomas Lockhart <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: SQL99 functions" }, { "msg_contents": "Zeugswetter Andreas SB <[email protected]> writes:\n>> Thomas Lockhart <[email protected]> writes:\n>>>> create table t1 (x int);\n>>>> create function foo (out int) returns int as ...;\n>>>> select foo(x) from t1;\n>> \n>>>> will give two columns for the result.\n>> \n>> You've *got* to be kidding.\n>> \n>> To name just one problem with that, where do I put an AS to relabel\n>> the extra column?\n\n> It gets the name of the variable ( here \"out\"), no way to rename it,\n> but no problem since you can name the variable however you like.\n\nWhat? OUT is a keyword there; I don't see any explicit name for the\nfunction's formal parameter at all.\n\n\nBTW, what happens when I use foo() in a place other than the top level\nof a SELECT list?\n\nExamples:\n\n\tSELECT (CASE WHEN x > 0 THEN foo(x) ELSE 42 END) FROM table\n\n\tSELECT x FROM table WHERE foo(x) > 0\n\n\nAlso, I still haven't figured out what INOUT is supposed to do.\n\n\nThere are some things that are too brain-dead to consider implementing,\nwhether they are in SQL99 or not, and this is sounding a lot like\none of them...\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 27 Jun 2000 11:15:01 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: SQL99 functions " } ]
[ { "msg_contents": "Hello all,\n\nThis is the result of my 6 days \"war\" with Linux kernel people...\nAny usefull comments on this????\n\n---------- Forwarded Message ----------\nSubject: Re: Fwd: Problem with recv syscall on socket when other side closed connection\nDate: Tue, 27 Jun 2000 16:21:55 +0400 (MSK DST)\nFrom: [email protected]\n\nHello!\n\n> Sorry... But seems that you did not understand the problem.\n> I talk about recv... Not write... write SHOULD give EPIPE on connection reset...\n> But not recv/read.\n\nI did understand. This error was for write(), but it became known\n_after_ you exited write(). So that it is delivered to read().\nIt is usual problem of all full-duplex pipes.\n\nWe could translate this EPIPE to ECONNRESET, when it is delivered\nto read(), but it does not change its sense.\n\nSolaris does not translate.\n\n\n> Usual way of handling connection reset when you do only read is to give\n> all data available and then return 0, indicating EOF.\n\nSorry? Think a bit.\n\nYou wrote to dead socket, right? It is the hardest error.\nIf the transport were local, you would get SIGPIPE and died painful death.\nIf an OS ignores such events, it is simply impossible to use,\nyou will get silently truncated data all the time.\n\n\n> Or some OSes (HPUX if I'm not mistaken) gives you all data available and then\n> ECONNRESET. But not other way around...\n\nThis approach has its merits, and it is acceptable in principle.\n\nBut Linux approach is evidently better, because errors are expedited.\nEach protocol, where out of band events are inlined to data\nis inclined to deadlocks.\n\nIn Linux scheme you know forward that stream is aborted.\nDepending on protocol you may choose to abort protocol\nor to continue to operate, parsing already received messages.\n\n\n> \t\tBut not other way around...\n\nYou have just seen a new way around. The correct one. 8)\n\nAlexey\n-------------------------------------------------------\n\n-- \nSincerely Yours,\nDenis Perchine\n\n----------------------------------\nE-Mail: [email protected]\nHomePage: http://www.perchine.com/dyp/\nFidoNet: 2:5000/120.5\n----------------------------------\n", "msg_date": "Tue, 27 Jun 2000 23:12:59 +0700", "msg_from": "Denis Perchine <[email protected]>", "msg_from_op": true, "msg_subject": "Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "Denis Perchine <[email protected]> writes:\n> This is the result of my 6 days \"war\" with Linux kernel people...\n> Any usefull comments on this????\n\nThe Linux people seem to be assuming (erroneously) that application\nprotocols are strictly I-send-and-then-you-send. That's too\nrestrictive, and in fact it falls down in exactly the case that\nwe're seeing in libpq: one side of the connection may send an\nerror message \"out of turn\" and then close the connection. If the\nother side of the connection was busy sending, the first thing it\nwill get is an EPIPE error on its send. By closing the connection\n*AND DISCARDING VALID USER DATA* at this point, the Linux kernel\nmakes it impossible to retrieve the error message --- which might\nhave contained essential information.\n\n> In Linux scheme you know forward that stream is aborted.\n> Depending on protocol you may choose to abort protocol\n> or to continue to operate, parsing already received messages.\n\nBut what about the messages you didn't get yet, but the other end\nsent in good faith? There's nothing in the TCP specs that says\na program can't close its end of the connection as soon as it has\nsent the last data it intends to send.\n\n>> But not other way around...\n\n> You have just seen a new way around. The correct one. 8)\n\nNo, just a new half-baked excuse for doing things wrong. The kernel at\nthe other side of the connection accepted the data for delivery. That\nmeans that both sides of the connection are going to make their best\nefforts to deliver it. By willfully failing to deliver that data, the\nLinux kernel is violating the fundamental premise of TCP (or any other\nreliable byte-stream protocol). This is not \"correct\", it is broken.\nDo I need to quote RFC chapter and verse at you?\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 27 Jun 2000 17:48:04 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "Hello!\n\n> > This is the result of my 6 days \"war\" with Linux kernel people...\n\nAlas, I did not understand that he was in the state of war.\nI believed he asked for an advice and/or reported a bug. 8)\n\n\n> *AND DISCARDING VALID USER DATA* at this point, the Linux kernel\n> makes it impossible to retrieve the error message --- which might\n> have contained essential information.\n\nDenis was explained that Linux _never_ discards any data...\nOugh, I beg pardon: Denis noticed this _himself_.\nPlease, do not desinform people.\n\nBTW look at this. It is RFC1122, 4.2.2.13. \n\n \t\t\t\t\t\tIf a TCP\n connection is closed by the remote site, the local\n application MUST be informed whether it closed normally or\n was aborted.\n\nSee? 8)\n\n\nAlso, David Miller explained (and Denis understood and accepted this)\nthat TCP does not guarantee data delivery, when an RST is issued.\nIt is solely due to intrinsic network unreliability and segment reordering.\nThe application depending on these aspects is broken.\n\n\n> But what about the messages you didn't get yet, but the other end\n> sent in good faith? There's nothing in the TCP specs that says\n> a program can't close its end of the connection as soon as it has\n> sent the last data it intends to send.\n\nTCP specs say directly and unabiguously, that sending data\nto half-closed pipe is followed by immediate abort (RFC1122, a bit\nbelow the cite above). More detailed explanation can be found\nin current draft-ietf-tcpimpl.\n\n> Do I need to quote RFC chapter and verse at you?\n\nOf course.\n\nAlexey\n", "msg_date": "Wed, 28 Jun 2000 20:24:24 +0400 (MSK DST)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "[ Sorry for delay in response, I had other things to do over the\nweekend. ]\n\[email protected] writes:\n> BTW look at this. It is RFC1122, 4.2.2.13. \n\n> \t\t\t\t\t\tIf a TCP\n> connection is closed by the remote site, the local\n> application MUST be informed whether it closed normally or\n> was aborted.\n\nSo? This is not relevant, because the connection was not aborted.\nThe sentence immediately preceding that one defines an abort as an event\nin which RST segment(s) are sent, but closure of a connection is defined\nto send FIN, not RST. (More about that below.)\n\nThe more relevant quote is the next paragraph,\n\n The normal TCP close sequence delivers buffered data\n reliably in both directions. Since the two directions of a\n TCP connection are closed independently, it is possible for\n a connection to be \"half closed,\" i.e., closed in only one\n direction, and a host is permitted to continue sending data\n in the open direction on a half-closed connection.\n\nI do not see how you can read the first sentence of that paragraph in\nany way but to say that data once sent must be delivered if at all\npossible. Another example is from RFC-793 (STD-7), section 3.8,\ndefinition of CLOSE:\n\n Closing connections is intended to be a graceful operation in\n the sense that outstanding SENDs will be transmitted (and\n retransmitted), as flow control permits, until all have been\n serviced. Thus, it should be acceptable to make several SEND\n calls, followed by a CLOSE, and expect all the data to be sent\n to the destination.\n\nIn our situation, the server sends (queues) some data and then closes\nits side of the connection. The server-side TCP stack should send the\ndata along with FIN and then go to FIN-WAIT-1 state. In this state\nthe server side may receive more data from the client side (since the\nclient isn't yet aware the server has quit). RFC-793 is perfectly\nclear that the server side must send a dummy ACK but *no* RST in this\ncase --- see section 3.4, almost the end of the section:\n\n 3. If the connection is in a synchronized state (ESTABLISHED,\n FIN-WAIT-1, FIN-WAIT-2, CLOSE-WAIT, CLOSING, LAST-ACK, TIME-WAIT),\n any unacceptable segment (out of window sequence number or\n unacceptible acknowledgment number) must elicit only an empty\n acknowledgment segment containing the current send-sequence number\n and an acknowledgment indicating the next sequence number expected\n to be received, and the connection remains in the same state.\n\nTherefore, sending data to a no-longer-present receiver does not cause\na connection reset (at least not in a spec-conforming TCP stack), and\nthere is no justification for discarding data that is coming the other\nway.\n\nThe Linux kernel's present behavior is contrary to the standard, unable\nto support an essential user capability (ie, delivery of last-gasp error\nmessages), and contrary to the behavior of all other TCP implementations\nthat I have worked with. There is a reason why you are in the minority\nhere...\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 18:34:37 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "> 3. If the connection is in a synchronized state (ESTABLISHED,\n> FIN-WAIT-1, FIN-WAIT-2, CLOSE-WAIT, CLOSING, LAST-ACK, TIME-WAIT),\n> any unacceptable segment (out of window sequence number or\n> unacceptible acknowledgment number) must elicit only an empty\n> acknowledgment segment containing the current send-sequence number\n> and an acknowledgment indicating the next sequence number expected\n> to be received, and the connection remains in the same state.\n> \n> Therefore, sending data to a no-longer-present receiver does not cause\n> a connection reset (at least not in a spec-conforming TCP stack), and\n> there is no justification for discarding data that is coming the other\n> way.\n> \n> The Linux kernel's present behavior is contrary to the standard, unable\n> to support an essential user capability (ie, delivery of last-gasp error\n> messages), and contrary to the behavior of all other TCP implementations\n> that I have worked with. There is a reason why you are in the minority\n> here...\n\nReread the 3. above. What it actually requires if you think about it is that\nthe receive window is shrunk to zero and the connection hangs for all\neternity the way you are arguing it.\n\nAlan\n\n", "msg_date": "Mon, 3 Jul 2000 23:41:30 +0100 (BST)", "msg_from": "Alan Cox <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "Alan Cox <[email protected]> writes:\n>> 3. If the connection is in a synchronized state (ESTABLISHED,\n>> FIN-WAIT-1, FIN-WAIT-2, CLOSE-WAIT, CLOSING, LAST-ACK, TIME-WAIT),\n>> any unacceptable segment (out of window sequence number or\n>> unacceptible acknowledgment number) must elicit only an empty\n>> acknowledgment segment containing the current send-sequence number\n>> and an acknowledgment indicating the next sequence number expected\n>> to be received, and the connection remains in the same state.\n\n> Reread the 3. above. What it actually requires if you think about it is that\n> the receive window is shrunk to zero and the connection hangs for all\n> eternity the way you are arguing it.\n\nNo, it doesn't \"hang for all eternity\", it sits in the same state until\n(a) the client side closes its sending side of the connection (ie, sends\nFIN), or (b) the FIN-WAIT-1 state times out. But given a normally\nresponsive client and no loss of physical connectivity or crash of\neither TCP stack, there is no connection reset and no failure to deliver\nsent data.\n\nThere would be no need for all the half-open-connection verbiage if the\nspec were meant to be read the way you are reading it.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 18:52:51 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "> No, it doesn't \"hang for all eternity\", it sits in the same state until\n> (a) the client side closes its sending side of the connection (ie, sends\n> FIN), or (b) the FIN-WAIT-1 state times out. But given a normally\n> responsive client and no loss of physical connectivity or crash of\n> either TCP stack, there is no connection reset and no failure to deliver\n> sent data.\n\nI cannot ack the data since it has not been read, so how can I ack the fin ?\n", "msg_date": "Tue, 4 Jul 2000 01:05:49 +0100 (BST)", "msg_from": "Alan Cox <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "Alan Cox <[email protected]> writes:\n>> No, it doesn't \"hang for all eternity\", it sits in the same state until\n>> (a) the client side closes its sending side of the connection (ie, sends\n>> FIN), or (b) the FIN-WAIT-1 state times out. But given a normally\n>> responsive client and no loss of physical connectivity or crash of\n>> either TCP stack, there is no connection reset and no failure to deliver\n>> sent data.\n\n> I cannot ack the data since it has not been read, so how can I ack the fin ?\n\nACK does not mean that you've delivered the data to the end user.\nRFC 793, 2.6:\n\n An acknowledgment by TCP does not guarantee that the data has been\n delivered to the end user, but only that the receiving TCP has taken\n the responsibility to do so.\n\nBit-bucketing the data because the end user app is no longer present to\naccept it (due to having already closed its input socket) is implicitly\nwithin the receiving TCP's authority here. I think this is the core of\nour disagreement, but I can see no justification for your position that\nACK implies the data has been delivered to the end user. Every TCP\nimplementation I've ever heard of sends ACK as soon as it's collected\ndata into kernel buffers, *not* after the application has executed\nrecv() to extract the data from the kernel. (Who's to say that\ncompletion of recv() represents final delivery of the data anyway?\nSending ACK cannot be considered a report of end-to-end delivery;\nthat has to be an application-level concept.)\n\nAlso observe that the discussion of segment-arrival processing in\nsection 3.9 explicitly says that the behavior in FIN-WAIT-1 and later\nstates is not different from the behavior in ESTABLISHED state. In\nparticular, if you do not like the segment:\n\n If an incoming segment is not acceptable, an acknowledgment\n should be sent in reply (unless the RST bit is set, if so drop\n the segment and return):\n\n <SEQ=SND.NXT><ACK=RCV.NXT><CTL=ACK>\n\n After sending the acknowledgment, drop the unacceptable segment\n and return.\n\nThere is no room here for the TCP to decide to send RST instead.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 20:47:18 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "Hello!\n\n> <SEQ=SND.NXT><ACK=RCV.NXT><CTL=ACK>\n> \n> After sending the acknowledgment, drop the unacceptable segment\n> and return.\n> \n> There is no room here for the TCP to decide to send RST instead.\n\nI apologize, but RFC793 is sort of incomplete. Please, look at\nerrata in RFC1122 and to bug alerts described in documents published\nby tcp-impl (draft-tcpimpl-*).\n\n\nI cited you corresponding paragraph of the RFC in previous mail.\nShortly:\n\n1. When new data arrive after half-duplex close, we must reset.\n2. When close occurs on connection, which has unread data, we\n must reset.\n\nIt is required from the viewpoint of TCP protocol. Any OS, which\nforgets to make this is buggy. By the way, I do not know about OSes,\nwhich do not make this.\n\n From the viewpoint of application, the behaviour is also correct.\nData arrived, when nobody plans to read it, unambiguously means\neither connection abort or hard bug in application protocol.\n\nAlexey\n", "msg_date": "Tue, 4 Jul 2000 20:04:41 +0400 (MSK DST)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "[email protected] writes:\n> I apologize, but RFC793 is sort of incomplete. Please, look at\n> errata in RFC1122 and to bug alerts described in documents published\n> by tcp-impl (draft-tcpimpl-*).\n\nThe errata in RFC 1122 do not recommend any changes in connection\nclosure behavior. You appear to be hanging your hat on this paragraph\nfrom 1122 4.2.2.13:\n\n A host MAY implement a \"half-duplex\" TCP close sequence, so\n that an application that has called CLOSE cannot continue to\n read data from the connection. If such a host issues a\n CLOSE call while received data is still pending in TCP, or\n if new data is received after CLOSE is called, its TCP\n SHOULD send a RST to show that data was lost.\n\nHowever I read this as a requirement pertaining only to half-duplex\nclose sequences. There is nothing half-duplex about closing a socket\ncompletely. In any case, it can hardly be a good idea to abort the\nflow of data in the outbound direction in order to report that data\nis being dropped in the inbound direction. If an application has done\na half-close to close its inbound side only, but wants to keep sending\noutbound data, it presumably has a good reason for doing so. Behaving\nas you suggest would render this mode of operation useless.\n\nAs for the drafts, I assume you are referring to sections 2.16 and 2.17\nof RFC 2525 --- I couldn't find anything about connection resets in\nthe other files ftp://ftp.isi.edu/internet-drafts/draft-ietf-tcpimpl-*.\nMay I remind you that 2525 is an informational RFC, not a\nstandards-track RFC, and accordingly it has not been reviewed to the\nextent that a proposed standards change would be?\n\nI shall be writing to the authors of 2525 to object to sections 2.16\nand 2.17 on the grounds that an RST causes data loss in the other\ndirection. We'll see what they have to say.\n\n> From the viewpoint of application, the behaviour is also correct.\n> Data arrived, when nobody plans to read it, unambiguously means\n> either connection abort or hard bug in application protocol.\n\nSure, it's a connection abort. My point is that RST is an unacceptably\nblunt instrument for reporting it, because it causes loss of data going\nin the other direction.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 13:46:44 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "Hello!\n\n> blunt instrument for reporting it, because it causes loss of data going\n> in the other direction.\n\nFirst. Data, which reached the host are not lost.\n\nSecond. TCP may lose this data because this data did not reach host\nbefore reset arrived, indeed.\n\nAfter the second we arrive at the next: if you send to dead pipe,\nor do not read some remnant of data before closing, it is _HARD_ bug\nin your application or in protocol.\n\nDo you understand what hard bug is? It is when further behaviour\nis unpredictable and the state cannot be recovered. Essentially,\nit is thing which exceptions and fatal signals are invented for. 8)\n\nAlexey\n", "msg_date": "Tue, 4 Jul 2000 21:57:28 +0400 (MSK DST)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "[email protected] writes:\n>> blunt instrument for reporting it, because it causes loss of data going\n>> in the other direction.\n\n> First. Data, which reached the host are not lost.\n\nAs I recall, the original complaint was precisely that Linux discards\nthe server->client data instead of allowing the client to read it. This\nwas on a single machine, so there's no issue of whether it got lost in\nthe network.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 15:52:30 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "Hello!\n\n> As I recall, the original complaint was precisely that Linux discards\n> the server->client data instead of allowing the client to read it. This\n> was on a single machine, so there's no issue of whether it got lost in\n> the network.\n\nI am sorry. I have already said: it is not truth.\n\nOriginal reporter (Denis) blamed particularly on the fact,\nthat Linux allows to read all queued data until EOF.\nTry yourself, if you do not believe.\n\nUnfortunately, I deleted that his mail, but you can find it\nin mail archives I think, it was to netdev or to linux-kernel.\n\nAlexey\n", "msg_date": "Wed, 5 Jul 2000 19:06:06 +0400 (MSK DST)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: Fwd: Re: Fwd: Problem with recv syscall on socket when other side\n\tclosed connection" }, { "msg_contents": "Hello kuznet,\n\nWednesday, July 05, 2000, 7:06:06 PM, you wrote:\n\nkmiar> Hello!\n\n>> As I recall, the original complaint was precisely that Linux discards\n>> the server->client data instead of allowing the client to read it. This\n>> was on a single machine, so there's no issue of whether it got lost in\n>> the network.\n\nkmiar> I am sorry. I have already said: it is not truth.\n\nkmiar> Original reporter (Denis) blamed particularly on the fact,\nkmiar> that Linux allows to read all queued data until EOF.\nkmiar> Try yourself, if you do not believe.\n\nkmiar> Unfortunately, I deleted that his mail, but you can find it\nkmiar> in mail archives I think, it was to netdev or to linux-kernel.\n\nI blamed that: Linux gives you EPIPE when you call recv before all\ndata available is retrieved. If you will try to read AFTER error you\nwill get all data. Problem is that it makes handling very complicated.\nIn the case of EPIPE you should try to read again. The problem is that\nyou should always try only once.\n\n-- \nBest regards,\n Denis mailto:[email protected]\n\n\n", "msg_date": "Wed, 5 Jul 2000 21:34:18 +0400", "msg_from": "Denis Perchine <[email protected]>", "msg_from_op": true, "msg_subject": "Re[2]: Fwd: Re: Fwd: Problem with recv syscall on socket when other\n\tside closed connection" }, { "msg_contents": "Denis Perchine <[email protected]> writes:\n> I blamed that: Linux gives you EPIPE when you call recv before all\n> data available is retrieved. If you will try to read AFTER error you\n> will get all data. Problem is that it makes handling very complicated.\n> In the case of EPIPE you should try to read again. The problem is that\n> you should always try only once.\n\nAh, thanks for the correction. Now, should we really program around\nthis behavior of the Linux kernel? I cannot think of any other OS where\nit is appropriate to retry on an EPIPE error condition, nor does it make\nany sense to do so given the normal meaning of that error code. \"Retry,\nbut only once\" is even sillier.\n\nI still think this behavior is a bug, not a feature. If you want to\nissue EPIPE (or more likely ECONNRESET) *after* all available data has\nbeen read, that's fine, and EPIPE for subsequent send attempts makes\nsense too. But EPIPE on read when there is more data available is just\nplain bizarre.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 05 Jul 2000 13:50:42 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re[2]: Fwd: Re: Fwd: Problem with recv syscall on socket when\n\tother side closed connection" }, { "msg_contents": "Hello!\n\n> I blamed that: Linux gives you EPIPE when you call recv before all\n> data available is retrieved. If you will try to read AFTER error you\n> will get all data. Problem is that it makes handling very complicated.\n> In the case of EPIPE you should try to read again. The problem is that\n> you should always try only once.\n\nWell, to me it does not look very essential, when asynchronous\nerror returned. Remember about EAGAIN and EINTR yet. You are not confused\nwith such erros, right? Why? 8)\n\nSeems, this order of issuing errors etc. at read() is specified in posix.\nI do not know really. I have said, error reporting only if no data are pending,\nlooks legal and has its merits. Main thing is not to forget to report error\nat all. 8)\n\n[ Alan, seems, all the comments about order of checks while read()\n are your ones. Can you comment? Maybe, it is really worth to change.\n]\n\nSide note: TLI really does not _allow_ any operations on endpoint\nin any direction until asynchronous error condition is cleared.\nIn fact, Linux does this on BSD sockets as well.\nThis is really natural, but I agree, it is inconvenient yet.\n\nAlexey\n", "msg_date": "Wed, 5 Jul 2000 22:00:52 +0400 (MSK DST)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: Re[2]: Fwd: Re: Fwd: Problem with recv syscall on socket when\n\tother side closed connection" } ]
[ { "msg_contents": "I'm in the process of fixing build problems in 7.0.2 for Solaris CC\n(not GCC).\n\n1. I'm curious why src/template/solaris_sparc_cc contains:\n -DDISABLE_COMPLEX_MACRO\nsrc/templates/solaris_i386_cc does not contains this definition.\n\nThis breaks the compilation:\ngmake[3]: Entering directory\n`/opt/build/postgresql-7.0.2/src/interfaces/ecpg/preproc'\ncc -o ecpg preproc.o pgc.o type.o ecpg.o ecpg_keywords.o output.o\nkeywords.o c_keywords.o ../lib/typename.o descriptor.o variable.o\n../../../backend/port/snprintf.o -R/opt/TWWfsw/pgsql702/lib\n-R/opt/TWWfsw/tcl83/lib -R/opt/TWWfsw/tk83/lib\n-R/opt/TWWfsw/readline4/lib -L/opt/TWWfsw/tcl83/lib\n-L/opt/TWWfsw/tk83/lib -L/opt/TWWfsw/readline4/lib -lgen -lsocket\n-lnsl -ldl -lm -lreadline -ltermcap -lcurses \nUndefined first referenced\n symbol in file\nnocachegetattr pgc.o\n\nThe reason for this is because nocachegetattr becomes a function when\nDISABLE_COMPLEX_MACRO is defined (nocachegetattr is defined in\nsrc/backend/access/common/heaptuple.c).\n\nThe first patch below removes this macro from\nsrc/templates/solaris_sparc_cc.\n\n2. Solaris CC does not like \"-K PIC\". It does like \"-KPIC\" though.\n\n3. Why is NAN defined in src/include/solaris_i386.h as:\n#define NAN (*(__const double *) __nan)\n#endif /* GCC. */\nif the compiler is not GCC? The Solaris C compiler does not like this.\nI removed it and the compilation succeeded.\n\n4. I moved some of the information out of\nsrc/interfaces/libpq++/Makefile.in into src/Makefile.shlib where it\nbelongs. Assuming that CXXFLAGS should take on the shared library\nflags for CFLAGS_SL is wrong though.\n\n5. added '-' before pl/tcl/Makefile so make does not complain if it\ncannot find pl/tcl/Makefile.tcldefs (which will be automatically\ngenerated).\n\n6. When building C++ shared libraries and archives under Solaris with\nthe Sun C/C++ compiler, use it to build the libraries and archives,\nnot ld/ar.\n\n7. Honor CXXFLAGS from configure command-line.\n\n8. In configure:\n a) It is evil to use a library if it exists:\n (AC_CHECK_LIB(util,main))\n It is far better to check for a function in the library. Better\n yet, make sure that function does not exist in libc or $LIBS\n before checking in external libraries. I have fixed this for\n some libraries but don't know what to check for in others\n (like ipc, IPC, lc, ld, compat, BSD).\n b) Added --with-readline=DIR to check for readline in DIR.\n\n-- \nalbert chin ([email protected])\n\n-- snip snip\n--- src/template/solaris_sparc_cc.orig\tTue Jun 20 19:22:44 2000\n+++ src/template/solaris_sparc_cc\tTue Jun 27 22:15:44 2000\n@@ -1,6 +1,6 @@\n AROPT:crs\n-CFLAGS:-Xa -v -D__sparc__ -D__sun__ -DDISABLE_COMPLEX_MACRO\n-SHARED_LIB:-K PIC\n+CFLAGS:-Xa -v -D__sparc__ -D__sun__\n+SHARED_LIB:-KPIC\n ALL:\n SRCH_INC:\n SRCH_LIB:\n--- src/template/solaris_i386_cc.orig\tTue Jun 27 21:36:15 2000\n+++ src/template/solaris_i386_cc\tTue Jun 27 21:36:19 2000\n@@ -1,6 +1,6 @@\n AROPT:cq\n CFLAGS:\n-SHARED_LIB:-K PIC\n+SHARED_LIB:-KPIC\n ALL:\n SRCH_INC:\n SRCH_LIB:\n--- src/include/port/solaris_i386.h.orig\tTue Jun 27 11:53:06 2000\n+++ src/include/port/solaris_i386.h\tTue Jun 27 11:53:20 2000\n@@ -31,8 +31,6 @@\n \t\t\t\t\t\t double __d; })\t\t\t\t\t\t\t \\\n \t\t\t\t { __nan_bytes }).__d)\n \n-#else\t\t\t\t\t\t\t/* Not GCC. */\n-#define\t\t\t\t NAN\t (*(__const double *) __nan)\n #endif\t /* GCC. */\n #endif\t /* NAN */\n \n--- src/interfaces/libpq++/Makefile.in.orig\tTue Jun 20 17:55:36 2000\n+++ src/interfaces/libpq++/Makefile.in\tTue Jun 27 21:27:23 2000\n@@ -39,22 +39,11 @@\n SHLIB_LINK= $(LIBPQ)\n endif\n \n-# For CC on IRIX, must use CC as linker/archiver of C++ libraries\n-ifeq ($(PORTNAME), irix5)\n- ifeq ($(CXX), CC)\n- AR = CC\n- AROPT = -ar -o\n- LD = CC\n- endif\n-endif\n-\n # Shared library stuff, also default 'all' target\n include $(SRCDIR)/Makefile.shlib\n \n-\n-# Pull shared-lib CFLAGS into CXXFLAGS\n+# Pull shared-lib CFLAGS into CXXFLAGS \n CXXFLAGS+= $(CFLAGS_SL)\n-\n \n .PHONY: examples\n examples:\n--- src/pl/tcl/Makefile.orig\tTue Jun 27 12:53:04 2000\n+++ src/pl/tcl/Makefile\tTue Jun 27 13:01:55 2000\n@@ -19,7 +19,7 @@\n # NOTE: GNU make will make this file automatically if it doesn't exist,\n # using the make rule that appears below. Cute, eh?\n #\n-include Makefile.tcldefs\n+-include Makefile.tcldefs\n \n #\n # Find out whether Tcl was built as a shared library --- if not,\n--- src/Makefile.shlib.orig\tTue Jun 27 12:22:09 2000\n+++ src/Makefile.shlib\tTue Jun 27 17:09:30 2000\n@@ -130,16 +130,23 @@\n ifeq ($(PORTNAME), hpux)\n install-shlib-dep\t:= install-shlib\n # HPUX doesn't believe in version numbers for shlibs\n- shlib\t\t\t\t:= lib$(NAME)$(DLSUFFIX)\n+ shlib\t\t\t:= lib$(NAME)$(DLSUFFIX)\n LDFLAGS_SL\t\t:= -b\n- CFLAGS\t\t\t+= $(CFLAGS_SL)\n+ CFLAGS\t\t+= $(CFLAGS_SL)\n endif\n \n ifeq ($(PORTNAME), irix5)\n- install-shlib-dep := install-shlib\n- shlib := lib$(NAME)$(DLSUFFIX).$(SO_MAJOR_VERSION)\n- LDFLAGS_SL := -shared -rpath $(LIBDIR) -set_version sgi$(SO_MAJOR_VERSION).$(SO_MINOR_VERSION)\n- CFLAGS += $(CFLAGS_SL)\n+ install-shlib-dep\t:= install-shlib\n+ shlib\t\t\t:= lib$(NAME)$(DLSUFFIX).$(SO_MAJOR_VERSION)\n+ LDFLAGS_SL\t\t:= -shared -rpath $(LIBDIR) -set_version sgi$(SO_MAJOR_VERSION).$(SO_MINOR_VERSION)\n+ CFLAGS\t\t+= $(CFLAGS_SL)\n+\n+ # if archiving or linking C++ objects, must use CC (according to CC(1))\n+ ifeq ($(CXX), CC)\n+ AR = CC\n+ AROPT = -ar -o\n+ LD = CC\n+ endif\n endif\n \n ifeq ($(PORTNAME), linux)\n@@ -156,7 +163,14 @@\n shlib\t\t\t\t:= lib$(NAME)$(DLSUFFIX).$(SO_MAJOR_VERSION).$(SO_MINOR_VERSION)\n LDFLAGS_SL\t\t:= -G\n SHLIB_LINK\t\t+= -ldl -lsocket -lresolv -lnsl -lm -lc\n- CFLAGS\t\t\t+= $(CFLAGS_SL)\n+ CFLAGS\t\t+= $(CFLAGS_SL)\n+\n+ # if archiving or linking C++ objects, must use CC (according to CC(1))\n+ ifeq ($(CXX), CC)\n+ AR\t\t\t:= CC\n+ AROPT\t\t:= -xar -o\n+ LD\t\t\t:= CC\n+ endif\n endif\n \n ifeq ($(PORTNAME), solaris_sparc)\n@@ -164,7 +178,14 @@\n shlib\t\t\t\t:= lib$(NAME)$(DLSUFFIX).$(SO_MAJOR_VERSION).$(SO_MINOR_VERSION)\n LDFLAGS_SL\t\t:= -G\n SHLIB_LINK\t\t+= -ldl -lsocket -lresolv -lnsl -lm -lc\n- CFLAGS\t\t\t+= $(CFLAGS_SL)\n+ CFLAGS\t\t+= $(CFLAGS_SL)\n+\n+ # if linking C++ objects, must use CC (according to CC(1))\n+ ifeq ($(CXX), CC)\n+ AR\t\t\t:= CC\n+ AROPT\t\t:= -xar -o\n+ LD\t\t\t:= CC\n+ endif\n endif\n \n ifeq ($(PORTNAME), alpha)\n--- src/Makefile.global.in.orig\tTue Jun 20 17:53:01 2000\n+++ src/Makefile.global.in\tTue Jun 20 17:53:14 2000\n@@ -209,6 +209,7 @@\n LEX= @LEX@\n AROPT= @AROPT@\n CFLAGS= -I$(SRCDIR)/include -I$(SRCDIR)/backend @CPPFLAGS@ @CFLAGS@\n+CXXFLAGS= @CXXFLAGS@\n CFLAGS_SL= @SHARED_LIB@\n PGSQL_INCLUDES= @PGSQL_INCLUDES@\n LIBS= @LIBS@\n--- src/configure.in.orig\tTue Jun 20 17:42:51 2000\n+++ src/configure.in\tTue Jun 27 16:00:11 2000\n@@ -675,24 +675,46 @@\n AC_SUBST(YFLAGS)\n \n AC_CHECK_LIB(sfio, main)\n-for curses in ncurses curses ; do\n-\tAC_CHECK_LIB(${curses}, main,\n-\t\t[LIBS=\"-l${curses} $LIBS\"; break])\n+\n+for termlib in ncurses curses termcap terminfo termlib; do\n+ AC_CHECK_LIB(${termlib}, tputs, [LIBS=\"-l${termlib} $LIBS\"; break])\n done\n-AC_CHECK_LIB(termcap, main)\n-AC_CHECK_LIB(readline, main)\n-AC_CHECK_LIB(readline, using_history, AC_DEFINE(HAVE_HISTORY_IN_READLINE),\n- AC_CHECK_LIB(history, main) )\n+AC_ARG_WITH(readline,\n+[ --with-readline=DIR search DIR for readline includes and libraries],[\n+ case \"$with_readline\" in\n+ no) ;;\n+ *)\n+ if test \"$with_readline\" != \"yes\"; then\n+ _cppflags=${CPPFLAGS}\n+ _ldflags=${LDFLAGS}\n+ CPPFLAGS=\"${CPPFLAGS} -I$with_readline/include\"\n+ LDFLAGS=\"${LDFLAGS} -L$with_readline/lib\"\n+ fi\n+\n+ AC_CHECK_HEADERS(readline.h readline/history.h readline/readline.h,\n+ AC_CHECK_LIB(readline, readline)\n+ AC_CHECK_LIB(readline, using_history,\n+\tAC_DEFINE(HAVE_HISTORY_IN_READLINE),\n+\tAC_CHECK_LIB(history, add_history)))\n+\n+ if test -z \"$ac_cv_header_readline_h\" -a \\\n+ -z \"$ac_cv_header_readline_readline_h\"; then\n+ if test -n \"${_cppflags}\"; then\n+\tCPPFLAGS=\"${_cppflags}\" LDFLAGS=\"${_ldflags}\"\n+ fi\n+ fi\n+ ;;\n+ esac])\n \n if test \"$PORTNAME\" != \"aix\" -a \"$PORTNAME\" != \"alpha\"\n then\n \tAC_CHECK_LIB(bsd, main)\n fi\n AC_CHECK_LIB(util, main)\n-AC_CHECK_LIB(m, main)\n-AC_CHECK_LIB(dl, main)\n-AC_CHECK_LIB(socket, main)\n-AC_CHECK_LIB(nsl, main)\n+AC_CHECK_FUNC(pow, , AC_CHECK_LIB(m, pow))\n+AC_CHECK_FUNC(dlopen, , AC_CHECK_LIB(dl, dlopen))\n+AC_CHECK_FUNC(gethostent, , AC_CHECK_LIB(nsl, gethostent))\n+AC_CHECK_FUNC(setsockopt, , AC_CHECK_LIB(socket, setsockopt))\n AC_CHECK_LIB(ipc, main)\n AC_CHECK_LIB(IPC, main)\n AC_CHECK_LIB(lc, main)\n@@ -701,7 +723,7 @@\n AC_CHECK_LIB(ld, main)\n AC_CHECK_LIB(compat, main)\n AC_CHECK_LIB(BSD, main)\n-AC_CHECK_LIB(crypt, main)\n+AC_CHECK_FUNC(crypt, , AC_CHECK_LIB(crypt, crypt))\n AC_CHECK_LIB(gen, main)\n AC_CHECK_LIB(PW, main)\n \n\n", "msg_date": "Tue, 27 Jun 2000 12:54:43 -0500 (CDT)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "7.0.2 on Solaris" }, { "msg_contents": "[email protected] writes:\n> 1. I'm curious why src/template/solaris_sparc_cc contains:\n> -DDISABLE_COMPLEX_MACRO\n\nPresumably because the Solaris Sparc compiler was once broken.\nDunno how long ago that was, but I have no objection to removing\nthe -D switch if it seems to work on not-too-ancient copies.\nHow recent is your Solaris?\n\n> The reason for this is because nocachegetattr becomes a function when\n> DISABLE_COMPLEX_MACRO is defined (nocachegetattr is defined in\n> src/backend/access/common/heaptuple.c).\n\n> The first patch below removes this macro from\n> src/templates/solaris_sparc_cc.\n\nIt's broken all right, but that's no fix --- it just masks the problem\non your platform. The breakage is that somebody decided a static\nfunction definition in an include file is a cool idea. Not when the\ninclude file is included by non-backend modules ... the non-macro\ndefinition of fastgetattr() needs to be made a plain extern routine.\n\n> 3. Why is NAN defined in src/include/solaris_i386.h as:\n> #define NAN (*(__const double *) __nan)\n> #endif /* GCC. */\n> if the compiler is not GCC? The Solaris C compiler does not like this.\n> I removed it and the compilation succeeded.\n\nHmm, I think this needs more investigation. Removing a definition of\nNAN that's inside \"#ifndef NAN\" cannot cause things to work unless there\nis an order dependency with another definition of NAN. Perhaps we need\nto be rearranging header includes or some such. Again, it seems like\nthe true issue might affect more than just solaris.\n\n> 4. I moved some of the information out of\n> src/interfaces/libpq++/Makefile.in into src/Makefile.shlib where it\n> belongs. Assuming that CXXFLAGS should take on the shared library\n> flags for CFLAGS_SL is wrong though.\n\nNope, try again. Your change breaks things for all non-C++ shared libs,\nbecause it changes AR and LD to C++-suitable definitions for *all* shlib\nmodules not just C++ modules. That might happen to work on your\ninstallation but it's a recipe for trouble. The reason that the C++\ninfo is in src/interfaces/libpq++/Makefile.in is that that's where it\nbelongs, to avoid messing up non-C++ modules. (If we had more than one\nC++ shlib I'd be more excited about figuring out how to keep the info in\nMakefile.shlib, but as things stand I think leaving it out is the path\nof least resistance.)\n\n> b) Added --with-readline=DIR to check for readline in DIR.\n\nWhat for? We have --with-includes and --with-libs; shall we now\nadd a custom \"--with-libfoo\" option for every single library we\ndepend on? This seems like useless clutter.\n\nSome of the other changes look good but I'm not in a position to\nseparate them out from the less-good stuff...\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 28 Jun 2000 01:56:49 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: 7.0.2 on Solaris " }, { "msg_contents": "On Wed, Jun 28, 2000 at 01:56:49AM -0400, Tom Lane wrote:\n> [email protected] writes:\n> > 1. I'm curious why src/template/solaris_sparc_cc contains:\n> > -DDISABLE_COMPLEX_MACRO\n> \n> Presumably because the Solaris Sparc compiler was once broken.\n> Dunno how long ago that was, but I have no objection to removing\n> the -D switch if it seems to work on not-too-ancient copies.\n> How recent is your Solaris?\n\nWe're runnig 5.0 with the latest set of patches. 6.0 just came out but\nwe don't have our copy yet.\n\n> > The reason for this is because nocachegetattr becomes a function when\n> > DISABLE_COMPLEX_MACRO is defined (nocachegetattr is defined in\n> > src/backend/access/common/heaptuple.c).\n> \n> > The first patch below removes this macro from\n> > src/templates/solaris_sparc_cc.\n> \n> It's broken all right, but that's no fix --- it just masks the problem\n> on your platform. The breakage is that somebody decided a static\n> function definition in an include file is a cool idea. Not when the\n> include file is included by non-backend modules ... the non-macro\n> definition of fastgetattr() needs to be made a plain extern routine.\n\nUnderstood. We took the easy way out. Considering noone else has\nreported a problem and that 5.0 has been out for awhile (though I\nimagine some people might still be using 4.2), it should be safe to\nremove.\n\n> > 3. Why is NAN defined in src/include/solaris_i386.h as:\n> > #define NAN (*(__const double *) __nan)\n> > #endif /* GCC. */\n> > if the compiler is not GCC? The Solaris C compiler does not like this.\n> > I removed it and the compilation succeeded.\n> \n> Hmm, I think this needs more investigation. Removing a definition of\n> NAN that's inside \"#ifndef NAN\" cannot cause things to work unless there\n> is an order dependency with another definition of NAN. Perhaps we need\n> to be rearranging header includes or some such. Again, it seems like\n> the true issue might affect more than just solaris.\n\nNote that Solaris/SPARC doesn't have this #ifndef NAN stuff. Why be\nspecific to Solaris/Intel? Solaris CC does not have __const so if left\nin it doesn't compile. \n\nThe reason the compile succeeds when the #else section is removed is\nbecause of the following in src/include/utils/timestamp.h:\n#ifdef NAN\n#define DT_INVALID (NAN)\n#else\n#define DT_INVALID (DBL_MIN+DBL_MIN)\n#endif\n\nSo, DBL_MIN+DBL_MIN is used rather than NAN. Dunno if this is a \"good\nthing\". There are other places, like src/backend/utils/adt/numeric.c,\nthat:\n#ifndef NAN\n#define NAN (0.0/0.0)\n#endif\n\nAgain, I don't know if this is a \"good thing\" (and if NAN is defined\nto be this here, why not similarly for src/include/utils/timestamp.h\nabove?).\n\n> > 4. I moved some of the information out of\n> > src/interfaces/libpq++/Makefile.in into src/Makefile.shlib where it\n> > belongs. Assuming that CXXFLAGS should take on the shared library\n> > flags for CFLAGS_SL is wrong though.\n> \n> Nope, try again. Your change breaks things for all non-C++ shared libs,\n> because it changes AR and LD to C++-suitable definitions for *all* shlib\n> modules not just C++ modules. That might happen to work on your\n> installation but it's a recipe for trouble. The reason that the C++\n> info is in src/interfaces/libpq++/Makefile.in is that that's where it\n> belongs, to avoid messing up non-C++ modules. (If we had more than one\n> C++ shlib I'd be more excited about figuring out how to keep the info in\n> Makefile.shlib, but as things stand I think leaving it out is the path\n> of least resistance.)\n\nOk. I now understand why it was in the Makefile in libpq++. I've\nattached a patch to add my Makefile.shlib changes to\nsrc/interfaces/libpq++/Makefile.in.\n\n> \n> > b) Added --with-readline=DIR to check for readline in DIR.\n> \n> What for? We have --with-includes and --with-libs; shall we now\n> add a custom \"--with-libfoo\" option for every single library we\n> depend on? This seems like useless clutter.\n\nI think so. Makes it easier when running ./configure to figure out\nwhat extra 3rd-party modules are being used. It also means every\nsubmodule does not necessary have to -L everything in --with-libs\nand -I everything in --with-includes. Can you guarantee that this\nwill always be safe?\n\n> \t\t\tregards, tom lane\n\n-- \nalbert chin ([email protected])\n\n-- snip snip\n--- src/interfaces/libpq++/Makefile.in.orig\tTue Jun 20 17:55:36 2000\n+++ src/interfaces/libpq++/Makefile.in\tWed Jun 28 08:55:30 2000\n@@ -39,22 +39,38 @@\n SHLIB_LINK= $(LIBPQ)\n endif\n \n-# For CC on IRIX, must use CC as linker/archiver of C++ libraries\n ifeq ($(PORTNAME), irix5)\n+ # if archiving or linking C++ objects, must use CC (according to CC(1))\n ifeq ($(CXX), CC)\n- AR = CC\n- AROPT = -ar -o\n- LD = CC\n+ AR\t\t\t:= CC\n+ AROPT\t\t:= -ar -o\n+ LD\t\t\t:= CC\n+ endif\n+endif\n+\n+ifeq ($(PORTNAME), solaris_i386)\n+ # if archiving or linking C++ objects, must use CC (according to CC(1))\n+ ifeq ($(CXX), CC) \n+ AR\t\t\t:= CC \n+ AROPT\t\t:= -xar -o\n+ LD\t\t\t:= CC \n+ endif \n+enfi\n+\n+ifeq ($(PORTNAME), solaris_sparc)\n+ # if linking C++ objects, must use CC (according to CC(1))\n+ ifeq ($(CXX), CC)\n+ AR\t\t\t:= CC\n+ AROPT\t\t:= -xar -o\n+ LD\t\t\t:= CC\n endif\n endif\n \n # Shared library stuff, also default 'all' target\n include $(SRCDIR)/Makefile.shlib\n \n-\n-# Pull shared-lib CFLAGS into CXXFLAGS\n+# Pull shared-lib CFLAGS into CXXFLAGS \n CXXFLAGS+= $(CFLAGS_SL)\n-\n \n .PHONY: examples\n examples:\n\n", "msg_date": "Wed, 28 Jun 2000 09:34:44 -0500 (CDT)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "Re: 7.0.2 on Solaris" }, { "msg_contents": "[email protected] writes:\n\n> 3. Why is NAN defined in src/include/solaris_i386.h as:\n\nWe probably need a real configure test for this.\n\n> 5. added '-' before pl/tcl/Makefile so make does not complain if it\n> cannot find pl/tcl/Makefile.tcldefs (which will be automatically\n> generated).\n\nYeah but if the automatic generation fails then it won't include it at\nall?\n\n> 7. Honor CXXFLAGS from configure command-line.\n\nIt already does that as far as I can tell. Defining the variable in\nMakefile.global instead of libpq++/Makefile doesn't make a difference.\n\n> 8. In configure:\n> a) It is evil to use a library if it exists:\n> (AC_CHECK_LIB(util,main))\n> It is far better to check for a function in the library.\n\nAmen to that, but unfortunately most of the information about which\nfunction was required from which library is lost in history. Btw., the\nproper solution to this is AC_SEARCH_LIBS, not what you wrote.\n\nIn general, if you want to contribute changes in this area it might be a\ngood idea to start from the current sources, because I'm all over that\ncode at the moment.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Wed, 28 Jun 2000 20:38:08 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: 7.0.2 on Solaris" }, { "msg_contents": "On Wed, Jun 28, 2000 at 08:38:08PM +0200, Peter Eisentraut wrote:\n> [email protected] writes:\n> \n> > 3. Why is NAN defined in src/include/solaris_i386.h as:\n> \n> We probably need a real configure test for this.\n\nProbably.\n\n> > 5. added '-' before pl/tcl/Makefile so make does not complain if it\n> > cannot find pl/tcl/Makefile.tcldefs (which will be automatically\n> > generated).\n> \n> Yeah but if the automatic generation fails then it won't include it at\n> all?\n\nNo. But, it will be silent so you won't know. However, you should get\nan error if the automatic generation fails.\n\n> > 7. Honor CXXFLAGS from configure command-line.\n> \n> It already does that as far as I can tell. Defining the variable in\n> Makefile.global instead of libpq++/Makefile doesn't make a difference.\n\nOk, thanks.\n\n> > 8. In configure:\n> > a) It is evil to use a library if it exists:\n> > (AC_CHECK_LIB(util,main))\n> > It is far better to check for a function in the library.\n> \n> Amen to that, but unfortunately most of the information about which\n> function was required from which library is lost in history. Btw., the\n> proper solution to this is AC_SEARCH_LIBS, not what you wrote.\n\nJust found out about AC_SEARCH_LIBS. Thanks. How about getting rid of\nall the AC_CHECK_LIB([library],main) code and adding things back in as\nthey break?\n\n> In general, if you want to contribute changes in this area it might be a\n> good idea to start from the current sources, because I'm all over that\n> code at the moment.\n\nOk. So should I get CVS and reapply my patches?\n\n-- \nalbert chin ([email protected])\n\n", "msg_date": "Wed, 28 Jun 2000 14:25:41 -0500 (CDT)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "Re: 7.0.2 on Solaris" }, { "msg_contents": "On Wed, Jun 28, 2000 at 08:38:08PM +0200, Peter Eisentraut wrote:\n> [email protected] writes:\n> \n> > 8. In configure:\n> > a) It is evil to use a library if it exists:\n> > (AC_CHECK_LIB(util,main))\n> > It is far better to check for a function in the library.\n> \n> Amen to that, but unfortunately most of the information about which\n> function was required from which library is lost in history. Btw., the\n> proper solution to this is AC_SEARCH_LIBS, not what you wrote.\n\nAC_SEARCH_LIBS is the correct solution providing you don't need the\nfeature of AC_CHECK_LIB that defines HAVE_LIB[LIBRARY] and adds\n-l[lib] to LIBS.\n\n-- \nalbert chin ([email protected])\n\n", "msg_date": "Thu, 29 Jun 2000 05:45:17 -0500 (CDT)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "Re: 7.0.2 on Solaris" }, { "msg_contents": "[email protected] writes:\n\n> Ok. So should I get CVS and reapply my patches?\n\nWell, let's see:\n\n> 1. -DDISABLE_COMPLEX_MACRO\n\nThis clearly needs to be fixed the right way. I'll take care of it.\n\n> 2. Solaris CC does not like \"-K PIC\". It does like \"-KPIC\" though.\n\nAlso no problem.\n\n> 3. NAN\n\nWe need a real solution for this. I'll look into it.\n\n> 4. I moved some of the information out of\n> src/interfaces/libpq++/Makefile.in into src/Makefile.shlib\n\nI think we don't want that.\n\n> 5. added '-' before pl/tcl/Makefile so make does not complain\n\nThat I can do as well. I confirmed that it works okay.\n\n> 6. When building C++ shared libraries and archives under Solaris with\n> the Sun C/C++ compiler, use it to build the libraries and archives,\n> not ld/ar.\n\nThis is similar to the IRIX case apparently, so okay.\n\n> 7. Honor CXXFLAGS from configure command-line.\n\nShould work fine.\n\n> 8. (AC_CHECK_LIB(util,main))\n\nI don't think we want to try that now. I happen to know that the -lutil\ncase was added for setproctitle() so I can change that, but the rest is\njust a cosmetics issue, unless you can prove breakage. \n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Fri, 30 Jun 2000 01:32:04 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: 7.0.2 on Solaris" }, { "msg_contents": "[email protected] writes:\n\n> 3. Why is NAN defined in src/include/solaris_i386.h as:\n> #define NAN (*(__const double *) __nan)\n> #endif /* GCC. */\n> if the compiler is not GCC? The Solaris C compiler does not like this.\n> I removed it and the compilation succeeded.\n\nWe never cleared out this item. What is the proper way to generate a `NaN'\non your compiler? Does `0.0/0.0' work?\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Sun, 2 Jul 2000 17:22:53 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: 7.0.2 on Solaris" }, { "msg_contents": "On Sun, Jul 02, 2000 at 05:22:53PM +0200, Peter Eisentraut wrote:\n> [email protected] writes:\n> \n> > 3. Why is NAN defined in src/include/solaris_i386.h as:\n> > #define NAN (*(__const double *) __nan)\n> > #endif /* GCC. */\n> > if the compiler is not GCC? The Solaris C compiler does not like this.\n> > I removed it and the compilation succeeded.\n> \n> We never cleared out this item. What is the proper way to generate a `NaN'\n> on your compiler? Does `0.0/0.0' work?\n\nI tried the following:\n#include <ieeefp.h>\n#include <stdio.h>\n\nint\nmain (void) {\n float f = 0.0/0.0;\n\n if (isnanf (f)) {\n puts (\"nan\");\n }\n}\n\nand the output was \"nan\".\n\n-- \nalbert chin ([email protected])\n\n", "msg_date": "Tue, 4 Jul 2000 12:26:34 -0500 (CDT)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "Re: 7.0.2 on Solaris" }, { "msg_contents": "I think our source tree has diverged quite a bit from 7.0.2. Could you\ngrab a snapshot and let me know what needs changing for Solaris?\nThanks.\n\n> I'm in the process of fixing build problems in 7.0.2 for Solaris CC\n> (not GCC).\n> \n> 1. I'm curious why src/template/solaris_sparc_cc contains:\n> -DDISABLE_COMPLEX_MACRO\n> src/templates/solaris_i386_cc does not contains this definition.\n> \n> This breaks the compilation:\n> gmake[3]: Entering directory\n> `/opt/build/postgresql-7.0.2/src/interfaces/ecpg/preproc'\n> cc -o ecpg preproc.o pgc.o type.o ecpg.o ecpg_keywords.o output.o\n> keywords.o c_keywords.o ../lib/typename.o descriptor.o variable.o\n> ../../../backend/port/snprintf.o -R/opt/TWWfsw/pgsql702/lib\n> -R/opt/TWWfsw/tcl83/lib -R/opt/TWWfsw/tk83/lib\n> -R/opt/TWWfsw/readline4/lib -L/opt/TWWfsw/tcl83/lib\n> -L/opt/TWWfsw/tk83/lib -L/opt/TWWfsw/readline4/lib -lgen -lsocket\n> -lnsl -ldl -lm -lreadline -ltermcap -lcurses \n> Undefined first referenced\n> symbol in file\n> nocachegetattr pgc.o\n> \n> The reason for this is because nocachegetattr becomes a function when\n> DISABLE_COMPLEX_MACRO is defined (nocachegetattr is defined in\n> src/backend/access/common/heaptuple.c).\n> \n> The first patch below removes this macro from\n> src/templates/solaris_sparc_cc.\n> \n> 2. Solaris CC does not like \"-K PIC\". It does like \"-KPIC\" though.\n> \n> 3. Why is NAN defined in src/include/solaris_i386.h as:\n> #define NAN (*(__const double *) __nan)\n> #endif /* GCC. */\n> if the compiler is not GCC? The Solaris C compiler does not like this.\n> I removed it and the compilation succeeded.\n> \n> 4. I moved some of the information out of\n> src/interfaces/libpq++/Makefile.in into src/Makefile.shlib where it\n> belongs. Assuming that CXXFLAGS should take on the shared library\n> flags for CFLAGS_SL is wrong though.\n> \n> 5. added '-' before pl/tcl/Makefile so make does not complain if it\n> cannot find pl/tcl/Makefile.tcldefs (which will be automatically\n> generated).\n> \n> 6. When building C++ shared libraries and archives under Solaris with\n> the Sun C/C++ compiler, use it to build the libraries and archives,\n> not ld/ar.\n> \n> 7. Honor CXXFLAGS from configure command-line.\n> \n> 8. In configure:\n> a) It is evil to use a library if it exists:\n> (AC_CHECK_LIB(util,main))\n> It is far better to check for a function in the library. Better\n> yet, make sure that function does not exist in libc or $LIBS\n> before checking in external libraries. I have fixed this for\n> some libraries but don't know what to check for in others\n> (like ipc, IPC, lc, ld, compat, BSD).\n> b) Added --with-readline=DIR to check for readline in DIR.\n> \n> -- \n> albert chin ([email protected])\n> \n> -- snip snip\n> --- src/template/solaris_sparc_cc.orig\tTue Jun 20 19:22:44 2000\n> +++ src/template/solaris_sparc_cc\tTue Jun 27 22:15:44 2000\n> @@ -1,6 +1,6 @@\n> AROPT:crs\n> -CFLAGS:-Xa -v -D__sparc__ -D__sun__ -DDISABLE_COMPLEX_MACRO\n> -SHARED_LIB:-K PIC\n> +CFLAGS:-Xa -v -D__sparc__ -D__sun__\n> +SHARED_LIB:-KPIC\n> ALL:\n> SRCH_INC:\n> SRCH_LIB:\n> --- src/template/solaris_i386_cc.orig\tTue Jun 27 21:36:15 2000\n> +++ src/template/solaris_i386_cc\tTue Jun 27 21:36:19 2000\n> @@ -1,6 +1,6 @@\n> AROPT:cq\n> CFLAGS:\n> -SHARED_LIB:-K PIC\n> +SHARED_LIB:-KPIC\n> ALL:\n> SRCH_INC:\n> SRCH_LIB:\n> --- src/include/port/solaris_i386.h.orig\tTue Jun 27 11:53:06 2000\n> +++ src/include/port/solaris_i386.h\tTue Jun 27 11:53:20 2000\n> @@ -31,8 +31,6 @@\n> \t\t\t\t\t\t double __d; })\t\t\t\t\t\t\t \\\n> \t\t\t\t { __nan_bytes }).__d)\n> \n> -#else\t\t\t\t\t\t\t/* Not GCC. */\n> -#define\t\t\t\t NAN\t (*(__const double *) __nan)\n> #endif\t /* GCC. */\n> #endif\t /* NAN */\n> \n> --- src/interfaces/libpq++/Makefile.in.orig\tTue Jun 20 17:55:36 2000\n> +++ src/interfaces/libpq++/Makefile.in\tTue Jun 27 21:27:23 2000\n> @@ -39,22 +39,11 @@\n> SHLIB_LINK= $(LIBPQ)\n> endif\n> \n> -# For CC on IRIX, must use CC as linker/archiver of C++ libraries\n> -ifeq ($(PORTNAME), irix5)\n> - ifeq ($(CXX), CC)\n> - AR = CC\n> - AROPT = -ar -o\n> - LD = CC\n> - endif\n> -endif\n> -\n> # Shared library stuff, also default 'all' target\n> include $(SRCDIR)/Makefile.shlib\n> \n> -\n> -# Pull shared-lib CFLAGS into CXXFLAGS\n> +# Pull shared-lib CFLAGS into CXXFLAGS \n> CXXFLAGS+= $(CFLAGS_SL)\n> -\n> \n> .PHONY: examples\n> examples:\n> --- src/pl/tcl/Makefile.orig\tTue Jun 27 12:53:04 2000\n> +++ src/pl/tcl/Makefile\tTue Jun 27 13:01:55 2000\n> @@ -19,7 +19,7 @@\n> # NOTE: GNU make will make this file automatically if it doesn't exist,\n> # using the make rule that appears below. Cute, eh?\n> #\n> -include Makefile.tcldefs\n> +-include Makefile.tcldefs\n> \n> #\n> # Find out whether Tcl was built as a shared library --- if not,\n> --- src/Makefile.shlib.orig\tTue Jun 27 12:22:09 2000\n> +++ src/Makefile.shlib\tTue Jun 27 17:09:30 2000\n> @@ -130,16 +130,23 @@\n> ifeq ($(PORTNAME), hpux)\n> install-shlib-dep\t:= install-shlib\n> # HPUX doesn't believe in version numbers for shlibs\n> - shlib\t\t\t\t:= lib$(NAME)$(DLSUFFIX)\n> + shlib\t\t\t:= lib$(NAME)$(DLSUFFIX)\n> LDFLAGS_SL\t\t:= -b\n> - CFLAGS\t\t\t+= $(CFLAGS_SL)\n> + CFLAGS\t\t+= $(CFLAGS_SL)\n> endif\n> \n> ifeq ($(PORTNAME), irix5)\n> - install-shlib-dep := install-shlib\n> - shlib := lib$(NAME)$(DLSUFFIX).$(SO_MAJOR_VERSION)\n> - LDFLAGS_SL := -shared -rpath $(LIBDIR) -set_version sgi$(SO_MAJOR_VERSION).$(SO_MINOR_VERSION)\n> - CFLAGS += $(CFLAGS_SL)\n> + install-shlib-dep\t:= install-shlib\n> + shlib\t\t\t:= lib$(NAME)$(DLSUFFIX).$(SO_MAJOR_VERSION)\n> + LDFLAGS_SL\t\t:= -shared -rpath $(LIBDIR) -set_version sgi$(SO_MAJOR_VERSION).$(SO_MINOR_VERSION)\n> + CFLAGS\t\t+= $(CFLAGS_SL)\n> +\n> + # if archiving or linking C++ objects, must use CC (according to CC(1))\n> + ifeq ($(CXX), CC)\n> + AR = CC\n> + AROPT = -ar -o\n> + LD = CC\n> + endif\n> endif\n> \n> ifeq ($(PORTNAME), linux)\n> @@ -156,7 +163,14 @@\n> shlib\t\t\t\t:= lib$(NAME)$(DLSUFFIX).$(SO_MAJOR_VERSION).$(SO_MINOR_VERSION)\n> LDFLAGS_SL\t\t:= -G\n> SHLIB_LINK\t\t+= -ldl -lsocket -lresolv -lnsl -lm -lc\n> - CFLAGS\t\t\t+= $(CFLAGS_SL)\n> + CFLAGS\t\t+= $(CFLAGS_SL)\n> +\n> + # if archiving or linking C++ objects, must use CC (according to CC(1))\n> + ifeq ($(CXX), CC)\n> + AR\t\t\t:= CC\n> + AROPT\t\t:= -xar -o\n> + LD\t\t\t:= CC\n> + endif\n> endif\n> \n> ifeq ($(PORTNAME), solaris_sparc)\n> @@ -164,7 +178,14 @@\n> shlib\t\t\t\t:= lib$(NAME)$(DLSUFFIX).$(SO_MAJOR_VERSION).$(SO_MINOR_VERSION)\n> LDFLAGS_SL\t\t:= -G\n> SHLIB_LINK\t\t+= -ldl -lsocket -lresolv -lnsl -lm -lc\n> - CFLAGS\t\t\t+= $(CFLAGS_SL)\n> + CFLAGS\t\t+= $(CFLAGS_SL)\n> +\n> + # if linking C++ objects, must use CC (according to CC(1))\n> + ifeq ($(CXX), CC)\n> + AR\t\t\t:= CC\n> + AROPT\t\t:= -xar -o\n> + LD\t\t\t:= CC\n> + endif\n> endif\n> \n> ifeq ($(PORTNAME), alpha)\n> --- src/Makefile.global.in.orig\tTue Jun 20 17:53:01 2000\n> +++ src/Makefile.global.in\tTue Jun 20 17:53:14 2000\n> @@ -209,6 +209,7 @@\n> LEX= @LEX@\n> AROPT= @AROPT@\n> CFLAGS= -I$(SRCDIR)/include -I$(SRCDIR)/backend @CPPFLAGS@ @CFLAGS@\n> +CXXFLAGS= @CXXFLAGS@\n> CFLAGS_SL= @SHARED_LIB@\n> PGSQL_INCLUDES= @PGSQL_INCLUDES@\n> LIBS= @LIBS@\n> --- src/configure.in.orig\tTue Jun 20 17:42:51 2000\n> +++ src/configure.in\tTue Jun 27 16:00:11 2000\n> @@ -675,24 +675,46 @@\n> AC_SUBST(YFLAGS)\n> \n> AC_CHECK_LIB(sfio, main)\n> -for curses in ncurses curses ; do\n> -\tAC_CHECK_LIB(${curses}, main,\n> -\t\t[LIBS=\"-l${curses} $LIBS\"; break])\n> +\n> +for termlib in ncurses curses termcap terminfo termlib; do\n> + AC_CHECK_LIB(${termlib}, tputs, [LIBS=\"-l${termlib} $LIBS\"; break])\n> done\n> -AC_CHECK_LIB(termcap, main)\n> -AC_CHECK_LIB(readline, main)\n> -AC_CHECK_LIB(readline, using_history, AC_DEFINE(HAVE_HISTORY_IN_READLINE),\n> - AC_CHECK_LIB(history, main) )\n> +AC_ARG_WITH(readline,\n> +[ --with-readline=DIR search DIR for readline includes and libraries],[\n> + case \"$with_readline\" in\n> + no) ;;\n> + *)\n> + if test \"$with_readline\" != \"yes\"; then\n> + _cppflags=${CPPFLAGS}\n> + _ldflags=${LDFLAGS}\n> + CPPFLAGS=\"${CPPFLAGS} -I$with_readline/include\"\n> + LDFLAGS=\"${LDFLAGS} -L$with_readline/lib\"\n> + fi\n> +\n> + AC_CHECK_HEADERS(readline.h readline/history.h readline/readline.h,\n> + AC_CHECK_LIB(readline, readline)\n> + AC_CHECK_LIB(readline, using_history,\n> +\tAC_DEFINE(HAVE_HISTORY_IN_READLINE),\n> +\tAC_CHECK_LIB(history, add_history)))\n> +\n> + if test -z \"$ac_cv_header_readline_h\" -a \\\n> + -z \"$ac_cv_header_readline_readline_h\"; then\n> + if test -n \"${_cppflags}\"; then\n> +\tCPPFLAGS=\"${_cppflags}\" LDFLAGS=\"${_ldflags}\"\n> + fi\n> + fi\n> + ;;\n> + esac])\n> \n> if test \"$PORTNAME\" != \"aix\" -a \"$PORTNAME\" != \"alpha\"\n> then\n> \tAC_CHECK_LIB(bsd, main)\n> fi\n> AC_CHECK_LIB(util, main)\n> -AC_CHECK_LIB(m, main)\n> -AC_CHECK_LIB(dl, main)\n> -AC_CHECK_LIB(socket, main)\n> -AC_CHECK_LIB(nsl, main)\n> +AC_CHECK_FUNC(pow, , AC_CHECK_LIB(m, pow))\n> +AC_CHECK_FUNC(dlopen, , AC_CHECK_LIB(dl, dlopen))\n> +AC_CHECK_FUNC(gethostent, , AC_CHECK_LIB(nsl, gethostent))\n> +AC_CHECK_FUNC(setsockopt, , AC_CHECK_LIB(socket, setsockopt))\n> AC_CHECK_LIB(ipc, main)\n> AC_CHECK_LIB(IPC, main)\n> AC_CHECK_LIB(lc, main)\n> @@ -701,7 +723,7 @@\n> AC_CHECK_LIB(ld, main)\n> AC_CHECK_LIB(compat, main)\n> AC_CHECK_LIB(BSD, main)\n> -AC_CHECK_LIB(crypt, main)\n> +AC_CHECK_FUNC(crypt, , AC_CHECK_LIB(crypt, crypt))\n> AC_CHECK_LIB(gen, main)\n> AC_CHECK_LIB(PW, main)\n> \n> \n> \n\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 10 Oct 2000 20:56:27 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: 7.0.2 on Solaris" }, { "msg_contents": "Bruce Momjian writes:\n\n> I think our source tree has diverged quite a bit from 7.0.2. Could you\n> grab a snapshot and let me know what needs changing for Solaris?\n\nThese fixes were integrated. Let's remember to make a more detailed\n\"Supported Platforms\" list next time, because this platform was listed and\nit evidently didn't work.\n\n> > I'm in the process of fixing build problems in 7.0.2 for Solaris CC\n> > (not GCC).\n\n-- \nPeter Eisentraut [email protected] http://yi.org/peter-e/\n\n", "msg_date": "Wed, 11 Oct 2000 18:25:35 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: 7.0.2 on Solaris" }, { "msg_contents": "Which version of Solaris was he using? I'm using Solaris 2.7, and 7.0.2\ncompiled, installed and works without a hitch. The previous release did\nnot work correctly, but 7.0.2 works fine for me.\n\n----------------------------------------------------------------\nTravis Bauer | CS Grad Student | IU |www.cs.indiana.edu/~trbauer\n----------------------------------------------------------------\n\nOn Wed, 11 Oct 2000, Peter Eisentraut wrote:\n\n> Bruce Momjian writes:\n> \n> > I think our source tree has diverged quite a bit from 7.0.2. Could you\n> > grab a snapshot and let me know what needs changing for Solaris?\n> \n> These fixes were integrated. Let's remember to make a more detailed\n> \"Supported Platforms\" list next time, because this platform was listed and\n> it evidently didn't work.\n> \n> > > I'm in the process of fixing build problems in 7.0.2 for Solaris CC\n> > > (not GCC).\n> \n> -- \n> Peter Eisentraut [email protected] http://yi.org/peter-e/\n> \n\n", "msg_date": "Wed, 11 Oct 2000 16:54:10 -0500 (EST)", "msg_from": "Travis Bauer <[email protected]>", "msg_from_op": false, "msg_subject": "Re: 7.0.2 on Solaris" }, { "msg_contents": "Travis Bauer writes:\n\n> Which version of Solaris was he using? I'm using Solaris 2.7, and 7.0.2\n> compiled, installed and works without a hitch.\n\nHe was using the Solaris compiler rather than GCC, and that couldn't\npossibly have worked, as it turned out.\n\n-- \nPeter Eisentraut [email protected] http://yi.org/peter-e/\n\n", "msg_date": "Thu, 12 Oct 2000 16:46:08 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: 7.0.2 on Solaris" } ]
[ { "msg_contents": "> > > The symlinks wouldn't do any good for what Bruce had in \n> > > mind anyway (IIRC, he wanted to get useful per-database\n> > > numbers from \"du\").\n> > \n> > Our database design seems to be in the opposite direction\n> > if it is restricted for the convenience of command calls.\n> \n> Well, I don't see any reason not to use tablespace/database \n> rather than just tablespace. Seems having fewer files in each directory\n\nOnce again - ability to use different tablespaces (disks) for tables/indices\nin the same schema. Schemas must not dictate where to store objects <-\nbad design.\n\n> will be a little faster, and if we can make administration easier,\n> why not?\n\nBecause you'll not be able use du/ls once we'll implement new smgr anyway.\n\nAnd, btw, - for what are we going implement tablespaces? Just to have\nfewer files in each dir ?!\n\nVadim\n", "msg_date": "Tue, 27 Jun 2000 10:54:55 -0700", "msg_from": "\"Mikheev, Vadim\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Big 7.1 open items" }, { "msg_contents": "[ Charset ISO-8859-1 unsupported, converting... ]\n> > > > The symlinks wouldn't do any good for what Bruce had in \n> > > > mind anyway (IIRC, he wanted to get useful per-database\n> > > > numbers from \"du\").\n> > > \n> > > Our database design seems to be in the opposite direction\n> > > if it is restricted for the convenience of command calls.\n> > \n> > Well, I don't see any reason not to use tablespace/database \n> > rather than just tablespace. Seems having fewer files in each directory\n> \n> Once again - ability to use different tablespaces (disks) for tables/indices\n> in the same schema. Schemas must not dictate where to store objects <-\n> bad design.\n\nI am suggesting this symlink:\n\n\tln -s data/base/testdb/myspace /var/myspace/testdb\n\nrather than:\n\n\tln -s data/base/testdb/myspace /var/myspace\n\nTablespaces still sit inside database directories, it is just that it\npoints to a subdirectory of myspace, rather than myspace itself.\n\nAm I missing something?\n\n> \n> > will be a little faster, and if we can make administration easier,\n> > why not?\n> \n> Because you'll not be able use du/ls once we'll implement new smgr anyway.\n\nAt least du will work. I doubt we will be putting tables from different\ndatabases in the same file.\n\n> \n> And, btw, - for what are we going implement tablespaces? Just to have\n> fewer files in each dir ?!\n\nNo, I thought it was to split files across drives.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 27 Jun 2000 14:18:07 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" } ]
[ { "msg_contents": "> > > Well, I don't see any reason not to use tablespace/database \n> > > rather than just tablespace. Seems having fewer files in \n> > > each directory\n> > \n> > Once again - ability to use different tablespaces (disks) \n> > for tables/indices in the same schema. Schemas must not dictate\n> > where to store objects <- bad design.\n> \n> I am suggesting this symlink:\n> \n> \tln -s data/base/testdb/myspace /var/myspace/testdb\n> \n> rather than:\n> \n> \tln -s data/base/testdb/myspace /var/myspace\n> \n> Tablespaces still sit inside database directories, it is just that it\n> points to a subdirectory of myspace, rather than myspace itself.\n^^^^^^^^^^^\n\nDidn't you mean \n\nln -s /var/myspace/testdb data/base/testdb/myspace\n\n?\n\nI thought that you don't like symlinks from data/base/... This is\nhow I understood Tom' words:\n\n> The symlinks wouldn't do any good for what Bruce had in mind anyway\n> (IIRC, he wanted to get useful per-database numbers from \"du\").\n\nVadim\n", "msg_date": "Tue, 27 Jun 2000 12:22:13 -0700", "msg_from": "\"Mikheev, Vadim\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Big 7.1 open items" }, { "msg_contents": "> > I am suggesting this symlink:\n> > \n> > \tln -s data/base/testdb/myspace /var/myspace/testdb\n> > \n> > rather than:\n> > \n> > \tln -s data/base/testdb/myspace /var/myspace\n> > \n> > Tablespaces still sit inside database directories, it is just that it\n> > points to a subdirectory of myspace, rather than myspace itself.\n> ^^^^^^^^^^^\n\nSorry, I should have said a symlink sits in data/base and points to the\ntablespace. My issue is having it point to tablespace/database and not\njust tablespace/.\n\n> \n> Didn't you mean \n> \n> ln -s /var/myspace/testdb data/base/testdb/myspace\n> \n\nNo, sorry for the confusion.\n\n\n> ?\n> \n> I thought that you don't like symlinks from data/base/... This is\n> how I understood Tom' words:\n> \n> > The symlinks wouldn't do any good for what Bruce had in mind anyway\n> > (IIRC, he wanted to get useful per-database numbers from \"du\").\n\nNo, I want symlinks in data/base. What I wanted was to have\nper-database directories in every tablespace so I can do a 'du' in the\ntablespace directory to see how much of each database is the tablespace.\nIt also nicely partitions the files, makes the directories smaller, and\nprevents any possible file name conflict.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 27 Jun 2000 15:34:42 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" } ]
[ { "msg_contents": "> > Do we need *both* database & tablespace to find table file ?!\n> > Imho, database shouldn't be used...\n> \n> Then the system tables from different databases would collide.\n\nActually, if we're going to use unique-ids for file names\nthen we have to know how to get system file names anyway.\nHm, OID+VERSION would make our life easier... Hiroshi?\n\nVadim\n", "msg_date": "Tue, 27 Jun 2000 12:31:43 -0700", "msg_from": "\"Mikheev, Vadim\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Big 7.1 open items " }, { "msg_contents": "[ Charset ISO-8859-1 unsupported, converting... ]\n> > > Do we need *both* database & tablespace to find table file ?!\n> > > Imho, database shouldn't be used...\n> > \n> > Then the system tables from different databases would collide.\n> \n> Actually, if we're going to use unique-ids for file names\n> then we have to know how to get system file names anyway.\n> Hm, OID+VERSION would make our life easier... Hiroshi?\n\nI assume we were going to have a pg_class.relversion to do that, but\nthat is per-database because pg_class is per-database.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 27 Jun 2000 15:40:27 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" } ]
[ { "msg_contents": "> > > Then the system tables from different databases would collide.\n> > \n> > Actually, if we're going to use unique-ids for file names\n> > then we have to know how to get system file names anyway.\n> > Hm, OID+VERSION would make our life easier... Hiroshi?\n> \n> I assume we were going to have a pg_class.relversion to do that, but\n ^^^^^^^^\nPG_CLASS_OID.VERSION_ID...\n\nJust a clarification -:)\n\n> that is per-database because pg_class is per-database.\n\nVadim\n", "msg_date": "Tue, 27 Jun 2000 12:37:34 -0700", "msg_from": "\"Mikheev, Vadim\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Big 7.1 open items" }, { "msg_contents": "[ Charset ISO-8859-1 unsupported, converting... ]\n> > > > Then the system tables from different databases would collide.\n> > > \n> > > Actually, if we're going to use unique-ids for file names\n> > > then we have to know how to get system file names anyway.\n> > > Hm, OID+VERSION would make our life easier... Hiroshi?\n> > \n> > I assume we were going to have a pg_class.relversion to do that, but\n> ^^^^^^^^\n> PG_CLASS_OID.VERSION_ID...\n> \n> Just a clarification -:)\n\nI actually meant I thought we were going to have a pg_class column\ncalled relversion that held the currently active version for that\nrelation.\n\nYes, the file name will be pg_class_oid.version_id.\n\nIs that OK?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 27 Jun 2000 15:45:42 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" } ]
[ { "msg_contents": "> I actually meant I thought we were going to have a pg_class column\n> called relversion that held the currently active version for that\n> relation.\n> \n> Yes, the file name will be pg_class_oid.version_id.\n> \n> Is that OK?\n\nWe recently discussed pure *unique-id* file names...\n\nVadim\n \n", "msg_date": "Tue, 27 Jun 2000 12:42:35 -0700", "msg_from": "\"Mikheev, Vadim\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Big 7.1 open items" }, { "msg_contents": "[ Charset ISO-8859-1 unsupported, converting... ]\n> > I actually meant I thought we were going to have a pg_class column\n> > called relversion that held the currently active version for that\n> > relation.\n> > \n> > Yes, the file name will be pg_class_oid.version_id.\n> > \n> > Is that OK?\n> \n> We recently discussed pure *unique-id* file names...\n> \n\nWell, that would allow us to mix database files in the same directory,\nif we wanted to do that. My opinion it is better to keep databases in\nseparate directories in each tablespace for clarity and performance\nreasons.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 27 Jun 2000 15:52:40 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n> Well, that would allow us to mix database files in the same directory,\n> if we wanted to do that. My opinion it is better to keep databases in\n> separate directories in each tablespace for clarity and performance\n> reasons.\n\nOne reason not to do that is that we'd still have to special-case\nthe system-wide relations. If it's just tablespace and OID in the\npath, then the system-wide rels look just the same as any other rel\nas far as the low-level stuff is concerned. That would be nice.\n\nMy feeling about the \"clarity and performance\" issue is that if a\ndbadmin wants to keep track of database contents separately, he can\nput different databases' tables into different tablespaces to start\nwith. If he puts several tables into one tablespace, he's saying\nhe doesn't care about distinguishing their space usage. There's\nno reason for us to force an additional level of directory lookup\nto be done whether the admin wants it or not.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 27 Jun 2000 17:19:31 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items " }, { "msg_contents": "> Bruce Momjian <[email protected]> writes:\n> > Well, that would allow us to mix database files in the same directory,\n> > if we wanted to do that. My opinion it is better to keep databases in\n> > separate directories in each tablespace for clarity and performance\n> > reasons.\n> \n> One reason not to do that is that we'd still have to special-case\n> the system-wide relations. If it's just tablespace and OID in the\n> path, then the system-wide rels look just the same as any other rel\n> as far as the low-level stuff is concerned. That would be nice.\n\nYes, good point about pg_shadow. They don't have databases. How do we\nget multiple pg_class tables in the same directory? Is the\npg_class.relversion file a number like 1,2,3,4, or does it come out of\nsome global counter like oid. If so, we could put them in the same\ndirectory.\n\nShould we be concerned about performance when 10-20 database are sitting\nin the same directory? I am thinking about open() and other calls that\nscan the directory. Certainly shorter file names will help.\n\n> \n> My feeling about the \"clarity and performance\" issue is that if a\n> dbadmin wants to keep track of database contents separately, he can\n> put different databases' tables into different tablespaces to start\n> with. If he puts several tables into one tablespace, he's saying\n> he doesn't care about distinguishing their space usage. There's\n> no reason for us to force an additional level of directory lookup\n> to be done whether the admin wants it or not.\n\nOK.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 27 Jun 2000 17:23:49 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n> Yes, good point about pg_shadow. They don't have databases. How do we\n> get multiple pg_class tables in the same directory? Is the\n> pg_class.relversion file a number like 1,2,3,4, or does it come out of\n> some global counter like oid. If so, we could put them in the same\n> directory.\n\nI think we could get away with insisting that each database store its\npg_class and friends in a separate tablespace (physically distinct\ndirectory) from any other database. That gets around the OID conflict.\n\nIt's still an open question whether OID+version is better than\nunique-ID for naming files that belong to different versions of the\nsame relation. I can see arguments on both sides.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 27 Jun 2000 17:30:17 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items " }, { "msg_contents": "Tom Lane wrote:\n\n> Bruce Momjian <[email protected]> writes:\n> > Yes, good point about pg_shadow. They don't have databases. How do we\n> > get multiple pg_class tables in the same directory? Is the\n> > pg_class.relversion file a number like 1,2,3,4, or does it come out of\n> > some global counter like oid. If so, we could put them in the same\n> > directory.\n>\n> I think we could get away with insisting that each database store its\n> pg_class and friends in a separate tablespace (physically distinct\n> directory) from any other database. That gets around the OID conflict.\n>\n> It's still an open question whether OID+version is better than\n> unique-ID for naming files that belong to different versions of the\n> same relation. I can see arguments on both sides.\n>\n\nI don't stick to unique-ID. My main point has always been the\ntransactional control of file allocation change.\nHowever *VERSION(_ID)* may be misleading because it couldn't\nmean the version of pg_class tuples.\n\nRegards.\n\nHiroshi Inoue\[email protected]\n\n\n", "msg_date": "Wed, 28 Jun 2000 08:16:27 +0900", "msg_from": "Hiroshi Inoue <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" }, { "msg_contents": "Bruce Momjian writes:\n\n> How do we get multiple pg_class tables in the same directory?\n\npg_class.DBOID\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n\n", "msg_date": "Wed, 28 Jun 2000 20:37:44 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" } ]
[ { "msg_contents": "This is a bit off topic but email to the author of the trigger bounced..\n\nIn the fti.c (full text index) trigger I see..\n\nsprintf(query, \"INSERT INTO %s (string, id) VALUES ($1, $2)\", indexname);\n\nbelow.. I need to add something else here.. Basically I'm eliminating the\nneed for the text table, I'm going to integrate the fti indexed table into\nan existing one I have but I need to insert naother ID into the resumes_fti\n(my indexed fti table)..\n\nThe above makes the query string, as I can see but what in the world is the\n$1 $2 and where did they come from? Since it looks like that's how the\nvalues of string and id are put into the query, I need to know so I can add\none more in there..\n\nBelow is the whole function I'm referring to in fti.c\n\nThanks!!!\n\n\nif (isinsert)\n {\n char *substring,\n *column;\n void *pplan;\n Oid *argtypes;\n Datum values[2];\n int colnum;\n struct varlena *data;\n EPlan *plan;\n\n sprintf(query, \"I%s$%s\", args[0], args[1]);\n plan = find_plan(query, &InsertPlans, &nInsertPlans);\n\n /* no plan yet, so allocate mem for argtypes */\n if (plan->nplans <= 0)\n {\n argtypes = (Oid *) palloc(2 * sizeof(Oid));\n\n argtypes[0] = VARCHAROID; /* create table t_name (string\n * varchar, */\n argtypes[1] = OIDOID; /* id oid); */\n\n /* prepare plan to gain speed */\n sprintf(query, \"INSERT INTO %s (string, id) VALUES ($1, $2)\",\n indexname);\n pplan = SPI_prepare(query, 2, argtypes);\n if (!pplan)\n elog(ERROR, \"Full Text Indexing: SPI_prepare returned NULL \"\n \"in insert\");\n\n pplan = SPI_saveplan(pplan);\n if (pplan == NULL)\n elog(ERROR, \"Full Text Indexing: SPI_saveplan returned NULL\"\n \" in insert\");\n\n plan->splan = (void **) malloc(sizeof(void *));\n *(plan->splan) = pplan;\n plan->nplans = 1;\n }\n\n\n\n", "msg_date": "Tue, 27 Jun 2000 15:54:11 -0400", "msg_from": "\"Mitch Vincent\" <[email protected]>", "msg_from_op": true, "msg_subject": "Trigger question" } ]
[ { "msg_contents": " Date: Tuesday, June 27, 2000 @ 23:31:54\nAuthor: tgl\n\nUpdate of /home/projects/pgsql/cvsroot/pgsql/src/backend/postmaster\n from hub.org:/home/projects/pgsql/tmp/cvs-serv57303/src/backend/postmaster\n\nModified Files:\n\tpostmaster.c \n\n----------------------------- Log Message -----------------------------\n\nFirst phase of memory management rewrite (see backend/utils/mmgr/README\nfor details). It doesn't really do that much yet, since there are no\nshort-term memory contexts in the executor, but the infrastructure is\nin place and long-term contexts are handled reasonably. A few long-\nstanding bugs have been fixed, such as 'VACUUM; anything' in a single\nquery string crashing. Also, out-of-memory is now considered a\nrecoverable ERROR, not FATAL.\nEliminate a large amount of crufty, now-dead code in and around\nmemory management.\nFix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and\nbackend startup.\n", "msg_date": "Tue, 27 Jun 2000 23:31:56 -0400 (EDT)", "msg_from": "Tom Lane <tgl>", "msg_from_op": true, "msg_subject": "pgsql/src/backend/postmaster (postmaster.c)" }, { "msg_contents": "PGPORT environment variable is ignored again.\nPostPortName seems to be reset in ResetAllOptions().\n\nRegards.\n\nHiroshi Inoue\[email protected]\n\n\n> -----Original Message-----\n> From: [email protected] [mailto:pgsql-committers-owner@hub.\n> org]On Behalf Of Tom Lane\n> Sent: Wednesday, June 28, 2000 12:32 PM\n> To: [email protected]\n> Subject: [COMMITTERS] pgsql/src/backend/postmaster (postmaster.c)\n> \n> \n> Date: Tuesday, June 27, 2000 @ 23:31:54\n> Author: tgl\n> \n> Update of /home/projects/pgsql/cvsroot/pgsql/src/backend/postmaster\n> from \n> hub.org:/home/projects/pgsql/tmp/cvs-serv57303/src/backend/postmaster\n> \n> Modified Files:\n> \tpostmaster.c \n> \n> ----------------------------- Log Message -----------------------------\n> \n> First phase of memory management rewrite (see backend/utils/mmgr/README\n> for details). It doesn't really do that much yet, since there are no\n> short-term memory contexts in the executor, but the infrastructure is\n> in place and long-term contexts are handled reasonably. A few long-\n> standing bugs have been fixed, such as 'VACUUM; anything' in a single\n> query string crashing. Also, out-of-memory is now considered a\n> recoverable ERROR, not FATAL.\n> Eliminate a large amount of crufty, now-dead code in and around\n> memory management.\n> Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and\n> backend startup.\n> \n", "msg_date": "Mon, 3 Jul 2000 12:57:56 +0900", "msg_from": "\"Hiroshi Inoue\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: pgsql/src/backend/postmaster (postmaster.c)" }, { "msg_contents": "\"Hiroshi Inoue\" <[email protected]> writes:\n> PGPORT environment variable is ignored again.\n> PostPortName seems to be reset in ResetAllOptions().\n\nThink you want to complain to Peter, not me. I haven't been touching\nany options-handling code...\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 02 Jul 2000 23:59:51 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: pgsql/src/backend/postmaster (postmaster.c) " }, { "msg_contents": "> -----Original Message-----\n> From: Tom Lane [mailto:[email protected]]\n> \n> \"Hiroshi Inoue\" <[email protected]> writes:\n> > PGPORT environment variable is ignored again.\n> > PostPortName seems to be reset in ResetAllOptions().\n> \n> Think you want to complain to Peter, not me. I haven't been touching\n> any options-handling code...\n>\n\nSeems you have changed the position of ResetAllOptions() as follows.\n\n@@ -378,15 +378,38 @@\n \t */\n \tumask((mode_t) 0077);\n \n-\tResetAllOptions();\n-\n \tMyProcPid = getpid();\n+\n+\t/*\n+\t * Fire up essential subsystems: error and memory management\n+\t */\n+\tEnableExceptionHandling(true);\n+\tMemoryContextInit();\n+\n+\t/*\n+\t * By default, palloc() requests in the postmaster will be allocated\n+\t * in the PostmasterContext, which is space that can be recycled by\n+\t * backends. Allocated data that needs to be available to backends\n+\t * should be allocated in TopMemoryContext.\n+\t */\n+\tPostmasterContext = AllocSetContextCreate(TopMemoryContext,\n+\t\t\t\t\t\t\t\t\t\t\t \"Postmaster\",\n+\t\t\t\t\t\t\t\t\t\t\t ALLOCSET_DEFAULT_MINSIZE,\n+\t\t\t\t\t\t\t\t\t\t\t ALLOCSET_DEFAULT_INITSIZE,\n+\t\t\t\t\t\t\t\t\t\t\t ALLOCSET_DEFAULT_MAXSIZE);\n+\tMemoryContextSwitchTo(PostmasterContext);\n+\n+\t/*\n+\t * Options setup\n+\t */\n \tif (getenv(\"PGDATA\"))\n \t\tDataDir = strdup(getenv(\"PGDATA\")); /* default value */\n \n \tif (getenv(\"PGPORT\"))\n \t\tPostPortName = atoi(getenv(\"PGPORT\"));\n \n+\tResetAllOptions();\n+\n\n", "msg_date": "Mon, 3 Jul 2000 13:26:59 +0900", "msg_from": "\"Hiroshi Inoue\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: pgsql/src/backend/postmaster (postmaster.c) " }, { "msg_contents": "\"Hiroshi Inoue\" <[email protected]> writes:\n> Seems you have changed the position of ResetAllOptions() as follows.\n\n> +\t/*\n> +\t * Options setup\n> +\t */\n> \tif (getenv(\"PGDATA\"))\n> \t\tDataDir = strdup(getenv(\"PGDATA\")); /* default value */\n \n> \tif (getenv(\"PGPORT\"))\n> \t\tPostPortName = atoi(getenv(\"PGPORT\"));\n \n> +\tResetAllOptions();\n\nHmm, actually I was trying to move it to after the memory-context\ncreation code, but I guess your point is that ResetAllOptions() is\nignoring these env-var-based settings. Seems like the correct fix\nis that guc.c ought to be handling this logic, not postmaster.c.\nPeter, what do you think?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 00:32:22 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: pgsql/src/backend/postmaster (postmaster.c) " }, { "msg_contents": "On Mon, 3 Jul 2000, Tom Lane wrote:\n\n> Hmm, actually I was trying to move it to after the memory-context\n> creation code, but I guess your point is that ResetAllOptions() is\n> ignoring these env-var-based settings. Seems like the correct fix\n> is that guc.c ought to be handling this logic, not postmaster.c.\n> Peter, what do you think?\n\nI concur.\n\n-- \nPeter Eisentraut Sernanders vaeg 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Mon, 3 Jul 2000 08:37:14 -0400 (EDT)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: pgsql/src/backend/postmaster (postmaster.c) " }, { "msg_contents": "[email protected] writes:\n>> ... Seems like the correct fix\n>> is that guc.c ought to be handling this logic, not postmaster.c.\n>> Peter, what do you think?\n\n> I concur.\n\nOK, do you want to fix it, or shall I?\n\nBTW, it is now possible for guc.c to use palloc in TopMemoryContext,\nif you want to do that. There's no real functional difference between\nthat and malloc --- except that MemoryContextStats() could include\nthe space used by guc.c --- but if it's just irritating you to use\nmalloc then you could change it.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 11:16:14 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: pgsql/src/backend/postmaster (postmaster.c) " }, { "msg_contents": "Tom Lane writes:\n\n> OK, do you want to fix it, or shall I?\n\nI will. I got some half-finished code lying around that touches that.\n\n> BTW, it is now possible for guc.c to use palloc in TopMemoryContext,\n> if you want to do that. There's no real functional difference between\n> that and malloc --- except that MemoryContextStats() could include\n> the space used by guc.c --- but if it's just irritating you to use\n> malloc then you could change it.\n\nWell, I would really like a way to register an entire memory context to be\nswept away by elog(ERROR). But that's probably too specialized. But yes, I\nwill look into that.\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Mon, 3 Jul 2000 18:47:26 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: pgsql/src/backend/postmaster (postmaster.c) " }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n> Well, I would really like a way to register an entire memory context to be\n> swept away by elog(ERROR). But that's probably too specialized.\n\nBe careful what you ask for, you might get it ;-).\n\nIf you really truly want a context that's destroyed by elog(ERROR),\nand not by any other event, then create it as a child of ErrorContext,\nand that's what will happen. But I suspect what you really want is a\ncontext that's a child of TopTransactionContext and will go away at\neither transaction commit or transaction abort.\n\nA difficulty I have been noticing with some of these special-purpose\ncontexts is that you'd like to keep a pointer to them in some global\nvariable, but there's no easy way to determine whether that pointer is\ncurrently valid (eg, whether the context has yet been created during\nthe current transaction). This would be particularly nasty with\nsomething like a child of ErrorContext --- has an error occurred since\nyou last used the context?\n\nAn example of how we currently have to get around this is\nbackend/commands/trigger.c's DeferredTriggerAbortXact() routine, which\nxact.c has to know to call at transaction abort. All it does is delete\ntrigger.c's private per-transaction context and reset the global pointer\nto same. Deleting the context is useless because that'd happen\nautomatically anyway when TopTransactionContext gets zapped. So,\nresetting the pointer is the only reason for this function to exist.\n\nI've been toying with the notion of adding an optional \"notify\" callback\nfunction to memory contexts. If specified, it'd be called immediately\nbefore resetting or deleting the context. This'd give you a clean way\nto reset pointers etc.\n\nComments anyone?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 15:20:31 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Memory-context slinging" }, { "msg_contents": "Tom Lane writes:\n\n> If you really truly want a context that's destroyed by elog(ERROR),\n> and not by any other event, then create it as a child of ErrorContext,\n> and that's what will happen. But I suspect what you really want is a\n> context that's a child of TopTransactionContext and will go away at\n> either transaction commit or transaction abort.\n\nNo, actually I wanted the former. Notice how the configuration file is\nfirst read in completely as a linked list. When something goes wrong\n(garbage in the file) I need to free the list (and other things) before\ncalling elog. The alternative would be to create a separate memory context\nfor this processing, but that's more of a purity issue, it wouldn't make\nit work better, AFAICT.\n\n> A difficulty I have been noticing with some of these special-purpose\n> contexts is that you'd like to keep a pointer to them in some global\n> variable, but there's no easy way to determine whether that pointer is\n> currently valid\n\nShouldn't it be NULL if it's not valid?\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Tue, 4 Jul 2000 21:57:37 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Memory-context slinging" }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n> Tom Lane writes:\n>> If you really truly want a context that's destroyed by elog(ERROR),\n>> and not by any other event, then create it as a child of ErrorContext,\n>> and that's what will happen. But I suspect what you really want is a\n>> context that's a child of TopTransactionContext and will go away at\n>> either transaction commit or transaction abort.\n\n> No, actually I wanted the former. Notice how the configuration file is\n> first read in completely as a linked list. When something goes wrong\n> (garbage in the file) I need to free the list (and other things) before\n> calling elog.\n\nAh, I see: normal exit from the routine would be to copy the validated\nlist to permanent storage and free the temporary junk, but in case of\nerror exit you'd like to have the list thrown away automatically.\n\nAs a general thing I'd still recommend a child of a transaction context,\nbecause that guarantees the temp storage will be cleaned up either way;\nyou can't accidentally forget about it. However, for guc.c's purposes\na child of ErrorContext is probably better, because (I think) guc.c\nneeds to be able to run outside of any transaction. You'll still leak\nmemory if you forget to free the context during normal exit, but at\nleast this technique handles the error-case cleanup for you.\n\n> The alternative would be to create a separate memory context\n> for this processing, but that's more of a purity issue, it wouldn't make\n> it work better, AFAICT.\n\nIf you leak permanent storage upon error, that's a bug that needs to be\nfixed, IMHO, not just a \"purity\" issue.\n\n>> A difficulty I have been noticing with some of these special-purpose\n>> contexts is that you'd like to keep a pointer to them in some global\n>> variable, but there's no easy way to determine whether that pointer is\n>> currently valid\n\n> Shouldn't it be NULL if it's not valid?\n\nExactly. The problem is how to make sure that that's true, if the\ncontext gets discarded by general-purpose memory cleanup code that has\nno idea you've got a pointer stashed away somewhere. You need a hook\nthat can clear your pointer to NULL at the right time. Basically this\nmechanism could take the place of (some of the) hard-wired calls in\nxact.c's abort processing, which would be a Good Thing --- consider\ndynamically loaded code that has no prospect of being directly known\nby xact.c.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 20:25:53 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Memory-context slinging " } ]
[ { "msg_contents": "> Yes, the file name will be pg_class_oid.version_id.\n\nWhat about segmented files (ie: those over 1Gb)?\n\n--\nPeter Mount\nEnterprise Support\nMaidstone Borough Council\nAny views stated are my own, and not those of Maidstone Borough Council\n\n", "msg_date": "Wed, 28 Jun 2000 08:21:00 +0100", "msg_from": "Peter Mount <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Big 7.1 open items" }, { "msg_contents": "Peter Mount <[email protected]> writes:\n>> Yes, the file name will be pg_class_oid.version_id.\n\n> What about segmented files (ie: those over 1Gb)?\n\nSeparate issue. Putting the segment number into the filename is\na bad idea because it doesn't give you any way to spread multiple\nsegments of a big table across filesystems. What's currently being\ndiscussed is paths that look like\n\n\tsomething/SEGNO/RELATIONOID.VERSIONID\n\nThis lets you control space allocation by making the SEGNO\nsubdirectories be symlinks to various places.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 28 Jun 2000 11:37:06 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items " } ]
[ { "msg_contents": "\n> I am suggesting this symlink:\n> \n> \tln -s data/base/testdb/myspace /var/myspace/testdb\n> \n> rather than:\n> \n> \tln -s data/base/testdb/myspace /var/myspace\n\nI guess on similar reasoning I would suggest inserting the extent\nsubdirectory, because it would be easier to create different \nfilesystems for them.\n\n\tln -s data/base/testdb/myspace/extent1 /var/myspace/extent1/testdb\n\nAndreas\n", "msg_date": "Wed, 28 Jun 2000 09:27:27 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: Big 7.1 open items" }, { "msg_contents": "> > I am suggesting this symlink:\n> > \n> > \tln -s data/base/testdb/myspace /var/myspace/testdb\n> > \n> > rather than:\n> > \n> > \tln -s data/base/testdb/myspace /var/myspace\n> \n> I guess on similar reasoning I would suggest inserting the extent\n> subdirectory, because it would be easier to create different \n> filesystems for them.\n> \n> \tln -s data/base/testdb/myspace/extent1 /var/myspace/extent1/testdb\n\nThe idea was to put the main files in the directory, and create Extent2,\nExtent3 directories for the extents.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 28 Jun 2000 10:14:14 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Big 7.1 open items" } ]
[ { "msg_contents": "\n> > I've been assuming that we would create a separate tablespace for\n> > each database, which would be the location of that database's\n> > system tables. It's probably also the default tablespace for user\n> > tables created in that database, though it wouldn't have to be.\n\nI think I would prefer the ability to place more than one database into \nthe same tablespace.\n\nAndreas\n", "msg_date": "Wed, 28 Jun 2000 09:36:08 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: Big 7.1 open items" }, { "msg_contents": "Zeugswetter Andreas SB <[email protected]> writes:\n>>>> I've been assuming that we would create a separate tablespace for\n>>>> each database, which would be the location of that database's\n>>>> system tables. It's probably also the default tablespace for user\n>>>> tables created in that database, though it wouldn't have to be.\n\n> I think I would prefer the ability to place more than one database into \n> the same tablespace.\n\nYou can put user tables from multiple databases into the same\ntablespace, under this proposal. Just not system tables.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 28 Jun 2000 10:57:52 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Big 7.1 open items " } ]
[ { "msg_contents": "\n> > I am suggesting this symlink:\n> > \n> > \tln -s data/base/testdb/myspace /var/myspace/testdb\n> > \n> > rather than:\n> > \n> > \tln -s data/base/testdb/myspace /var/myspace\n> \n> I guess on similar reasoning I would suggest inserting the extent\n> subdirectory, because it would be easier to create different \n> filesystems for them.\n> \n> \tln -s data/base/testdb/myspace/extent1 /var/myspace/extent1/testdb\n\nGrmpf, I meant:\n\tln -s /var/myspace/extent1/testdb data/base/testdb/myspace/extent1\n\nAndreas \n", "msg_date": "Wed, 28 Jun 2000 09:48:54 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: Big 7.1 open items" }, { "msg_contents": "Zeugswetter Andreas SB <[email protected]> writes:\n>> I guess on similar reasoning I would suggest inserting the extent\n>> subdirectory, because it would be easier to create different \n>> filesystems for them.\n>> \n>> ln -s data/base/testdb/myspace/extent1 /var/myspace/extent1/testdb\n\n> Grmpf, I meant:\n> \tln -s /var/myspace/extent1/testdb data/base/testdb/myspace/extent1\n\nThat would mean more bookkeeping: everytime you add an extent to a\ntablespace, you'd have to go around and find all the referencing\ndatabases and add a symlink to each one.\n\nBut I think the direction we're headed in is that the data/base/DBNAME\ndirectories are going to disappear entirely, so this argument about\nwhat symlinks they need to have is a bit pointless ;-). Databases\nare going to become a higher-level concept that's not directly reflected\nin the physical layout.\n\nThe way I'm currently envisioning it is that we have paths like\n\n\tdata/spaces/TABLESPACE/EXTENT/RELATION.VERSION\n\n(ignoring the details about whether we use names or OIDs and which\ndirectory levels might be symlinks). Since we will require each logical\ndatabase to have a distinct \"home tablespace\" in which its system tables\nlive, that \"home tablespace\" can be the runtime working directory for\nbackends running in that database. If you like you can think of the\nhome tablespace directory as being equivalent to the old database\ndirectory, but it's really a different notion --- and in particular,\nit's got nothing to do with how the backend addresses tables that are\nin other tablespaces.\n\nBTW, it occurs to me that we ought to have some frammish whereby temp\nfiles and tables created by backends running in a particular database\ncan be directed to a different tablespace. If we do nothing, then\nthey'd always appear in the database's home tablespace, but I can sure\nsee a dbadmin wanting to push his large sort temp files off to someplace\nelse...\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 28 Jun 2000 11:13:47 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Big 7.1 open items " }, { "msg_contents": "> BTW, it occurs to me that we ought to have some frammish whereby temp\n> files and tables created by backends running in a particular database\n> can be directed to a different tablespace. If we do nothing, then\n> they'd always appear in the database's home tablespace, but I can sure\n> see a dbadmin wanting to push his large sort temp files off to someplace\n> else...\n\nSort tables are an area Ingres does round-robin so the tape file can be\non different drives.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 28 Jun 2000 11:15:25 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Big 7.1 open items" } ]
[ { "msg_contents": " \n> Zeugswetter Andreas SB <[email protected]> writes:\n> >> Thomas Lockhart <[email protected]> writes:\n> >>>> create table t1 (x int);\n> >>>> create function foo (out int) returns int as ...;\n> >>>> select foo(x) from t1;\n> >> \n> >>>> will give two columns for the result.\n> >> \n> >> You've *got* to be kidding.\n> >> \n> >> To name just one problem with that, where do I put an AS to relabel\n> >> the extra column?\n> \n> > It gets the name of the variable ( here \"out\"), no way to rename it,\n> > but no problem since you can name the variable however you like.\n> \n> What? OUT is a keyword there; I don't see any explicit name for the\n> function's formal parameter at all.\n\nTo not look *that* stupid the mistake I made was:\n\tI thought the parameter name was not optional\n\nThe default for in|out|\"in out\" is \"in\" thus I read the function as foo\n(\"out\" IN int)\n(of course if it was an IN var it does'nt produce an out column and \nthus need no label, mistake 2)\n\nIn Oracle the syntax is:\n\tvarname [IN|OUT|IN OUT] type\n\nAndreas\n", "msg_date": "Wed, 28 Jun 2000 10:16:16 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: SQL99 functions " } ]
[ { "msg_contents": "\n> I don't suppose there is a 'disable constraints' command? Or \n> the ability to\n> set all constraints as deferrred until commit-time?\n\nI think this is already available, but I am not sure it is suited for \na restore, since it uses memory for the constraint checking queue.\n\nAndreas\n", "msg_date": "Wed, 28 Jun 2000 10:26:22 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: Proposal: More flexible backup/restore via pg_d\n\tump" } ]
[ { "msg_contents": "\n> > > > The symlinks wouldn't do any good for what Bruce had in \n> > > > mind anyway (IIRC, he wanted to get useful per-database\n> > > > numbers from \"du\").\n> > > \n> > > Our database design seems to be in the opposite direction\n> > > if it is restricted for the convenience of command calls.\n> > \n> > Well, I don't see any reason not to use tablespace/database \n> > rather than just tablespace. Seems having fewer files in \n> each directory\n> \n> Once again - ability to use different tablespaces (disks) for \n> tables/indices\n> in the same schema. Schemas must not dictate where to store objects <-\n> bad design.\n\nCan we agree, that the schema is below the database hierarchy, and thus\nis something different than a database ?\nA table under another schema will simply get another oid, and thus no\ncollision.\nBut I agree that schema should not dictate storage location,\nbut the schema might imply a default storage location like in Oracle\n(default tablespaces for a user).\n\n> > will be a little faster, and if we can make administration easier,\n> > why not?\n> \n> Because you'll not be able use du/ls once we'll implement new \n> smgr anyway.\n\nLeaving the file per table design imho does imply an order of magnitude \nincrease in the impact of errors in the smgr. Now an error is likely to\ndestroy \none table only, then it can destroy a whole tablespace.\nBut I am still a fan of the single file/raw device per tablespace design,\nsince it can remove a lot of the OS overhead. \n\n> And, btw, - for what are we going implement tablespaces? Just to have\n> fewer files in each dir ?!\n\nNo, I guess the idea is to have a tool to manipulate physical distribution \nof objects (which disk, which filesystem ...)\n\nAndreas\n", "msg_date": "Wed, 28 Jun 2000 11:11:35 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: Big 7.1 open items" }, { "msg_contents": "Zeugswetter Andreas SB wrote:\n\n> > > > > The symlinks wouldn't do any good for what Bruce had in\n> > > > > mind anyway (IIRC, he wanted to get useful per-database\n> > > > > numbers from \"du\").\n> > > >\n> > > > Our database design seems to be in the opposite direction\n> > > > if it is restricted for the convenience of command calls.\n> > >\n> > > Well, I don't see any reason not to use tablespace/database\n> > > rather than just tablespace. Seems having fewer files in\n> > each directory\n> >\n> > Once again - ability to use different tablespaces (disks) for\n> > tables/indices\n> > in the same schema. Schemas must not dictate where to store objects <-\n> > bad design.\n>\n> Can we agree, that the schema is below the database hierarchy, and thus\n> is something different than a database ?\n\nI don't think we have a common understanding for PG's *database*\n(created by createdb). Every one seems to have his own *database*.\n\nAccording to your another posting,your *database* hierarchy is\n instance -> database -> schema -> object\nlike Oracle.\n\nHowever SQL92 seems to have another hierarchy:\n cluster -> catalog -> schema -> object\nand dot notation catalog.schema.object could be used.\n\nI couldn't find clear correspondense between PG's *database*\nand above hierarchy because we have no dot notation for\nobjects currently.\n\n\n>\n> A table under another schema will simply get another oid, and thus no\n> collision.\n> But I agree that schema should not dictate storage location,\n> but the schema might imply a default storage location like in Oracle\n> (default tablespaces for a user).\n\nAFAIK,schema is independent from user in SQL92.\nSo default_tablespace_per_user doesn't necessarily imply\ndefault_tablespace_per_schema.\n\nRegards.\n\nHiroshi Inoue\n\n", "msg_date": "Wed, 28 Jun 2000 19:47:43 +0900", "msg_from": "Hiroshi Inoue <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Big 7.1 open items" }, { "msg_contents": "Hiroshi Inoue writes:\n\n> According to your another posting,your *database* hierarchy is\n> instance -> database -> schema -> object\n> like Oracle.\n> \n> However SQL92 seems to have another hierarchy:\n> cluster -> catalog -> schema -> object\n> and dot notation catalog.schema.object could be used.\n\nFYI:\n\nAn \"instance\" is a \"cluster\". I don't know where the word instance came\nfrom, the docs sometimes call it \"installation\" or \"site\", which is even\nworse. I have been using \"database cluster\" for the latest documentation\nwork. My dictionary defines a cluster as \"a group of things gathered or\noccurring closely together\", which is what this is. Call it a \"data area\"\nor an \"initdb'ed thing\", etc.\n\nA \"catalog\" can be equated with our \"database\". The method of creating\ncatalogs is implementation defined, so our CREATE DATABASE command is in\nperfect compliance with the standard. We don't support the\ncatalog.schema.object notation but that notation only makes sense when you\ncan access more than one catalog at a time. We don't allow that and SQL\ndoesn't require it. We could allow that notation and throw an error when\nthe catalog name doesn't match the current database, but that's mere\ncosmetic work.\n\nIn entry level SQL 92, a \"schema\" is essentially the same as table\nownership. You can execute the command CREATE SCHEMA AUTHORIZATION\n\"peter\", which means that user \"peter\" (where he came from is\n\"implementation-defined\") can now create tables under his name. There is\nno such thing as a table owner, there's the \"containing schema\" and its\nowner. The tables \"peter\" creates can then be referenced by the dotted\nnotation. But it is not correct to equate this with CREATE USER. Even if\nthere was no schema for \"peter\" he could still connect and query other\npeople's tables.\n\nMoving beyond SQL 92 you can also create schemas with a different name\nthan your user name. This is merely a little more naming flexibility.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Thu, 29 Jun 2000 19:27:15 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Big 7.1 open items" }, { "msg_contents": "Peter Eisentraut wrote:\n\n> Hiroshi Inoue writes:\n>\n> > According to your another posting,your *database* hierarchy is\n> > instance -> database -> schema -> object\n> > like Oracle.\n> >\n> > However SQL92 seems to have another hierarchy:\n> > cluster -> catalog -> schema -> object\n> > and dot notation catalog.schema.object could be used.\n>\n> FYI:\n\nThanks.\nI'm asking to all what our *DATABASE* is.\nDifferent from you,I couldn't see any decisive feature in our *DATABASE*.\n\n>\n>\n> An \"instance\" is a \"cluster\". I don't know where the word instance came\n\nI could find the word in Oracle.\nIMHO,it corresponds to our initdb'ed thing(a postmaster controls).\n\n>\n> from, the docs sometimes call it \"installation\" or \"site\", which is even\n> worse. I have been using \"database cluster\" for the latest documentation\n> work. My dictionary defines a cluster as \"a group of things gathered or\n> occurring closely together\", which is what this is. Call it a \"data area\"\n> or an \"initdb'ed thing\", etc.\n>\n\nSQL92 seems to say that a cluster corresponds to a target of connection\nand has no name(after connection was established). Isn't it same as our\n*DATABASE* ?\n\n>\n> A \"catalog\" can be equated with our \"database\". The method of creating\n> catalogs is implementation defined, so our CREATE DATABASE command is in\n> perfect compliance with the standard. We don't support the\n> catalog.schema.object notation but that notation only makes sense when you\n> can access more than one catalog at a time.\n\nYes,it's most essential that we couldn't access more than one catalog.\nThis means that we have only one (noname) \"catalog\" per \"cluster\".\n\n> We don't allow that and SQL\n> doesn't require it. We could allow that notation and throw an error when\n> the catalog name doesn't match the current database, but that's mere\n> cosmetic work.\n>\n> In entry level SQL 92, a \"schema\" is essentially the same as table\n> ownership. You can execute the command CREATE SCHEMA AUTHORIZATION\n> \"peter\", which means that user \"peter\" (where he came from is\n> \"implementation-defined\") can now create tables under his name. There is\n> no such thing as a table owner, there's the \"containing schema\" and its\n> owner. The tables \"peter\" creates can then be referenced by the dotted\n> notation. But it is not correct to equate this with CREATE USER. Even if\n> there was no schema for \"peter\" he could still connect and query other\n> people's tables.\n>\n\nI've used *username* \"schema\"s in Oracle for a long time but I've never\nthought that it's the essence of \"schema\". If I recoginze correctly,the\nconcept of \"catalog\" hasn't necessarily been important while \"schema\"\n= \"user\". The conflict of \"schema\" name is equivalent to the conflict\nof \"user\" name if \"schema\" = \"user\". IMHO,SQL92 has required the\nconcept of \"catalog\" because \"schema\" has been changed to be\nindependent of \"user\".\n\nAnyway in current PG \"cluster\":\"catalog\":\"schema\"=1:1:1(0) and\nour *DATABASE* is an only confusing concept in the hierarchy..\n\nRegards,\n\nHiroshi Inoue\n\n\n", "msg_date": "Fri, 30 Jun 2000 08:59:49 +0900", "msg_from": "Hiroshi Inoue <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Big 7.1 open items" } ]
[ { "msg_contents": "\n> AFAIK,schema is independent from user in SQL92.\n> So default_tablespace_per_user doesn't necessarily imply\n> default_tablespace_per_schema.\n\nWell, sombody must be interpreting this wrong, because \nin Informix and Oracle the schema corresponds to the owner\nand they say they conform to ansi in this regard.\nIn both, a user can access other schemas or switch to another\nschema, so in that sense you could say that the schema is \nindependent of users.\n\n> However SQL92 seems to have another hierarchy:\n> \tcluster -> catalog -> schema -> object\n\nI would say our \"database\" corresponds to \"catalog\" and \n\"instance\" corresponds to \"cluster\" in the SQL92 hierarchy.\nInstance is probably a bad wording in respect to multiple\nmachine clusters where you can access all objects \nfrom every node.\nDatabase was probably not used, because this is often used \nto describe the whole hierarchy.\n\n> I couldn't find clear correspondense between PG's *database*\n> and above hierarchy because we have no dot notation for\n> objects currently.\n\nThis will definitely be a problem because of our current nested dot\ninterpretation towards functions taking one opaque or _class_type\nargument.\n\nAndreas\n", "msg_date": "Wed, 28 Jun 2000 14:07:33 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: Big 7.1 open items" }, { "msg_contents": "On Wed, Jun 28, 2000 at 02:07:33PM +0200, Zeugswetter Andreas SB wrote:\n> \n> > AFAIK,schema is independent from user in SQL92.\n> > So default_tablespace_per_user doesn't necessarily imply\n> > default_tablespace_per_schema.\n> \n> Well, sombody must be interpreting this wrong, because \n> in Informix and Oracle the schema corresponds to the owner\n> and they say they conform to ansi in this regard.\n\nTo quote from the SQL92 standard for CERATE SCHEMA:\n\n <schema definition> ::=\n CREATE SCHEMA <schema name clause>\n [ <schema character set specification> ]\n [ <schema element>... ]\n\n <schema name clause> ::=\n <schema name>\n | AUTHORIZATION <schema authorization identifier>\n | <schema name> AUTHORIZATION <schema authorization identifier>\n\n\n 1) If <schema name> is not specified, then a <schema name> equal to\n <schema authorization identifier> is implicit.\n\n 2) If AUTHORIZATION <schema authorization identifier> is not speci-\n fied, then\n\n Case:\n\n a) If the <schema definition> is contained in a <module> that\n has a <module authorization identifier> specified, then an\n <authorization identifier> equal to that <module authoriza-\n tion identifier> is implicit for the <schema definition>.\n\n b) Otherwise, an <authorization identifier> equal to the SQL-\n session <authorization identifier> is implicit.\n\nSo, we see that the SQL92 default fora schema is the session username.\n\n> In both, a user can access other schemas or switch to another\n> schema, so in that sense you could say that the schema is \n> independent of users.\n\nNot only in a sense, it is in fact.\n> \n> This will definitely be a problem because of our current nested dot\n> interpretation towards functions taking one opaque or _class_type\n> argument.\n\nRight. If we're going to support SQL92 dot notation (which I think we\nshould) we'll either need to lose the function notion completely, or\ncome up with some really clever hack about applying them in order.\n\nRoss\n-- \nRoss J. Reedstrom, Ph.D., <[email protected]> \nNSBRI Research Scientist/Programmer\nComputer and Information Technology Institute\nRice University, 6100 S. Main St., Houston, TX 77005\n", "msg_date": "Wed, 28 Jun 2000 12:03:18 -0500", "msg_from": "\"Ross J. Reedstrom\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Big 7.1 open items" }, { "msg_contents": "> -----Original Message-----\n> From: Zeugswetter Andreas SB\n> \n> > AFAIK,schema is independent from user in SQL92.\n> > So default_tablespace_per_user doesn't necessarily imply\n> > default_tablespace_per_schema.\n> \n> Well, sombody must be interpreting this wrong, because \n> in Informix and Oracle the schema corresponds to the owner\n> and they say they conform to ansi in this regard.\n\nIs there really a schema:user=1:1 limitation in SQL-92 ?\nThough both SQL-86 and SQL-89 had the limitation\nSQL-92 removed it AFAIK.\n\nRegards.\n\nHiroshi Inoue\n", "msg_date": "Thu, 29 Jun 2000 02:05:31 +0900", "msg_from": "\"Hiroshi Inoue\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: AW: Big 7.1 open items" }, { "msg_contents": "On Thu, Jun 29, 2000 at 02:05:31AM +0900, Hiroshi Inoue wrote:\n> \n> Is there really a schema:user=1:1 limitation in SQL-92 ?\n> Though both SQL-86 and SQL-89 had the limitation\n> SQL-92 removed it AFAIK.\n\nSee my other post. In SQL92, the username is the default schema name.\n\nRoss\n-- \nRoss J. Reedstrom, Ph.D., <[email protected]> \nNSBRI Research Scientist/Programmer\nComputer and Information Technology Institute\nRice University, 6100 S. Main St., Houston, TX 77005\n", "msg_date": "Wed, 28 Jun 2000 12:19:54 -0500", "msg_from": "\"Ross J. Reedstrom\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Big 7.1 open items" } ]
[ { "msg_contents": "\n ....current CVS:\n\n./configure:\n\nsed: can't read ./src/bin/pg_dump/Makefile.in: No such file or directory\nsed: can't read ./src/bin/pg_version/Makefile.in: No such file or directory\nsed: can't read ./src/bin/psql/Makefile.in: No such file or directory\nsed: can't read ./src/interfaces/odbc/Makefile.global.in: No such file or\nsed: can't read ./src/pl/Makefile.in: No such file or directory\nsed: can't read ./src/pl/plpgsql/src/Makefile.in: No such file or directory\nsed: can't read ./src/pl/plperl/GNUmakefile.in: No such file or directory\n\n and some Makefiles are zeroized...\n\n\n\t\t\t\t\tKarel\n\n", "msg_date": "Wed, 28 Jun 2000 17:36:43 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": true, "msg_subject": "./configure bug in CVS" }, { "msg_contents": "Karel Zak <[email protected]> writes:\n> ....current CVS:\n\n> ./configure:\n\n> sed: can't read ./src/bin/pg_dump/Makefile.in: No such file or directory\n\nSounds like you have an out-of-date copy of configure. Might want to\ndo a full CVS checkout to ensure your tree is in sync again...\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 28 Jun 2000 12:22:35 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: ./configure bug in CVS " }, { "msg_contents": "These files are removed in CVS but configure was also updated. (I checked\non the CVS web interface -- it is.) So your cvs pull got messed up. Rerun\nautoconf to fix the problem locally.\n\nKarel Zak writes:\n\n> \n> ....current CVS:\n> \n> ./configure:\n> \n> sed: can't read ./src/bin/pg_dump/Makefile.in: No such file or directory\n> sed: can't read ./src/bin/pg_version/Makefile.in: No such file or directory\n> sed: can't read ./src/bin/psql/Makefile.in: No such file or directory\n> sed: can't read ./src/interfaces/odbc/Makefile.global.in: No such file or\n> sed: can't read ./src/pl/Makefile.in: No such file or directory\n> sed: can't read ./src/pl/plpgsql/src/Makefile.in: No such file or directory\n> sed: can't read ./src/pl/plperl/GNUmakefile.in: No such file or directory\n> \n> and some Makefiles are zeroized...\n> \n> \n> \t\t\t\t\tKarel\n> \n> \n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Wed, 28 Jun 2000 18:37:09 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: ./configure bug in CVS" }, { "msg_contents": "\nOn Wed, 28 Jun 2000, Tom Lane wrote:\n\n> Karel Zak <[email protected]> writes:\n> > ....current CVS:\n> \n> > ./configure:\n> \n> > sed: can't read ./src/bin/pg_dump/Makefile.in: No such file or directory\n> \n> Sounds like you have an out-of-date copy of configure. Might want to\n> do a full CVS checkout to ensure your tree is in sync again...\n> \n\n Oh yes, my local CVS pool was mazy. Now is all right.\n\n Thanks.\n\t\t\t\t\tKarel\n\n", "msg_date": "Wed, 28 Jun 2000 18:46:30 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": true, "msg_subject": "Re: ./configure bug in CVS " } ]
[ { "msg_contents": "Version: 7.0.2\n\nThere seems to be a problem with creating a unique constraint, like this:\n\nbray=# CREATE TABLE supplier ( terms CHAR(2) ,ac_code CHAR(5) \nUNIQUE) INHERITS (organisation);\nNOTICE: CREATE TABLE/UNIQUE will create implicit index 'supplier_ac_code_key' \nfor table 'supplier'\nERROR: RelationClearRelation: relation 2951924 modified while in use\n\nbut without UNIQUE:\n\nbray=# create table supplier (terms char(2), acs_code char(5)) inherits \n(organisation);\nCREATE\n\nbray=# \\d organisation\n Table \"organisation\"\n Attribute | Type | Modifier \n------------+----------+--------------------\n ptype | smallint | \n id | char(10) | not null\n name | text | not null\n address | integer | \n salutation | text | default 'Dear Sir'\n envelope | text | \n email | text | \n www | text | \n contact | char(10) | \n structure | char(1) | \n department | text | \n parent_id | char(10) | \nConstraints: (((department ISNULL) AND (parent_id ISNULL)) OR ((department > \n''::text) AND (parent_id > ''::bpchar)))\n ((((structure = 'L'::bpchar) OR (structure = 'C'::bpchar)) OR \n(structure = 'U'::bpchar)) OR (structure = 'O'::bpchar))\nRules: organisation_type_ins_rule,\n organisation_type_upd_rule\nTriggers: RI_ConstraintTrigger_2950794,\n RI_ConstraintTrigger_2950240\n\nbray=# \n\n-- \nOliver Elphick [email protected]\nIsle of Wight http://www.lfix.co.uk/oliver\nPGP: 1024R/32B8FAA1: 97 EA 1D 47 72 3F 28 47 6B 7E 39 CC 56 E4 C1 47\nGPG: 1024D/3E1D0C1C: CA12 09E0 E8D5 8870 5839 932A 614D 4C34 3E1D 0C1C\n ========================================\n \"The Spirit of the Lord is upon me, because he hath \n anointed me to preach the gospel to the poor; he hath \n sent me to heal the brokenhearted, to preach \n deliverance to the captives, and recovering of sight \n to the blind, to set at liberty them that are \n bruised...\" Luke 4:18 \n\n\n", "msg_date": "Wed, 28 Jun 2000 16:55:10 +0100", "msg_from": "\"Oliver Elphick\" <[email protected]>", "msg_from_op": true, "msg_subject": "Problem creating UNIQUE constraint " } ]
[ { "msg_contents": "Daniel Freedman wrote:\n> \n> Hi,\n> \n> Anyway, Yahoo reported this morning\n> (http://biz.yahoo.com/bw/000628/ca_va_linu.html) that MySQL formed a\n> strategic alliance with VA Linux and released their MySQL product under\n> the GPL license.\n\nby going GPL and adding transactions they deserve at least a look :)\n\nI wonder how they resolve(d) the problem with SleepyCat/BSD-DB (which \nthey use for transaction support) _not_ being GPL .\n\n---------------\nHannu\n", "msg_date": "Wed, 28 Jun 2000 19:09:05 +0300", "msg_from": "Hannu Krosing <[email protected]>", "msg_from_op": true, "msg_subject": "Re: MySQL has gone GPL" }, { "msg_contents": "\nHi,\n\nThis is quite off-topic, but thought people might be interested in it, in\ncase they hadn't heard yet (and I thought, considered it is off-topic in\ngeneral, I'd be better off putting it to hackers than general). Apologies\nin advance if I'm wrong to do so.\n\nAnyway, Yahoo reported this morning\n(http://biz.yahoo.com/bw/000628/ca_va_linu.html) that MySQL formed a\nstrategic alliance with VA Linux and released their MySQL product under\nthe GPL license. Now, I've never used MySQL, and, in the learning stages\nof transferring some of my Oracle knowledge to Postgres, I've really\nstarted to love PGSql, but I thought the developers might want to be aware\nof this development.\n\nTake care,\n\nDaniel\n\n", "msg_date": "Wed, 28 Jun 2000 12:39:45 -0400 (EDT)", "msg_from": "Daniel Freedman <[email protected]>", "msg_from_op": false, "msg_subject": "MySQL has gone GPL" }, { "msg_contents": "\n> \n> Hi,\n> \n> This is quite off-topic, but thought people might be interested in it, in\n> case they hadn't heard yet (and I thought, considered it is off-topic in\n> general, I'd be better off putting it to hackers than general). Apologies\n> in advance if I'm wrong to do so.\n> \n> Anyway, Yahoo reported this morning\n> (http://biz.yahoo.com/bw/000628/ca_va_linu.html) that MySQL formed a\n> strategic alliance with VA Linux and released their MySQL product under\n> the GPL license. Now, I've never used MySQL, and, in the learning stages\n> of transferring some of my Oracle knowledge to Postgres, I've really\n> started to love PGSql, but I thought the developers might want to be aware\n> of this development.\n\n On http://www.mysql.com is something about it too.\n\nInteresting is that mySQL will in \"strategic alliance\" is Progress too..\n\nBTW. --- MySQL with or without GPL, still it is same MySQL :-)\n\n\t\t\t\t\tKarel\n\n", "msg_date": "Wed, 28 Jun 2000 18:49:47 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: MySQL has gone GPL" } ]
[ { "msg_contents": "Bruce Momjian <[email protected]> writes:\n> If we put multiple database tables in the same directory, have we\n> considered how to drop databases? Right now we do rm -rf:\n\nrm -rf will no longer work in a tablespaces environment anyway.\n(Even if you kept symlinks underneath the DB directory, rm -rf\nwouldn't follow them.)\n\nDROP DATABASE will have to be implemented honestly: run through\npg_class and do a regular DROP on each user table.\n\nOnce you've got rid of the user tables, rm -rf should suffice to\nget rid of the \"home tablespace\" as I've been calling it, with\nall the system tables therein.\n\nNow that you mention it, this is another reason why system tables for\neach database have to live in a separate tablespace directory: there's\nno other good way to do that final stage of DROP DATABASE. The\nDROP-each-table approach doesn't work for system tables (somewhere along\nabout the point where you drop pg_attribute, DROP TABLE itself would\nstop working ;-)).\n\nHowever I do see a bit of a problem here: since DROP DATABASE is\nordinarily executed by a backend that's running in a different database,\nhow's it going to read pg_class of the target database? Perhaps it will\nbe necessary to fire up a sub-backend that runs in the target DB for\nlong enough to kill all the user tables. Looking messy...\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 28 Jun 2000 12:11:40 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Big 7.1 open items " }, { "msg_contents": "> However I do see a bit of a problem here: since DROP DATABASE is\n> ordinarily executed by a backend that's running in a different database,\n> how's it going to read pg_class of the target database? Perhaps it will\n> be necessary to fire up a sub-backend that runs in the target DB for\n> long enough to kill all the user tables. Looking messy...\n\nThat was my feeling. Imagine another issue. If you see a file, how do\nyou figure out what database it belong to? You would have to cycle\nthrough the pg_class relations for every database. Seems such reverse\nlookups would not be impossible. Not sure if it will ever be required.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 28 Jun 2000 12:40:53 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" }, { "msg_contents": "> -----Original Message-----\n> From: Tom Lane [mailto:[email protected]]\n> \n> However I do see a bit of a problem here: since DROP DATABASE is\n> ordinarily executed by a backend that's running in a different database,\n> how's it going to read pg_class of the target database? Perhaps it will\n> be necessary to fire up a sub-backend that runs in the target DB for\n> long enough to kill all the user tables. Looking messy...\n>\n\nWhy do we have to have system tables per *database* ?\nIs there anything wrong with global system tables ?\nAnd how about adding dbid to pg_class,pg_proc etc ?\n\nRegards.\n\nHiroshi Inoue\n", "msg_date": "Thu, 29 Jun 2000 02:28:45 +0900", "msg_from": "\"Hiroshi Inoue\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: Big 7.1 open items " }, { "msg_contents": "\"Hiroshi Inoue\" <[email protected]> writes:\n> Why do we have to have system tables per *database* ?\n> Is there anything wrong with global system tables ?\n> And how about adding dbid to pg_class,pg_proc etc ?\n\nWe could, but I think I'd vote against it on two grounds:\n\n1. Reliability. If something corrupts pg_class, do you want to\nlose your whole installation, or just one database?\n\n2. Increased locking overhead/loss of concurrency. Currently, there\nis very little lock contention between backends running in different\ndatabases. A shared pg_class will be a single point of locking (as\nwell as a single point of failure) for the whole installation.\n\nIt would solve the DROP DATABASE problem kind of nicely, but really\nit'd just be downgrading DROP DATABASE to a DROP SCHEMA operation...\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 28 Jun 2000 13:39:32 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Big 7.1 open items " }, { "msg_contents": "Tom Lane wrote:\n\n> \"Hiroshi Inoue\" <[email protected]> writes:\n> > Why do we have to have system tables per *database* ?\n> > Is there anything wrong with global system tables ?\n> > And how about adding dbid to pg_class,pg_proc etc ?\n>\n> We could, but I think I'd vote against it on two grounds:\n>\n> 1. Reliability. If something corrupts pg_class, do you want to\n> lose your whole installation, or just one database?\n>\n> 2. Increased locking overhead/loss of concurrency. Currently, there\n> is very little lock contention between backends running in different\n> databases. A shared pg_class will be a single point of locking (as\n> well as a single point of failure) for the whole installation.\n\nIsn't current design of PG's *database* for dropdb using \"rm -rf\"\nrather than for above 1.2. ?\nIf we couldn't rely on our db itself and our locking mechanism is\npoor,we could start different postmasters for different *database*s.\n\n\n> It would solve the DROP DATABASE problem kind of nicely, but really\n> it'd just be downgrading DROP DATABASE to a DROP SCHEMA operation...\n>\n\nWhat is our *DATABASE* ?\nIs it clear to all people ?\nAt least it's a vague concept for me.\nCould you please tell me what kind of objects are our *DATABASE*\nobjects but could not be schema objects ?\n\nRegards.\n\nHiroshi Inoue\n\n\n", "msg_date": "Thu, 29 Jun 2000 08:53:03 +0900", "msg_from": "Hiroshi Inoue <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" }, { "msg_contents": "Tom Lane wrote:\n\n> Now that you mention it, this is another reason why system tables for\n> each database have to live in a separate tablespace directory: there's\n> no other good way to do that final stage of DROP DATABASE. The\n> DROP-each-table approach doesn't work for system tables (somewhere along\n> about the point where you drop pg_attribute, DROP TABLE itself would\n> stop working ;-)).\n\nIf drop table is extended to drop multiple tables at once, then you read\nand cache everything you need before doing all the destruction in the\nsecond stage.\n", "msg_date": "Sat, 01 Jul 2000 13:59:26 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Big 7.1 open items" } ]
[ { "msg_contents": "I'm about halfway done with revising backend memory management per\nthe discussions on pghackers around the end of April (see the\njust-committed src/backend/utils/mmgr/README for latest version\nof the proposal). There are several things that weren't previously\ndiscussed that are now looking like good ideas. I don't think\nthese'll be too controversial, but I wanted to mention them in case\nanyone gets unhappy:\n\n1. src/backend/nodes/freefuncs.c is currently dead code; it's not\ncalled from anywhere. I had proposed reviving it to use for freeing\nrule trees that are stored in relcache entries (currently, a relcache\nflush leaks storage for any rules associated with flushed entries)\nbut Hiroshi and others objected that freeObject() is inherently unsafe\nsince it can't cope with nodetrees having multiple links to the same\nnode --- and we do create such trees in places. We now have a much\nbetter, safer, and faster mechanism for getting rid of arbitrary node\ntrees: construct them in a private memory context that you can delete\n(or just reset) when you don't want the tree anymore. Therefore I'm\nnow persuaded that freefuncs.c never will be anything but dead code.\nI propose removing it entirely, so that people won't feel they have\nto maintain it when adding/revising/deleting node types. Any\nobjections?\n\n2. Currently there is some code in aset.c that arranges to wipe\nfreed space immediately upon its being freed, as an aid in detecting\nattempted uses of already-freed objects. It's conditionally compiled\nbased on \"#ifdef CLOBBER_FREED_MEMORY\", which ought to be mentioned\nin config.h.in (but isn't yet). This slows things down a little bit\nso I wouldn't recommend having it turned on for production use, but\nI think it would be a good idea to keep it turned on during the\ndevelopment and early beta phases of 7.1. With the memory\nmanagement changes I think we will be at increased risk of having\nuse-of-freed-memory bugs, so I'd like to get as much testing as\npossible done with this code enabled. Will anyone object if I make\nCLOBBER_FREED_MEMORY defined by default until 7.1 release is near?\n\n3. The portal (\"DECLARE CURSOR\") code is now critically dependent on\nbeing able to copy parse and plan trees with copyObject(). Formerly it\ndidn't copy them, instead using a hack that involved relabeling the\nworking \"blank\" portal as a permanent portal so that its contents\nwouldn't go away at end of statement. That won't work anymore (blank\nportals are history) so the constructed trees have to be explicitly\ncopied into the memory context that's created for the cursor portal.\nThe reason I bring this up is that we've had bugs in the past with some\nnode types or tree structures not being understood or properly copied by\ncopyObject(). Such a bug would now manifest itself as a failure\noccurring only when some particular feature is used in a DECLARE CURSOR.\nTo help catch such problems sooner, it'd be a good idea to have\nconditionally-compiled test code that forces *every* generated parse and\nplan tree to be passed through copyObject(). Again, I propose setting\nup config.h symbols for this and turning them on by default during the\n7.1 development phase.\n\nComments, objections?\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 28 Jun 2000 13:04:27 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "Misc. consequences of backend memory management changes" }, { "msg_contents": "> I'm about halfway done with revising backend memory management per\n> the discussions on pghackers around the end of April (see the\n> just-committed src/backend/utils/mmgr/README for latest version\n> of the proposal). There are several things that weren't previously\n> discussed that are now looking like good ideas. I don't think\n> these'll be too controversial, but I wanted to mention them in case\n> anyone gets unhappy:\n> \n> 1. src/backend/nodes/freefuncs.c is currently dead code; it's not\n> called from anywhere. I had proposed reviving it to use for freeing\n> rule trees that are stored in relcache entries (currently, a relcache\n> flush leaks storage for any rules associated with flushed entries)\n> but Hiroshi and others objected that freeObject() is inherently unsafe\n> since it can't cope with nodetrees having multiple links to the same\n> node --- and we do create such trees in places. We now have a much\n> better, safer, and faster mechanism for getting rid of arbitrary node\n> trees: construct them in a private memory context that you can delete\n> (or just reset) when you don't want the tree anymore. Therefore I'm\n> now persuaded that freefuncs.c never will be anything but dead code.\n> I propose removing it entirely, so that people won't feel they have\n> to maintain it when adding/revising/deleting node types. Any\n> objections?\n\nSure.\n\n\n> \n> 2. Currently there is some code in aset.c that arranges to wipe\n> freed space immediately upon its being freed, as an aid in detecting\n> attempted uses of already-freed objects. It's conditionally compiled\n> based on \"#ifdef CLOBBER_FREED_MEMORY\", which ought to be mentioned\n> in config.h.in (but isn't yet). This slows things down a little bit\n> so I wouldn't recommend having it turned on for production use, but\n> I think it would be a good idea to keep it turned on during the\n> development and early beta phases of 7.1. With the memory\n> management changes I think we will be at increased risk of having\n> use-of-freed-memory bugs, so I'd like to get as much testing as\n> possible done with this code enabled. Will anyone object if I make\n> CLOBBER_FREED_MEMORY defined by default until 7.1 release is near?\n\nWhy not have it enabled when cassert is enabled?\n\n> \n> 3. The portal (\"DECLARE CURSOR\") code is now critically dependent on\n> being able to copy parse and plan trees with copyObject(). Formerly it\n> didn't copy them, instead using a hack that involved relabeling the\n> working \"blank\" portal as a permanent portal so that its contents\n> wouldn't go away at end of statement. That won't work anymore (blank\n> portals are history) so the constructed trees have to be explicitly\n> copied into the memory context that's created for the cursor portal.\n> The reason I bring this up is that we've had bugs in the past with some\n> node types or tree structures not being understood or properly copied by\n> copyObject(). Such a bug would now manifest itself as a failure\n> occurring only when some particular feature is used in a DECLARE CURSOR.\n> To help catch such problems sooner, it'd be a good idea to have\n> conditionally-compiled test code that forces *every* generated parse and\n> plan tree to be passed through copyObject(). Again, I propose setting\n> up config.h symbols for this and turning them on by default during the\n> 7.1 development phase.\n\nSame, why not use cassert, or maybe change cassert to debug and all\ndevelopers will enable it?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 28 Jun 2000 13:27:35 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Misc. consequences of backend memory management changes" }, { "msg_contents": "\nOn Wed, 28 Jun 2000, Tom Lane wrote:\n\n> I'm about halfway done with revising backend memory management per\n\n haha, I just prepare first query cache snapshot and what I see, \nTom already rewrite aset/mcxt and my routines are out-of-date. \nBut it is right, with your changes it will query cache better :-) \n\n Well, I a little surprise that you remove all aset definiton from\nheader files to aset.c. If anyone will write new context type, he\ncan't uses some already exist definition.\n\n IMHO not will bad if all context types (now exists aset type only)\nwill more transparently and will own header files and at first sight will\nvisible what is common memory routines and what specific.\n\n I believe that current memory managemet changes create more _modular_\nmem code. \n\n\nAbout context tree --- what will happen if in PG will exist some context \nthat not will in context tree? Yes, it is curios question...explication: \n I have in the query cache contexts (for each plan) that are persisten (in \nshmem) across connection (and across TopMemoryContext live-time). How \nintegrate this context to the contect tree? Or skip for this specific \nvariant context type independent MemoryContextCreate and init this common \npart itself? - (I vote for this)\n\n New pfree(), repalloc() are nice. \n\n Plan you some changes in SPI? I have new code for SPI save plan\n(per-context and via query cache). \n\n And last question, is current mcxt.c API final? I want port my query cache\nto compatible with current interface. \n\n\t\t\t\t\tKarel\n\n", "msg_date": "Wed, 28 Jun 2000 20:26:59 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Misc. consequences of backend memory management changes" }, { "msg_contents": "\nOn Wed, 28 Jun 2000, Karel Zak wrote:\n\n> \n> About context tree --- what will happen if in PG will exist some context \n> that not will in context tree? Yes, it is curios question...explication: \n> I have in the query cache contexts (for each plan) that are persisten (in \n> shmem) across connection (and across TopMemoryContext live-time). How \n> integrate this context to the contect tree? Or skip for this specific \n> variant context type independent MemoryContextCreate and init this common \n> part itself? - (I vote for this)\n\n I a little speculate about it and I mean that query cache contexts can be\nin separate contest tree based on master query cache context that is create\nduring global shmem initialization. Agree?\n\n\t\t\t\t\t\tKarel\n\n", "msg_date": "Wed, 28 Jun 2000 21:40:35 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Misc. consequences of backend memory management changes" }, { "msg_contents": "Karel Zak <[email protected]> writes:\n> Well, I a little surprise that you remove all aset definiton from\n> header files to aset.c. If anyone will write new context type, he\n> can't uses some already exist definition.\n\n> IMHO not will bad if all context types (now exists aset type only)\n> will more transparently and will own header files and at first sight will\n> visible what is common memory routines and what specific.\n\nI left AllocSetContext in memnodes.h temporarily, but I think in the\nlong run it should only be defined inside aset.c. Anyone else have\nan opinion on that?\n\n> About context tree --- what will happen if in PG will exist some context \n> that not will in context tree?\n\nYou can certainly have top-level contexts that aren't children of\nanything. There's no real functional distinction between a context\ncreated like that and one that is a child of TopMemoryContext, because\nwe never reset/delete children of TopMemoryContext anyway. I was\ncareful to make all the other contexts descendants of TopMemoryContext\nbecause I wanted to be able to call MemoryContextStats(TopMemoryContext)\nto get a picture of all the memory usage. Other than that\nconsideration, CacheMemoryContext, ErrorContext, and so forth could\nperfectly well have been top-level contexts with no parent.\n\nA context representing shared memory probably should be a top-level\ncontext, though. That would make sense if you think of TopMemoryContext\nas the top-level of all the backend's *local* memory.\n\n> Or skip for this specific variant context type independent\n> MemoryContextCreate and init this common part itself? - (I vote for\n> this)\n\nNo, certainly not. Just pass a NULL for parent if you don't want to\nconnect it up to the local context tree.\n\n> Plan you some changes in SPI? I have new code for SPI save plan\n> (per-context and via query cache). \n\nI have been wondering about that. I don't like SPI saving plans in\nTopMemoryContext because I don't think it knows how to get rid of them;\na plan saved there is effectively a permanent memory leak. Not sure\nwhat to do about it though.\n\n> And last question, is current mcxt.c API final? I want port my query cache\n> to compatible with current interface. \n\nI think it is done, though I reserve the right to change it if I find out\nsomething needs to be done differently ;-). But what I am expecting to\ndo next is just modify the planner and executor to make better use of\nthe facilities that are there now. I won't need to change mcxt.c any\nmore unless I find it's too hard to use or doesn't do quite the right\nthings...\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 28 Jun 2000 21:37:35 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Misc. consequences of backend memory management changes " }, { "msg_contents": "On Wed, 28 Jun 2000, Tom Lane wrote:\n\n> Karel Zak <[email protected]> writes:\n> > Well, I a little surprise that you remove all aset definiton from\n> > header files to aset.c. If anyone will write new context type, he\n> > can't uses some already exist definition.\n> \n> > IMHO not will bad if all context types (now exists aset type only)\n> > will more transparently and will own header files and at first sight will\n> > visible what is common memory routines and what specific.\n> \n> I left AllocSetContext in memnodes.h temporarily, but I think in the\n> long run it should only be defined inside aset.c. Anyone else have\n> an opinion on that?\n\n For example I use AllocSet's block definition.\n\n> \n> > About context tree --- what will happen if in PG will exist some context \n> > that not will in context tree?\n> \n> You can certainly have top-level contexts that aren't children of\n> anything. There's no real functional distinction between a context\n> created like that and one that is a child of TopMemoryContext, because\n> we never reset/delete children of TopMemoryContext anyway. I was\n> careful to make all the other contexts descendants of TopMemoryContext\n> because I wanted to be able to call MemoryContextStats(TopMemoryContext)\n> to get a picture of all the memory usage. Other than that\n\n Yes, I see, it is good idea.\n\n> consideration, CacheMemoryContext, ErrorContext, and so forth could\n> perfectly well have been top-level contexts with no parent.\n> \n> A context representing shared memory probably should be a top-level\n> context, though. That would make sense if you think of TopMemoryContext\n> as the top-level of all the backend's *local* memory.\n> \n> > Or skip for this specific variant context type independent\n> > MemoryContextCreate and init this common part itself? - (I vote for\n> > this)\n> \n> No, certainly not. Just pass a NULL for parent if you don't want to\n> connect it up to the local context tree.\n\n Well, the query cache shared memory pool and qCache_TopContext (parent of\nall cache entry) are init during postmarter start up. In this time not exist\nTopMemoryContext and 'node' for this context will alloc via malloc(). It is \nright and I understand here. This qCache_TopContext is used for internal \nquery cache data and all data is in shmem.\n \n*But* if I add entry to query cache I create for each entry separate memory \ncontext and it must be in shared memory, but in shared memory must be not\nonly data allocated via this context, in shmem must be _all_ for context\nrelevant data --- so, context independent data too (bacause context must\noverlive backend end). A problem is that common MemoryContextCreate() expects \nthat context 'node' is allocated in the TopMemoryContext or by force of \nmalloc(). Not is possible allocate context node in a parent context.\n\nI read your comment \"since the node must survive resets of its parent\ncontext\", but how resolve this?\n \n \n I have two method for save cached planns. First in the hash table in shared \nmemory (about it we talking) and second in hash table in local memory. For\n'local part' context I can set as parent CacheMemoryContext or will better\nuse TopMemoryConetxt.\n\n> > Plan you some changes in SPI? I have new code for SPI save plan\n> > (per-context and via query cache). \n> \n> I have been wondering about that. I don't like SPI saving plans in\n> TopMemoryContext because I don't think it knows how to get rid of them;\n> a plan saved there is effectively a permanent memory leak. Not sure\n> what to do about it though.\n\n We already told about it with Hiroshi. I rewrite SPI_sevaplan(), now it\nsave each plan to separate context and is possible use SPI_freeplan().\nAlso I add 'by_key' interface to SPI that allows save planns to the query \ncache under some key (string or binary key). My SPI *not* use\nTopMemoryContext.\n\n\n\t\t\t\t\t\tKarel\n\n", "msg_date": "Thu, 29 Jun 2000 11:39:07 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Misc. consequences of backend memory management changes " }, { "msg_contents": "Karel Zak <[email protected]> writes:\n>>>> Or skip for this specific variant context type independent\n>>>> MemoryContextCreate and init this common part itself? - (I vote for\n>>>> this)\n>> \n>> No, certainly not. Just pass a NULL for parent if you don't want to\n>> connect it up to the local context tree.\n\n> Well, the query cache shared memory pool and qCache_TopContext (parent of\n> all cache entry) are init during postmarter start up. In this time not exist\n> TopMemoryContext and 'node' for this context will alloc via malloc().\n\nLook again ;-). TopMemoryContext is created as the first or second step\nin postmaster startup, long before we attach to shared mem.\n\n> *But* if I add entry to query cache I create for each entry separate memory \n> context and it must be in shared memory, but in shared memory must be not\n> only data allocated via this context, in shmem must be _all_ for context\n> relevant data --- so, context independent data too (bacause context must\n> overlive backend end). A problem is that common MemoryContextCreate() expects \n> that context 'node' is allocated in the TopMemoryContext or by force of \n> malloc(). Not is possible allocate context node in a parent context.\n\nGood point.\n\n> I read your comment \"since the node must survive resets of its parent\n> context\", but how resolve this?\n\nWell, there's a fundamental problem with keeping memory contexts that\nmeet this API in shared memory anyway: the contexts have to contain\nparent/child pointers pointing at other contexts, and the contexts\nhave to contain pointers to context-management function pointer structs.\nNeither of these are very safe for shared memory.\n\nYou've probably noticed that we keep cross-pointers in shared memory in\nthe form of offsets from the start of the shmem block, not as absolute\npointers. This is to support the possibility that the shmem block is\nmapped at different addresses in different backends. (It's possible\nthat we could give up that flexibility now that all the backends are\nforked from a postmaster that's already attached the shmem block.\nBut I'd rather not wire such an assumption into the contents of shmem.)\n\nSimilarly, it would be risky to assume that the functions all appear\nat the same address in every backend connected to the shmem block.\n\nWhat I think you probably want to do is create \"placeholder\" context\nnodes in backend local memory to represent the shared-memory contexts.\nThe placeholder would contain the MemoryContext fields and a shmem\noffset to a struct in shared memory that is the \"real\" control info\nfor the context.\n\nIf every backend has a placeholder node for every context in shmem\nthen you have a bit of a problem deleting shmem contexts, but I\nthink you probably don't need to do that. There should be a permanent\nplaceholder for the top-level shmem context (create this during\nbackend startup), but placeholder context nodes for per-plan contexts\nonly need to live for as long as you are copying nodetrees into the\ncontext. They don't have to stick around longer than that.\n\n\nA few minutes later:\n\nNow that I think about it, this whole notion of copying nodetrees\ninto shared memory has got a pointer problem, because the constructed\ntree is going to be full of pointers that will only be valid for a\nbackend that has shmem attached at the same address that the creating\nbackend had.\n\nDo we really want to give up flexibility of shmem attach address in\norder to have a shared-memory plan cache? I've been dubious about\nthe notion of a shared plan cache from day one (I think the locking\noverhead is going to be intolerable) and this may be the problem\nthat kills it completely.\n\nBefore you say \"we will never again care about flexibility of attach\naddress\", consider these two scenarios:\n1. Examining shmem using a debugging process started independently\n of the postmaster.\n2. \"Hot swapping\" a new compilation of the postmaster/backend without\n service interruption. This is possible (for changes that don't\n change the layout of shmem, of course) by killing and restarting\n the postmaster without killing extant backends. But you couldn't\n assume that the new incarnation of the postmaster would attach to\n the existing shmem blocks at the exact same addresses the previous\n incarnation had been assigned.\nWe don't have either of these today but we might want to do them\nat some point in the future. There may be other reasons that I\ndidn't come up with in a moment's thought.\n\n\n>>>> Plan you some changes in SPI? I have new code for SPI save plan\n>>>> (per-context and via query cache). \n>> \n>> I have been wondering about that. I don't like SPI saving plans in\n>> TopMemoryContext because I don't think it knows how to get rid of them;\n>> a plan saved there is effectively a permanent memory leak. Not sure\n>> what to do about it though.\n\n> We already told about it with Hiroshi. I rewrite SPI_sevaplan(), now it\n> save each plan to separate context and is possible use SPI_freeplan().\n> Also I add 'by_key' interface to SPI that allows save planns to the query \n> cache under some key (string or binary key). My SPI *not* use\n> TopMemoryContext.\n\nThat all sounds good. Plan cache in local memory seems reasonable to\nme, I'm just worried about the shared-mem aspect of it.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Jun 2000 10:35:01 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Misc. consequences of backend memory management changes " }, { "msg_contents": "\nOn Thu, 29 Jun 2000, Tom Lane wrote:\n\n> Karel Zak <[email protected]> writes:\n> >>>> Or skip for this specific variant context type independent\n> >>>> MemoryContextCreate and init this common part itself? - (I vote for\n> >>>> this)\n> >> \n> >> No, certainly not. Just pass a NULL for parent if you don't want to\n> >> connect it up to the local context tree.\n> \n> > Well, the query cache shared memory pool and qCache_TopContext (parent of\n> > all cache entry) are init during postmarter start up. In this time not exist\n> > TopMemoryContext and 'node' for this context will alloc via malloc().\n> \n> Look again ;-). TopMemoryContext is created as the first or second step\n> in postmaster startup, long before we attach to shared mem.\n\n I looking :-), I don't say that here is a problem. I create master \nqCache_TopContext like standard memory context, but only alloc method will\nnon-AllocSet. No problem. \n\n> \n> > *But* if I add entry to query cache I create for each entry separate memory \n> > context and it must be in shared memory, but in shared memory must be not\n> > only data allocated via this context, in shmem must be _all_ for context\n> > relevant data --- so, context independent data too (bacause context must\n> > overlive backend end). A problem is that common MemoryContextCreate() expects \n> > that context 'node' is allocated in the TopMemoryContext or by force of \n> > malloc(). Not is possible allocate context node in a parent context.\n> \n> Good point.\n> \n> > I read your comment \"since the node must survive resets of its parent\n> > context\", but how resolve this?\n> \n> Well, there's a fundamental problem with keeping memory contexts that\n> meet this API in shared memory anyway: the contexts have to contain\n> parent/child pointers pointing at other contexts, and the contexts\n> have to contain pointers to context-management function pointer structs.\n> Neither of these are very safe for shared memory.\n\n First I must say that _all_ with shmem/local query cache _works_ now \nfor me in current (7.0) memory design (I can send it). And I hope it must\nworks in 7.1 too.\n\n For better understanding I a little describe this cache (qcache).\nIn postmaster startup and shmem init qcache create shmem pool. This pool \nis split to blocks (now 1024b) these blocks are used in qcache contexts.\nAll free blocks are in freeblock list. During initialization is created\nhash table (in shmem pool) for cached entry and qCache_TopContext (and \nalloc methods) too. All it is in postmaster. \n If user (PREPARE command) or SPI (by_key interface) add a plan to qcache\nit create new memory context in shmem pool and as parent is used master\nqCache_TopContext. New plan is copy to shmem via copyObject().\n If user (EXECUTE) or SPI needs a plan in hash table is found relevant\npointer with plan and from this is copy to executor (or ..etc) context. \n To the shared qcache pool has access only qcache's routines and all \naccess are \"dress\" in spin locks. For real work in backend is always used\nsome plan copy from qcache. --- It query cache. \n\n\n> You've probably noticed that we keep cross-pointers in shared memory in\n> the form of offsets from the start of the shmem block, not as absolute\n> pointers. This is to support the possibility that the shmem block is\n> mapped at different addresses in different backends. (It's possible\n> that we could give up that flexibility now that all the backends are\n> forked from a postmaster that's already attached the shmem block.\n> But I'd rather not wire such an assumption into the contents of shmem.)\n> \n> Similarly, it would be risky to assume that the functions all appear\n> at the same address in every backend connected to the shmem block.\n\n All work with shmem must be assure via locks.\n\n> \n> What I think you probably want to do is create \"placeholder\" context\n> nodes in backend local memory to represent the shared-memory contexts.\n> The placeholder would contain the MemoryContext fields and a shmem\n> offset to a struct in shared memory that is the \"real\" control info\n> for the context.\n\n I not sure if backend already needs information about shmem pool. Now \nI have all meta-information about shmem pool in one struct (in shmem) \nthat is at for all backend known pointer. If I need some information\nabout my pool I see this meta-data or scan hash table.\n\n I mean: we can create separate memory context tree based on some \nqCache_TopContext and init in postmaster. Standard backend memory routines\nnot will know and access to this tree. Only query cache will the bridge\nbetween shmem and standard contexts, but shmem contexts are compatible and\nis possible use it in palloc()/pfree(). \n\n But I will think about your \"placeholder\" idea :-) I still a little not\nunderstand who will need placeholder information - debug? The qcache\nknows all from own structs.\n\n> If every backend has a placeholder node for every context in shmem\n> then you have a bit of a problem deleting shmem contexts, but I\n> think you probably don't need to do that. There should be a permanent\n\nIt must be possible destroy context. If query cache is full, it itself\nremove a oldest plan that is mark as \"removeable\" (an example plans from \nSPI). \n\n> placeholder for the top-level shmem context (create this during\n> backend startup), but placeholder context nodes for per-plan contexts\n> only need to live for as long as you are copying nodetrees into the\n> context. They don't have to stick around longer than that.\n> \n> \n> A few minutes later:\n\nI always have \"a few minutes later\" after mail sending :-) \n\n> Now that I think about it, this whole notion of copying nodetrees\n> into shared memory has got a pointer problem, because the constructed\n> tree is going to be full of pointers that will only be valid for a\n> backend that has shmem attached at the same address that the creating\n> backend had.\n\nI don't understand, \n\ntest=# PREPARE aaa AS SELECT relname FROM pg_class \n WHERE relname ~~ $1 USING text;\nPREPARE\ntest=# \\c test\nYou are now connected to database test. <---------- new backend\ntest=# EXECUTE aaa USING 'pg_c%';\n relname\n------------------------\n pg_class\n pg_class_oid_index\n pg_class_relname_index\n\nit is in 7.0\n\n\n> Before you say \"we will never again care about flexibility of attach\n> address\", consider these two scenarios:\n> 1. Examining shmem using a debugging process started independently\n> of the postmaster.\n> 2. \"Hot swapping\" a new compilation of the postmaster/backend without\n> service interruption. This is possible (for changes that don't\n> change the layout of shmem, of course) by killing and restarting\n> the postmaster without killing extant backends. But you couldn't\n> assume that the new incarnation of the postmaster would attach to\n> the existing shmem blocks at the exact same addresses the previous\n> incarnation had been assigned.\n\n Ops, now I understand, you want on-the-fly change backend. \n\nI still mean \"we will never again care about....\" (I'm hard-headed :-) \n\n(!) All Data in the query cache must be dispensable. User's SPI routines must\nallows rebuild query and user (PREPARE) must knows that prepared planns are\nnot VIEW (!)\n\n BTW. - if we want really \"Hot swapping\" for qcache too, we can (also)\nsave native not parsed query strings into query cache and rebuild it after \non-the-fly backend change. \n\nI have some scenario too:\n\n- PostgreSQL is backend for web application and this application very \noften send to PG same query and PG has feature that allows save 95% \ntime of query parsing :-) \t\n\n- or user has trigger (RI?) and this trigger always run only \nSPI_execp_by_key()...\n\nWell, I prepare it with '#ifdef HAVE_QCACHE'. \n\n\t\t\t\t\tKarel\n\n", "msg_date": "Thu, 29 Jun 2000 18:14:00 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Misc. consequences of backend memory management changes " }, { "msg_contents": "Tom Lane writes:\n\n> Will anyone object if I make CLOBBER_FREED_MEMORY defined by default\n> until 7.1 release is near?\n\n> To help catch such problems sooner, it'd be a good idea to have\n> conditionally-compiled test code that forces *every* generated parse\n> and plan tree to be passed through copyObject(). Again, I propose\n> setting up config.h symbols for this and turning them on by default\n> during the 7.1 development phase.\n\nI've been thinking that we need a configure option for this sort of stuff,\nlike\n\n --enable-debug=memory,lock,foo,noipc\n\nwhich would result in\n\n#define MEMORY_DEBUG 1\n#define LOCK_DEBUG 1\n#define FOO_DEBUG 1\n#undef IPC_DEBUG\n\nComments?\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Thu, 29 Jun 2000 19:27:33 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Misc. consequences of backend memory management changes" }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n>> Will anyone object if I make CLOBBER_FREED_MEMORY defined by default\n>> until 7.1 release is near?\n\n> I've been thinking that we need a configure option for this sort of stuff,\n> like\n\n> --enable-debug=memory,lock,foo,noipc\n\n> which would result in\n\n> #define MEMORY_DEBUG 1\n> #define LOCK_DEBUG 1\n> #define FOO_DEBUG 1\n> #undef IPC_DEBUG\n\n> Comments?\n\nNot unreasonable, though it would tend to encourage use of short,\nhard-to-understand names for debug options :-(. We could live with\nthat as long as there were adequate documentation about what each\noption does in config.h.in.\n\nIf you wanna do it then I'd suggest folding cassert into the same\nmechanism.\n\nLast night I rearranged config.h.in so that the configure-driven symbols\nare more clearly separated from the non-configurable symbols, and I also\nseparated out the debug symbols from the feature/limit symbols. Should\nmake a good starting point for seeing what needs to be dealt with.\nUnfortunately the existing debug symbols were mostly undocumented, and\nI didn't take the time to dig to see what they do ... if anything ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Jun 2000 14:30:07 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Misc. consequences of backend memory management changes " }, { "msg_contents": "Tom Lane writes:\n\n> Last night I rearranged config.h.in so that the configure-driven\n> symbols are more clearly separated from the non-configurable symbols,\n\nI'd find it convenient if we could eventually create config.h.in\nautomatically from configure.in with autoheader, but in order to do that\nconfigure needs to know about everything that's going on in there. It's\nnot nice to invite your users to edit header files anyway...\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Fri, 30 Jun 2000 01:37:49 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Misc. consequences of backend memory management changes" }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n> I'd find it convenient if we could eventually create config.h.in\n> automatically from configure.in with autoheader, but in order to do that\n> configure needs to know about everything that's going on in there. It's\n> not nice to invite your users to edit header files anyway...\n\nI looked at autoheader and it strikes me as yet another glorified\n'cat'. If we used it, we'd basically have an acconfig.h that contains\nexactly what's now in the handgenerated config.h.in, and autoheader\nwould copy it all to config.h.in.\n\nIf we had a bunch of config.h.in's that could share code, this might\nbe useful, but for our present purposes I'm missing where the win is.\n\nOne thing that does occur to me is that a very large fraction of\nconfig.h is now symbols that are supposed to be set by configure,\nand as you say it's not good to give people the idea that they\nshould tweak those results by hand after configuring. What do you\nthink of pulling the remaining hand-settable symbols out into a\nseparate file, maybe called something like \"siteconfig.h\"? Then\nconfig.h becomes purely a machine-generated file. This would\nalso solve the problem of losing hand-set config choices if you\nrerun configure.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Jun 2000 21:01:21 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "config.h (was Re: Misc. consequences of backend memory management\n\tchanges)" }, { "msg_contents": " One thing that does occur to me is that a very large fraction of\n config.h is now symbols that are supposed to be set by configure,\n and as you say it's not good to give people the idea that they\n should tweak those results by hand after configuring. What do you\n think of pulling the remaining hand-settable symbols out into a\n separate file, maybe called something like \"siteconfig.h\"? Then\n config.h becomes purely a machine-generated file. This would\n also solve the problem of losing hand-set config choices if you\n rerun configure.\n\nconfig.h should contain only machine generated stuff for exactly the\nreasons you mention. Perhaps config.h.in should be something like:\n\n/* siteconfig.h overrides various variables */\n#include \"siteconfig.h\"\n\n/* enable feature FOO (can be overridden by siteconfig.h) */\n#ifndef FOO\n#undefine FOO\n#endif\n\nCheers,\nBrook\n", "msg_date": "Fri, 30 Jun 2000 08:56:26 -0600 (MDT)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": false, "msg_subject": "Re: config.h (was Re: Misc. consequences of backend memory management\n\tchanges)" }, { "msg_contents": "Tom Lane wrote:\n\n> 2. Currently there is some code in aset.c that arranges to wipe\n> freed space immediately upon its being freed, as an aid in detecting\n> attempted uses of already-freed objects. It's conditionally compiled\n> based on \"#ifdef CLOBBER_FREED_MEMORY\", which ought to be mentioned\n> in config.h.in (but isn't yet). This slows things down a little bit\n> so I wouldn't recommend having it turned on for production use, but\n> I think it would be a good idea to keep it turned on during the\n> development and early beta phases of 7.1. With the memory\n> management changes I think we will be at increased risk of having\n> use-of-freed-memory bugs, so I'd like to get as much testing as\n> possible done with this code enabled. Will anyone object if I make\n> CLOBBER_FREED_MEMORY defined by default until 7.1 release is near?\n\nCould we instead have some configure option --devel or something that\nturns on all the checks?\n", "msg_date": "Sat, 01 Jul 2000 14:04:19 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Misc. consequences of backend memory management changes" }, { "msg_contents": "Tom Lane writes:\n\n> I looked at autoheader and it strikes me as yet another glorified\n> 'cat'. If we used it, we'd basically have an acconfig.h that contains\n> exactly what's now in the handgenerated config.h.in, and autoheader\n> would copy it all to config.h.in.\n\nThat mechanism is obsolete, autoheader is much smarter now (2.13). In\norder to get the comments onto the symbols you write\n\nAC_DEFINE(HAVE_FOO, [], [Define if you have foo])\n\nI.e., you document the stuff where you define it.\n\n> What do you think of pulling the remaining hand-settable symbols out\n> into a separate file, maybe called something like \"siteconfig.h\"? \n\nThat seems like a good idea.\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Sat, 1 Jul 2000 17:04:04 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: config.h (was Re: Misc. consequences of backend memory\n\tmanagement changes)" } ]
[ { "msg_contents": "Hi guys !!!!\n\nJust a little question, I'd want to know, if the following situation is \na normal restriction(is should be strange..), or if it's a bug , or maybe\nif I'm missing something ...\n\nventasge2000=# create table t1 (c1 char(10),c2 varchar(10));\nCREATE\nventasge2000=# insert into t1 values('XXX666','XXX666');\nINSERT 182218 1\nventasge2000=# select * from t1 where c1 ~ '666$';\n c1 | c2 \n----+----\n(0 rows)\n\nventasge2000=# select * from t1 where c2 ~ '666$';\n c1 | c2 \n------------+--------\n XXX666 | XXX666\n(1 row)\n\n\nDoesn't regular expressions(in particular the $ metachar) work properly \nwith char columns????\n\n\nThanks in advance\n\nSEYA\nDario Estepario ...\n\nPS: I'm using version 7.02, but I also see this trouble in 7.0\n", "msg_date": "Wed, 28 Jun 2000 19:31:59 -0500", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "regular expressions troubles with char cols" }, { "msg_contents": "[email protected] writes:\n> ventasge2000=# create table t1 (c1 char(10),c2 varchar(10));\n> CREATE\n> ventasge2000=# insert into t1 values('XXX666','XXX666');\n> INSERT 182218 1\n> ventasge2000=# select * from t1 where c1 ~ '666$';\n> c1 | c2 \n> ----+----\n> (0 rows)\n> ventasge2000=# select * from t1 where c2 ~ '666$';\n> c1 | c2 \n> ------------+--------\n> XXX666 | XXX666\n> (1 row)\n\n> Doesn't regular expressions(in particular the $ metachar) work properly \n> with char columns????\n\nI see no bug there --- you've forgotten about the trailing spaces in\nthe char(10) column. Try c1 ~ '666 *$' if you want to match against a\nvariable amount of padding in a char(N) column. But really I'd suggest\nusing the appropriate datatype to begin with, and evidently char(N) is\nnot it for your use of this table.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 10:11:23 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: regular expressions troubles with char cols " } ]
[ { "msg_contents": "Where can I get more information about programming in SPI. \n(more than in PostgreSQL documentation).\nThanks for any help\nAdam\n\n\n\n\n\n\n\n\nWhere can I get more information about programming  in \nSPI. \n(more than in PostgreSQL documentation).\nThanks for any help\nAdam", "msg_date": "Thu, 29 Jun 2000 09:37:20 +0200", "msg_from": "\"Adam Walczykiewicz\" <[email protected]>", "msg_from_op": true, "msg_subject": "SPI - documentation" }, { "msg_contents": "I'm not sure that there is any other documentation other than what's in the\nprogrammer's manual.. I've been playing with the SPI this weekend and would\nbe willing to help where I can, if I only knew what you were trying to\naccomplish :-)\n\n-Mitch\n\n\n----- Original Message -----\nFrom: Adam Walczykiewicz <[email protected]>\nTo: <[email protected]>\nSent: Thursday, June 29, 2000 3:37 AM\nSubject: [HACKERS] SPI - documentation\n\n\n> Where can I get more information about programming in SPI.\n> (more than in PostgreSQL documentation).\n> Thanks for any help\n> Adam\n>\n>\n\n", "msg_date": "Mon, 3 Jul 2000 15:48:15 -0400", "msg_from": "\"Mitch Vincent\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: SPI - documentation" } ]
[ { "msg_contents": "\n> > \tln -s data/base/testdb/myspace/extent1 /var/myspace/extent1/testdb\n> \n> The idea was to put the main files in the directory, and create Extent2,\n> Extent3 directories for the extents.\n\nThe reasoning was, that the database subdir should be below the extentdir,\nso that creating different fs for each extent would be easier, and not\ndepend\non the database name.\n\nIt is easy to create fs for:\n\t/var/myspace\nor\n\t/var/myspace[/extent1]\n\t/var/myspace/extent2\nbut not if it has dbname in it.\n\nAndreas\n", "msg_date": "Thu, 29 Jun 2000 09:43:14 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: Big 7.1 open items" } ]
[ { "msg_contents": "\n> Zeugswetter Andreas SB <[email protected]> writes:\n> >>>> I've been assuming that we would create a separate tablespace for\n> >>>> each database, which would be the location of that database's\n> >>>> system tables. It's probably also the default \n> tablespace for user\n> >>>> tables created in that database, though it wouldn't have to be.\n> \n> > I think I would prefer the ability to place more than one \n> database into \n> > the same tablespace.\n> \n> You can put user tables from multiple databases into the same\n> tablespace, under this proposal. Just not system tables.\n\nYes, but then it is only half baked.\n\nAndreas\n", "msg_date": "Thu, 29 Jun 2000 09:44:33 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: Big 7.1 open items " } ]
[ { "msg_contents": "\n> > > AFAIK,schema is independent from user in SQL92.\n> > > So default_tablespace_per_user doesn't necessarily imply\n> > > default_tablespace_per_schema.\n> > \n> > Well, sombody must be interpreting this wrong, because \n> > in Informix and Oracle the schema corresponds to the owner\n> > and they say they conform to ansi in this regard.\n> \n> Is there really a schema:user=1:1 limitation in SQL-92 ?\n> Though both SQL-86 and SQL-89 had the limitation\n> SQL-92 removed it AFAIK.\n\nAs I said in another posting a user does not need to exist\nfor each schema. The dba can create objects under any \nschema name.\n\nAndreas\n", "msg_date": "Thu, 29 Jun 2000 10:00:06 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: Big 7.1 open items" }, { "msg_contents": "> -----Original Message-----\n> From: [email protected] [mailto:[email protected]]On\n> Behalf Of Zeugswetter Andreas SB\n> \n> > > > AFAIK,schema is independent from user in SQL92.\n> > > > So default_tablespace_per_user doesn't necessarily imply\n> > > > default_tablespace_per_schema.\n> > > \n> > > Well, sombody must be interpreting this wrong, because \n> > > in Informix and Oracle the schema corresponds to the owner\n> > > and they say they conform to ansi in this regard.\n> > \n> > Is there really a schema:user=1:1 limitation in SQL-92 ?\n> > Though both SQL-86 and SQL-89 had the limitation\n> > SQL-92 removed it AFAIK.\n> \n> As I said in another posting a user does not need to exist\n> for each schema. The dba can create objects under any \n> schema name.\n>\n\nSorry for my poor understanding.\nWhat I meant was that SQL92 allows the following.\n\n\tschema\t\towner\n\t---------------------------\n\tschema1\tuser1\n\tschema2\tuser1\n\tschema3\tuser2\n\tschema4\tuser3\n\tschema5\tuser3\n\tschema6\tuser3\n\nIs my understaning same as yours ?\n\nRegards.\n\nHiroshi Inoue\[email protected] \n", "msg_date": "Thu, 29 Jun 2000 17:21:57 +0900", "msg_from": "\"Hiroshi Inoue\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: AW: Big 7.1 open items" } ]
[ { "msg_contents": "The SQL7 way is the schema is the username, with the exception of \"dba\" -\nit's used as a \"global\" schema.\n\n--\nPeter Mount\nEnterprise Support\nMaidstone Borough Council\nAny views stated are my own, and not those of Maidstone Borough Council\n\n\n-----Original Message-----\nFrom: Hiroshi Inoue [mailto:[email protected]]\nSent: Thursday, June 29, 2000 9:22 AM\nTo: Zeugswetter Andreas SB\nCc: PostgreSQL-development\nSubject: RE: AW: [HACKERS] Big 7.1 open items\n\n\n> -----Original Message-----\n> From: [email protected] [mailto:[email protected]]On\n> Behalf Of Zeugswetter Andreas SB\n> \n> > > > AFAIK,schema is independent from user in SQL92.\n> > > > So default_tablespace_per_user doesn't necessarily imply\n> > > > default_tablespace_per_schema.\n> > > \n> > > Well, sombody must be interpreting this wrong, because \n> > > in Informix and Oracle the schema corresponds to the owner\n> > > and they say they conform to ansi in this regard.\n> > \n> > Is there really a schema:user=1:1 limitation in SQL-92 ?\n> > Though both SQL-86 and SQL-89 had the limitation\n> > SQL-92 removed it AFAIK.\n> \n> As I said in another posting a user does not need to exist\n> for each schema. The dba can create objects under any \n> schema name.\n>\n\nSorry for my poor understanding.\nWhat I meant was that SQL92 allows the following.\n\n\tschema\t\towner\n\t---------------------------\n\tschema1\tuser1\n\tschema2\tuser1\n\tschema3\tuser2\n\tschema4\tuser3\n\tschema5\tuser3\n\tschema6\tuser3\n\nIs my understaning same as yours ?\n\nRegards.\n\nHiroshi Inoue\[email protected] \n", "msg_date": "Thu, 29 Jun 2000 09:45:34 +0100", "msg_from": "Peter Mount <[email protected]>", "msg_from_op": true, "msg_subject": "RE: AW: Big 7.1 open items" } ]
[ { "msg_contents": "\n>Date: Thu, 29 Jun 2000 10:57:37 +0200\n>To: [email protected]\n>From: stanislas pinte <[email protected]>\n>Subject: URGENT PROBLEM: connection time-outs of postgresql backend/frontend\n>\n>hello,\n>\n>I posted several questions in the mailing list \"general\", and received no \n>response. I think that it is not documented where to manage the\n>connection time-out as far as TCP/IP connections are concerned.\n>\n>My problem is the following: using a JDBC client, and long-lived \n>connections, I see my connections being timed-out by the postmaster after \n>having been idle for a given time. How could I modify the time-out time?\n>\n>Could the postgresql dev team answer this?\n>\n>thanks a lot,\n>\n>Stan. Trademine corp.\n\n", "msg_date": "Thu, 29 Jun 2000 12:08:19 +0200", "msg_from": "stanislas pinte <[email protected]>", "msg_from_op": true, "msg_subject": "Fwd: URGENT PROBLEM: connection time-outs of postgresql\n\tbackend/frontend" } ]
[ { "msg_contents": "\n> > As I said in another posting a user does not need to exist\n> > for each schema. The dba can create objects under any \n> > schema name.\n> >\n> \n> Sorry for my poor understanding.\n> What I meant was that SQL92 allows the following.\n> \n> \tschema\t\towner\n> \t---------------------------\n> \tschema1\tuser1\n> \tschema2\tuser1\n> \tschema3\tuser2\n> \tschema4\tuser3\n> \tschema5\tuser3\n> \tschema6\tuser3\n> \n> Is my understaning same as yours ?\n\nYes, this is how I read the spec 99. Also:\n\tschema1\tuser1\n\tschema1\tuser2\n\nI doubt that this really buys any features that a simple grant cannot give.\nI mean, if a user creates an object with a schema name that is different \nfrom his user name we could simply grant him all rights on this object\n(if he isn't dba).\n\nAndreas\n", "msg_date": "Thu, 29 Jun 2000 12:13:21 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: Big 7.1 open items" } ]
[ { "msg_contents": "\n> > > I think I would prefer the ability to place more than one \n> > database into \n> > > the same tablespace.\n> > \n> > You can put user tables from multiple databases into the same\n> > tablespace, under this proposal. Just not system tables.\n> \n> Yes, but then it is only half baked.\n\nHalf baked or not, I think I am starting to like it.\nI think I would restrict such an automagically created tablespace\n(tblspace name = db name) to only contain tables from this database.\n\nAndreas\n", "msg_date": "Thu, 29 Jun 2000 12:33:39 +0200", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: Big 7.1 open items " } ]
[ { "msg_contents": "The web page\n\nhttp://www.paulandlesley.org/gmake/autodep.html\n\ndescribes the automatic dependency generation mechanism I plan to\nimplement (in particular the \"Advanced\" section). I intend to get rid of\nthe make depend targets afterwards.\n\nIf you're interested, take a look. It's really quite ingenious.\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Thu, 29 Jun 2000 17:31:19 +0200 (MET DST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Automatic dependency generation" } ]
[ { "msg_contents": "Can someone explain why 'volatile' is used in psql/mainloop.c? Seems\nvolatile is only for hardware-specific variables that can not be\noptimized.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 29 Jun 2000 12:17:19 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "psql use of 'volatile'" }, { "msg_contents": "On Thu, 29 Jun 2000, Bruce Momjian wrote:\n\n> Can someone explain why 'volatile' is used in psql/mainloop.c?\n\nIf you remove them then you get tons of warnings about variables possibly\ngetting clobbered. The reason is the longjmp business that's going on when\nyou press Control-C. (The fact that the variables would get clobbered is\nnot critical since they're reinitialized immediately anyway, but who am I\nto argue with the compiler.)\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Thu, 29 Jun 2000 18:36:48 +0200 (MET DST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: psql use of 'volatile'" }, { "msg_contents": "[ Charset ISO-8859-1 unsupported, converting... ]\n> On Thu, 29 Jun 2000, Bruce Momjian wrote:\n> \n> > Can someone explain why 'volatile' is used in psql/mainloop.c?\n> \n> If you remove them then you get tons of warnings about variables possibly\n> getting clobbered. The reason is the longjmp business that's going on when\n> you press Control-C. (The fact that the variables would get clobbered is\n> not critical since they're reinitialized immediately anyway, but who am I\n> to argue with the compiler.)\n\nAgreed.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 29 Jun 2000 12:37:21 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: psql use of 'volatile'" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n> Can someone explain why 'volatile' is used in psql/mainloop.c? Seems\n> volatile is only for hardware-specific variables that can not be\n> optimized.\n\nNot at all. The ANSI C spec specifically says that you must declare\nas volatile any local variables that you need to change after a\nsetjmp() call. This prevents the compiler from putting them in\nregisters, which would cause their post-longjmp values to be\nuncertain. The exact statement is that after a longjmp,\n\n [#3] All accessible objects have values as of the time\n longjmp was called, except that the values of objects of\n automatic storage duration that are local to the function\n containing the invocation of the corresponding setjmp macro\n that do not have volatile-qualified type and have been\n changed between the setjmp invocation and longjmp call are\n indeterminate.\n\nThe reason they're \"indeterminate\" is that longjmp restores all the\nmachine registers to the values saved by setjmp. So, if your local\nvariable foo was allocated in a register, then foo reverts to its\nvalue as of the time of setjmp; but if it's in memory, it doesn't\nrevert. The behavior is really perfectly \"determinate\", but you\ncan't safely rely on which will happen. Unless you force foo to be\nallocated in memory, which you do by labeling it \"volatile\".\n\nThis is sort of an abuse of the intuitive meaning of \"volatile\",\nbut it falls right out of what the qualifier actually means to a C\ncompiler, which is \"keep this in memory, not in a register, and\nfetch it afresh from memory every single time the program references\nit. No common-subexpression elimination here!\"\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Jun 2000 21:19:58 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: psql use of 'volatile' " } ]
[ { "msg_contents": "Is volatile being used because the variables are modified in a signal\nhandler? Maybe.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 29 Jun 2000 12:19:00 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Use of volatile" }, { "msg_contents": "* Bruce Momjian <[email protected]> [000629 09:24] wrote:\n> Is volatile being used because the variables are modified in a signal\n> handler? Maybe.\n\nYes, that's another reason to use volatile.\n\n-Alfred\n", "msg_date": "Thu, 29 Jun 2000 09:51:18 -0700", "msg_from": "Alfred Perlstein <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Use of volatile" } ]
[ { "msg_contents": "I need someone for a project I need done. The project will require upfront\npricing no hourly bids will be accepted. All interested will need to\nintergrate a database into a website. If you know how to do this email me at\[email protected]. I will give you the scoop on what I will need from\nthere.\n\nScott\n\n\n\n\n\n\n-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----\nhttp://www.newsfeeds.com - The #1 Newsgroup Service in the World!\n-----== Over 80,000 Newsgroups - 16 Different Servers! =-----\n", "msg_date": "Thu, 29 Jun 2000 12:42:53 -0500", "msg_from": "\"Scott Winterstein\" <[email protected]>", "msg_from_op": true, "msg_subject": "database into website" } ]
[ { "msg_contents": "I maintain an R package that allows one to access postgresql tables from\nwithin R. (If anyone is curious, the url is\nhttp://rpgsql.sourceforge.net.) I've found it really difficult to write\nan automated configure script that will find the postgresql library and\ninclude files, because different packagers put them in different\nplaces. I'd like to make a request. It would be convenient if there\nwere a small utility distributed with postgresql that would report the\nlocation of the library and include directories a la gnome-config. All\nthat is needed is a small script with autoconf substitutions for the\nlibrary and include paths.\n\nCheers,\nTim\n\n-- \nTimothy H. Keitt\nNational Center for Ecological Analysis and Synthesis\n735 State Street, Suite 300, Santa Barbara, CA 93101\nPhone: 805-892-2519, FAX: 805-892-2510\nhttp://www.nceas.ucsb.edu/~keitt/\n", "msg_date": "Thu, 29 Jun 2000 12:06:32 -0700", "msg_from": "\"Timothy H. Keitt\" <[email protected]>", "msg_from_op": true, "msg_subject": "finding lib/include dirs" }, { "msg_contents": "Timothy H. Keitt writes:\n\n> It would be convenient if there were a small utility distributed with\n> postgresql that would report the location of the library and include\n> directories a la gnome-config.\n\nI'm not familiar with gnome-config. Do you mean like this:\n\n$ sh pg-config.sh\npg_includedir=/usr/local/pgsql/include\npg_libdir=/usr/local/pgsql/lib\n\nHow would you find that script? Would you expect it in in the path?\n\nWe could also provide an Autoconf macro a la AC_PATH_X (perhaps as a\nwrapper around pg-config.sh).\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n\n", "msg_date": "Fri, 30 Jun 2000 18:22:50 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: finding lib/include dirs" } ]
[ { "msg_contents": "Here are a few changes to the installation layout for your approval.\n\nThe *.sample files and the initdb input files (BKI) should go into\nPREFIX/share, not lib. Since only initdb has to know about this there\nshould be no problems. As for finding these files, the easiest and safest\nmethod would be to substitute this path into initdb at build time.\nOverride with -L is of course possible. (The \"lib\" mnemonic would be gone,\nthink of it as \"location\".)\n\nAt that time, could we rename these bki files to something readable, like\n\nglobal.bki\nglobal.description\ntemplate1.bki\ntemplate1.description\n\nIt is my understanding that originally the *.bki.source files were\nconverted to *.bki at some point (forgot where), but we don't do that\nanymore. And we don't support more than one set of input files either\n(global1.bki, global2.bki?). Again, only initdb needs to know about this.\n\n\nThe odbcinst.ini file has been installed somewhere between PREFIX,\nPREFIX/etc, PREFIX/share or just /share or just /etc depending on which\nsort of installation procedure you chose or which of these directories\nalready existed. I suggest we settle on PREFIX/etc. There's still that\n--with-odbcinst option for those who prefer differently.\n\n\nThe same for the Kerberos 5 keytab file. Can't be in PREFIX/, ought to be\nin PREFIX/etc.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Fri, 30 Jun 2000 01:34:02 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Installation layout" }, { "msg_contents": "> At that time, could we rename these bki files to something readable, like\n> \n> global.bki\n> global.description\n> template1.bki\n> template1.description\n> \n> It is my understanding that originally the *.bki.source files were\n> converted to *.bki at some point (forgot where), but we don't do that\n> anymore. And we don't support more than one set of input files either\n> (global1.bki, global2.bki?). Again, only initdb needs to know about this.\n> \n\nAgreed.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 29 Jun 2000 19:52:24 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Installation layout" }, { "msg_contents": "> Here are a few changes to the installation layout for your approval.\n\nLooks fine to me.\n\n> At that time, could we rename these bki files to something readable, like\n\n> global.bki\n> global.description\n> template1.bki\n> template1.description\n\nAs long as you're renaming, maybe \".bki\" should be replaced with\nsomething clearer, like \".init\"? I never can remember what bki\nstands for anyway...\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Jun 2000 21:26:20 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Installation layout " }, { "msg_contents": "> > Here are a few changes to the installation layout for your approval.\n> \n> Looks fine to me.\n> \n> > At that time, could we rename these bki files to something readable, like\n> \n> > global.bki\n> > global.description\n> > template1.bki\n> > template1.description\n> \n> As long as you're renaming, maybe \".bki\" should be replaced with\n> something clearer, like \".init\"? I never can remember what bki\n> stands for anyway...\n\nMust be backend init, but it is a misnomer.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 29 Jun 2000 21:48:04 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Installation layout" }, { "msg_contents": " Here are a few changes to the installation layout for your approval.\n\n The *.sample files and the initdb input files (BKI) should go into\n PREFIX/share, not lib. Since only initdb has to know about this there\n should be no problems. As for finding these files, the easiest and safest\n method would be to substitute this path into initdb at build time.\n Override with -L is of course possible. (The \"lib\" mnemonic would be gone,\n think of it as \"location\".)\n\nAs a point of reference, the NetBSD pkgsrc package for postgresql\nalready puts the following in /usr/pkg/share/postgresql:\n\n global1.bki.source\n global1.description\n local1_template1.bki.source\n local1_template1.description\n\nCheers,\nBrook\n", "msg_date": "Fri, 30 Jun 2000 09:21:50 -0600 (MDT)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Installation layout" } ]
[ { "msg_contents": "I'd like to change some things about how the code handles the information\nabout its own version number.\n\nThe version number would be set in configure.in as in\n\nVERSION='1.2.3'\n\n(Keep it in configure.in because it would be nice to have this information\navailable globally. One fine day we might even automatically make tarballs\nwith this.)\n\nconfig.h defines the symbol PG_VERSION (VERSION is probably too prone to\nclashes) as \"1.2.3\", which applications can then use to display --version\ninformation of whatever.\n\nThe canonical string \"PostgreSQL x.y.z compiled by zz\" is still available.\n\nApplications that want to test for \"release compatibility\" like 7.0.1 vs\n7.0.2 can use strtod(PG_VERSION, NULL) and then use a simple `==' or `<',\netc. (Of course you're free to parse yourself \"number\"-\"dot\"-\"number\".\nThat's not hard.)\n\nRight now I think we should set it to '7.1devel', later to '7.1b1',\netc. It's really quite wrong to label it '7.1' already now.\n\nI also think that we could scrap the pg_version program. Its function is\nto create a file PG_VERSION in the directory given as argument, and it's\ncalled by initdb. Why not just substitute the version number into initdb\nat build time and just `echo > PG_VERSION'? One less level of redirection.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Fri, 30 Jun 2000 01:34:08 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Changes to handling version numbers internally" }, { "msg_contents": "On Fri, 30 Jun 2000, Peter Eisentraut wrote:\n\n> I'd like to change some things about how the code handles the information\n> about its own version number.\n> \n> The version number would be set in configure.in as in\n> \n> VERSION='1.2.3'\n> \n> (Keep it in configure.in because it would be nice to have this information\n> available globally. One fine day we might even automatically make tarballs\n> with this.)\n\nSo, we have to change and autoconf that periodically, instead of just\nhaving a simple 'version.h' file that we vi?\n\nIt doesn't sound to me like something that *belongs* in configure.in, as\nit is not required for any tests that configure requires ...\n\n\n\n", "msg_date": "Thu, 29 Jun 2000 22:11:25 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally" }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n> The version number would be set in configure.in as in\n\nMakes sense to me. Right now there is a list someplace of all the\nplaces that have to be edited to update the version number :-(.\nIf configure could help centralize/automate that, it'd be great.\n\nI also concur that supporting non-numeric suffixes on the version\nstring would be a big improvement.\n\n> I also think that we could scrap the pg_version program.\n\nYes, it seems like dead weight to me too. I think that it\npredates our use of configure, so at the time it might have\nbeen necessary mechanism, but now...\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Jun 2000 21:31:33 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "Two issues with the new versioning ideas:\n\n- The version number would be set in configure.in as in VERSION='1.2.3'\n\n- So, we have to change and autoconf that periodically, instead of just\n having a simple 'version.h' file that we vi?\n\n It doesn't sound to me like something that *belongs* in configure.in, as\n it is not required for any tests that configure requires ...\n\nWhat about having a single file (e.g., VERSION) somewhere that\ncontains the relevant information (e.g., it only contains 1.2.3). Any\nMakefiles that need it could cat the value into a make variable and\nuse it; any C files that need it could include it as appropriate.\nAlternatively, though probably not so cleanly in some ways, configure\nitself could use the value to define a variable that would be\npropagated as needed (probably only Makefile.global and config.h).\n\nAnother point:\n\n Right now I think we should set it to '7.1devel', later to '7.1b1',\n etc. It's really quite wrong to label it '7.1' already now.\n\nI suggest a numbering sequence that is strictly orderable. This is\nreally useful for the versioning of the NetBSD code, because the\nlocation of each version within a largaer sequence is immediately\nobvious. The sequence goes something like:\n\ntrunk -+--> 1.4A ---> 1.4B ---> 1.4C ---> ... -+--> 1.5A ---> 1.5B ---> ...\n | |\n | |\nreleases +--> 1.4Alpha ---> 1.4Beta ---> 1.4 +--> 1.5Alpha ---> 1.5Beta ---> 1.5\n |\n |\n +--> ... ---> 1.4.1 ---> ...\n\nThis may not exactly correspond to the NetBSD scheme, but the general\npoints should be clear: the main trunk versions are incremented in a\nnicely sortable order whenever major interfaces change (corresponding\npresumably to our need for initdb), and the branches have a similar\nclear ordering. This seems preferable to using versions like 7.1devel\nor 7.1b1 which are less formalized and less sortable.\n\nCheers,\nBrook\n", "msg_date": "Fri, 30 Jun 2000 09:17:27 -0600 (MDT)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally" }, { "msg_contents": "The Hermit Hacker writes:\n\n> It doesn't sound to me like something that *belongs* in configure.in,\n> as it is not required for any tests that configure requires ...\n\nThis way we can also make it available to shell scripts and make\nfiles. The next autoconf version will have built-in support for this\nsetup, so it's not like I'm making this up. :)\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n\n", "msg_date": "Sat, 1 Jul 2000 17:03:17 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Changes to handling version numbers internally" }, { "msg_contents": "On Sat, 1 Jul 2000, Peter Eisentraut wrote:\n\n> The Hermit Hacker writes:\n> \n> > It doesn't sound to me like something that *belongs* in configure.in,\n> > as it is not required for any tests that configure requires ...\n> \n> This way we can also make it available to shell scripts and make\n> files. The next autoconf version will have built-in support for this\n> setup, so it's not like I'm making this up. :)\n\nI see you've already committed the changes, which I still disagree with\n... I don't like the fact that someone has to remember to update the\nconfigure.in and autoconf it to change the version number. We already\nhave the mess with template files, why not create a VERSION file that\ncould just as easily be read in by configure, without having to do an\nautoconf, to change the version?\n\n\n\n", "msg_date": "Sun, 2 Jul 2000 22:41:20 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally" }, { "msg_contents": "The Hermit Hacker <[email protected]> writes:\n>> This way we can also make it available to shell scripts and make\n>> files. The next autoconf version will have built-in support for this\n>> setup, so it's not like I'm making this up. :)\n\n> I see you've already committed the changes, which I still disagree with\n> ... I don't like the fact that someone has to remember to update the\n> configure.in and autoconf it to change the version number.\n\nIt's a lot better than it was before ;-)\n\n> We already have the mess with template files, why not create a VERSION\n> file that could just as easily be read in by configure, without having\n> to do an autoconf, to change the version?\n\nGilding the lily. The people who are likely to be bumping the version\nnumber are all perfectly familar with autoconf. Also, we've already\ngot bad-enough problems with dependencies (see nearby thread) without\nintroducing a dependency from a file that's read by configure to\nwho-knows-where --- a dependency that can't be handled by 'make depend'.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 02 Jul 2000 21:50:17 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "On Sun, 2 Jul 2000, Tom Lane wrote:\n\n> The Hermit Hacker <[email protected]> writes:\n> >> This way we can also make it available to shell scripts and make\n> >> files. The next autoconf version will have built-in support for this\n> >> setup, so it's not like I'm making this up. :)\n> \n> > I see you've already committed the changes, which I still disagree with\n> > ... I don't like the fact that someone has to remember to update the\n> > configure.in and autoconf it to change the version number.\n> \n> It's a lot better than it was before ;-)\n> \n> > We already have the mess with template files, why not create a VERSION\n> > file that could just as easily be read in by configure, without having\n> > to do an autoconf, to change the version?\n> \n> Gilding the lily. The people who are likely to be bumping the version\n> number are all perfectly familar with autoconf. \n\nYa, I've been fighting myself internally on this point ... I just hate the\nthough of embed'ng it into configure, that's all ...\n\npeter mentioned something about next version of autoconf doing this? \n\n", "msg_date": "Sun, 2 Jul 2000 23:19:31 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "On Sun, 2 Jul 2000, The Hermit Hacker wrote:\n\n> peter mentioned something about next version of autoconf doing this? \n\nYeah, you write AC_INIT(Package, Version) and that creates the VERSION\nvariable and substitutes it, just like I did. (It also creates the PACKAGE\nvariable, in case you ever want to rename the package. :)\n\nIf you don't like it in configure.in, I understand, but it seems like that\nwould be just one more file to find.\n\n-- \nPeter Eisentraut Sernanders vaeg 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Mon, 3 Jul 2000 08:33:15 -0400 (EDT)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "On Mon, 3 Jul 2000 [email protected] wrote:\n\n> On Sun, 2 Jul 2000, The Hermit Hacker wrote:\n> \n> > peter mentioned something about next version of autoconf doing this? \n> \n> Yeah, you write AC_INIT(Package, Version) and that creates the VERSION\n> variable and substitutes it, just like I did. (It also creates the PACKAGE\n> variable, in case you ever want to rename the package. :)\n> \n> If you don't like it in configure.in, I understand, but it seems like that\n> would be just one more file to find.\n\nTo me, it just seems that it will more often be forgotten then putting it\ninto a VERSION text file ... but it isn't something that I'm going to turn\nblue in face over ... since its as 'clean' as above, leave it ... \n\nOne thought, and with all your recent changes you may have done this, but\nis there any way of putting some sort of text file as part of the install\nprocedure that \"stores\" the ./configure command line? Something that, if\nI were to remove all sources but decide later that I want to add, let's\nsay, multi-byte, I could do 'cat ~pgsql/CONFIGURE' to find out how the\nsystem is already compiled? \n\n\n", "msg_date": "Tue, 4 Jul 2000 14:40:32 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "The Hermit Hacker <[email protected]> writes:\n> One thought, and with all your recent changes you may have done this, but\n> is there any way of putting some sort of text file as part of the install\n> procedure that \"stores\" the ./configure command line? Something that, if\n> I were to remove all sources but decide later that I want to add, let's\n> say, multi-byte, I could do 'cat ~pgsql/CONFIGURE' to find out how the\n> system is already compiled? \n\nconfig.status has always done this.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 15:56:31 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "The Hermit Hacker writes:\n\n> One thought, and with all your recent changes you may have done this, but\n> is there any way of putting some sort of text file as part of the install\n> procedure that \"stores\" the ./configure command line? Something that, if\n> I were to remove all sources but decide later that I want to add, let's\n> say, multi-byte, I could do 'cat ~pgsql/CONFIGURE' to find out how the\n> system is already compiled? \n\nconfig.status stores the current configuration. To reuse it, run\n`./config.status --recheck'. If you just want to *know* what the current\nconfiguration is, then `head config.status' should answer that pretty\nwell.\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Wed, 5 Jul 2000 02:12:24 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "\nyou guys missed the key part of my message \"if I were to remove all\nsources but decide later ...\" :) if I remove sources, I no longer have\nconfig-status to work with :)\n\nOn Wed, 5 Jul 2000, Peter Eisentraut wrote:\n\n> The Hermit Hacker writes:\n> \n> > One thought, and with all your recent changes you may have done this, but\n> > is there any way of putting some sort of text file as part of the install\n> > procedure that \"stores\" the ./configure command line? Something that, if\n> > I were to remove all sources but decide later that I want to add, let's\n> > say, multi-byte, I could do 'cat ~pgsql/CONFIGURE' to find out how the\n> > system is already compiled? \n> \n> config.status stores the current configuration. To reuse it, run\n> `./config.status --recheck'. If you just want to *know* what the current\n> configuration is, then `head config.status' should answer that pretty\n> well.\n> \n> -- \n> Peter Eisentraut Sernanders v�g 10:115\n> [email protected] 75262 Uppsala\n> http://yi.org/peter-e/ Sweden\n> \n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n", "msg_date": "Wed, 5 Jul 2000 19:31:02 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "The Hermit Hacker <[email protected]> writes:\n> you guys missed the key part of my message \"if I were to remove all\n> sources but decide later ...\" :) if I remove sources, I no longer have\n> config-status to work with :)\n\nNo, I didn't miss it. IMHO \"make distclean\" should remove every file\ncreated by configure --- that would include any specialized \"CONFIGURE\"\noutput file you might want it to create. So there's no advantage to\nhaving that over the existing config.status output file.\n\nIf you don't want to lose the configure outputs then do \"make clean\",\nnot \"make distclean\".\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 05 Jul 2000 18:44:50 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "The Hermit Hacker writes:\n\n> you guys missed the key part of my message \"if I were to remove all\n> sources but decide later ...\" :) if I remove sources, I no longer have\n> config-status to work with :)\n\nWell, if you remove everything, then where is that configure command line\nsaving file going to survive?\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Thu, 6 Jul 2000 02:19:02 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "On Thu, 6 Jul 2000, Peter Eisentraut wrote:\n\n> The Hermit Hacker writes:\n> \n> > you guys missed the key part of my message \"if I were to remove all\n> > sources but decide later ...\" :) if I remove sources, I no longer have\n> > config-status to work with :)\n> \n> Well, if you remove everything, then where is that configure command line\n> saving file going to survive?\n\nthat is what I'm kinda asking for ... maybe some POSTGRES/etc directory\ncan be created, or create a file under lib that just contains that one\ncommand?\n\nthink about Lamar's RPMs ... might not be many that would, but some might\nlike to be able to look at a file and see what args he used to configure\nit, in case they want to keep what is currently there, but add an option\nthat he didn't put in ...\n\n", "msg_date": "Wed, 5 Jul 2000 23:30:38 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "The Hermit Hacker writes:\n\n> that is what I'm kinda asking for ... maybe some POSTGRES/etc directory\n> can be created, or create a file under lib that just contains that one\n> command?\n> \n> think about Lamar's RPMs ... might not be many that would, but some might\n> like to be able to look at a file and see what args he used to configure\n> it, in case they want to keep what is currently there, but add an option\n> that he didn't put in ...\n\nGood point. (Although the RPMs do way too much custom patching right now\nfor this to be very useful.)\n\nSomeone recently asked for a utility along the lines of gnome-config so\nthey can get the installation paths for the libraries and include files\nfor their package. Maybe we can work the configure command line into that.\n\n$ pg-config --includedir\n/usr/local/pgsql/include\n$ pg-config --libdir\n/usr/local/pgsql/lib\n$ pg-config --configure\n--enable-debug --with-CXX\n\nComments?\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Sat, 8 Jul 2000 02:09:51 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n> Someone recently asked for a utility along the lines of gnome-config so\n> they can get the installation paths for the libraries and include files\n> for their package. Maybe we can work the configure command line into that.\n\n> $ pg-config --includedir\n> /usr/local/pgsql/include\n> $ pg-config --libdir\n> /usr/local/pgsql/lib\n> $ pg-config --configure\n> --enable-debug --with-CXX\n\n> Comments?\n\nI like it. We can still say that make distclean wipes the source tree\nclean of configure output ... but the install tree retains documentation\nabout how it was built.\n\nAside from configure options, it'd at least be important to save the\nselected OS template name. Not sure what else.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 07 Jul 2000 20:28:23 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " }, { "msg_contents": "On Sat, 8 Jul 2000, Peter Eisentraut wrote:\n\n> The Hermit Hacker writes:\n> \n> > that is what I'm kinda asking for ... maybe some POSTGRES/etc directory\n> > can be created, or create a file under lib that just contains that one\n> > command?\n> > \n> > think about Lamar's RPMs ... might not be many that would, but some might\n> > like to be able to look at a file and see what args he used to configure\n> > it, in case they want to keep what is currently there, but add an option\n> > that he didn't put in ...\n> \n> Good point. (Although the RPMs do way too much custom patching right now\n> for this to be very useful.)\n> \n> Someone recently asked for a utility along the lines of gnome-config so\n> they can get the installation paths for the libraries and include files\n> for their package. Maybe we can work the configure command line into that.\n> \n> $ pg-config --includedir\n> /usr/local/pgsql/include\n> $ pg-config --libdir\n> /usr/local/pgsql/lib\n> $ pg-config --configure\n> --enable-debug --with-CXX\n> \n> Comments?\n\nThat would be perfect also\n\n\n", "msg_date": "Sat, 8 Jul 2000 13:02:34 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Changes to handling version numbers internally " } ]
[ { "msg_contents": "\nIf anyone is interested in being sent the current sources for the new\npg_dump & pg_restore, please let me know.\n\nThe utilities now seem to work, but need testing.\n\nThe basic idea is to use pg_dump to dump an *entire* database, and then use\npg_restore to choose what is restored.\n\nThe salient features are as follows:\n\n- pg_dump still used to dump database; all output is via new interface\n(virtually all of the pg_dump code is changed, but not the logic). The\nchanges are relatively minor, all the same.\n\n- the '-c' option is not used in pg-dump: it now dumps the commands to\ndelete the schema, and it is up to the user of pg_restore to decide if they\nare output.\n\n- the default output file format is a custom format with compressed\nsections (the data dumps). It is NOT a text file.\n\n- pg_restore reads the backup file and, depending on the options chosen,\nproduces a script (to stdout) that can be sent to psql.\n\n- by default pg_restore outputs the schema/data in the order it was sent\nfrom pg_dump, but the --oid flag will send the output in order of\nincreasing OID, and the --rearrange flag will put all 'non-parental' (??)\nitems at the end, after the data. (eg. indexes, acls, triggers etc).\nNeedless to say that the best results com from using both of these options.\n\n- If the -c (clear) option is chosen in pg_restore, it also dumps the\n'drop' commands in reverse order at the start of the script. This *should*\nmake it more reliable than dumping them when the item is defined. It also\nmeans that triggers can be dropped.\n\n- The --toc option shows a summary of the restore operation that would be\nperformed if the --toc were not there.\n\nPlease send me an email if you are interested and have the time to test them.\n\nThanks,\n\nPhilip Warner.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Fri, 30 Jun 2000 15:27:53 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Modified pg_dump & new pg_restore need testing..." }, { "msg_contents": "\nSounds like something that is a suitable replacement for what we have now\n... I particularly like the 'compressed output' *but* is there some way,\nfor the paranoid, of doing an uncompressed format\n(--dont-compress?)? Then again, how does this handle someone wanting to\ndump the database for reloading into a non-PgSQL database? \n\nOn Fri, 30 Jun 2000, Philip Warner wrote:\n\n> \n> If anyone is interested in being sent the current sources for the new\n> pg_dump & pg_restore, please let me know.\n> \n> The utilities now seem to work, but need testing.\n> \n> The basic idea is to use pg_dump to dump an *entire* database, and then use\n> pg_restore to choose what is restored.\n> \n> The salient features are as follows:\n> \n> - pg_dump still used to dump database; all output is via new interface\n> (virtually all of the pg_dump code is changed, but not the logic). The\n> changes are relatively minor, all the same.\n> \n> - the '-c' option is not used in pg-dump: it now dumps the commands to\n> delete the schema, and it is up to the user of pg_restore to decide if they\n> are output.\n> \n> - the default output file format is a custom format with compressed\n> sections (the data dumps). It is NOT a text file.\n> \n> - pg_restore reads the backup file and, depending on the options chosen,\n> produces a script (to stdout) that can be sent to psql.\n> \n> - by default pg_restore outputs the schema/data in the order it was sent\n> from pg_dump, but the --oid flag will send the output in order of\n> increasing OID, and the --rearrange flag will put all 'non-parental' (??)\n> items at the end, after the data. (eg. indexes, acls, triggers etc).\n> Needless to say that the best results com from using both of these options.\n> \n> - If the -c (clear) option is chosen in pg_restore, it also dumps the\n> 'drop' commands in reverse order at the start of the script. This *should*\n> make it more reliable than dumping them when the item is defined. It also\n> means that triggers can be dropped.\n> \n> - The --toc option shows a summary of the restore operation that would be\n> performed if the --toc were not there.\n> \n> Please send me an email if you are interested and have the time to test them.\n> \n> Thanks,\n> \n> Philip Warner.\n> \n> \n> ----------------------------------------------------------------\n> Philip Warner | __---_____\n> Albatross Consulting Pty. Ltd. |----/ - \\\n> (A.C.N. 008 659 498) | /(@) ______---_\n> Tel: (+61) 0500 83 82 81 | _________ \\\n> Fax: (+61) 0500 83 82 82 | ___________ |\n> Http://www.rhyme.com.au | / \\|\n> | --________--\n> PGP key available upon request, | /\n> and from pgp5.ai.mit.edu:11371 |/\n> \n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n", "msg_date": "Fri, 30 Jun 2000 13:53:36 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Modified pg_dump & new pg_restore need testing..." }, { "msg_contents": "At 13:53 30/06/00 -0300, The Hermit Hacker wrote:\n>\n>Sounds like something that is a suitable replacement for what we have now\n>... I particularly like the 'compressed output' *but* is there some way,\n>for the paranoid, of doing an uncompressed format\n>(--dont-compress?)? Then again, how does this handle someone wanting to\n>dump the database for reloading into a non-PgSQL database? \n>\n\nThere will be; you can specify -z0 on the pg_dump, but it still has binary\njunk in the output. I will add a '--plain-text, -p' option to dump it into\ntext. You can also use pg_restore to dump the file to text.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Sat, 01 Jul 2000 09:52:57 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Modified pg_dump & new pg_restore need testing..." }, { "msg_contents": "Philip, where did we leave this?\n\n> \n> If anyone is interested in being sent the current sources for the new\n> pg_dump & pg_restore, please let me know.\n> \n> The utilities now seem to work, but need testing.\n> \n> The basic idea is to use pg_dump to dump an *entire* database, and then use\n> pg_restore to choose what is restored.\n> \n> The salient features are as follows:\n> \n> - pg_dump still used to dump database; all output is via new interface\n> (virtually all of the pg_dump code is changed, but not the logic). The\n> changes are relatively minor, all the same.\n> \n> - the '-c' option is not used in pg-dump: it now dumps the commands to\n> delete the schema, and it is up to the user of pg_restore to decide if they\n> are output.\n> \n> - the default output file format is a custom format with compressed\n> sections (the data dumps). It is NOT a text file.\n> \n> - pg_restore reads the backup file and, depending on the options chosen,\n> produces a script (to stdout) that can be sent to psql.\n> \n> - by default pg_restore outputs the schema/data in the order it was sent\n> from pg_dump, but the --oid flag will send the output in order of\n> increasing OID, and the --rearrange flag will put all 'non-parental' (??)\n> items at the end, after the data. (eg. indexes, acls, triggers etc).\n> Needless to say that the best results com from using both of these options.\n> \n> - If the -c (clear) option is chosen in pg_restore, it also dumps the\n> 'drop' commands in reverse order at the start of the script. This *should*\n> make it more reliable than dumping them when the item is defined. It also\n> means that triggers can be dropped.\n> \n> - The --toc option shows a summary of the restore operation that would be\n> performed if the --toc were not there.\n> \n> Please send me an email if you are interested and have the time to test them.\n> \n> Thanks,\n> \n> Philip Warner.\n> \n> \n> ----------------------------------------------------------------\n> Philip Warner | __---_____\n> Albatross Consulting Pty. Ltd. |----/ - \\\n> (A.C.N. 008 659 498) | /(@) ______---_\n> Tel: (+61) 0500 83 82 81 | _________ \\\n> Fax: (+61) 0500 83 82 82 | ___________ |\n> Http://www.rhyme.com.au | / \\|\n> | --________--\n> PGP key available upon request, | /\n> and from pgp5.ai.mit.edu:11371 |/\n> \n\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 10 Oct 2000 21:01:02 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Modified pg_dump & new pg_restore need testing..." }, { "msg_contents": "At 21:01 10/10/00 -0400, Bruce Momjian wrote:\n>Philip, where did we leave this?\n\nIn CVS. ;-).\n\nThe longer answer is that it's been in CVS for a while, and I have a\nversion for 7.0.2 which I have also been using for a while. Various people\nhave tested it, but not as many or as much as I would like (or lots of\npeople have tested it with no problems, which seems unlikely).\n\nThe docs for pg_dump are with Thomas and I am about to start working on an\nunrelated bug in the way pg_dump handles sequences. When that's done, I\nwill document pg_restore. Hopefully, all in time for beta...\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.B.N. 75 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 11 Oct 2000 12:13:15 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Modified pg_dump & new pg_restore need testing..." } ]
[ { "msg_contents": "Hi,\n\n For speed I had a programme (actually a CORBA server) that would\nreceive a lot of records, copy them into a binary record and copied them\ninto the database. As this is a CORBA server on the same machine as the\ndatabase, there are no endedness problems or anything like that. This\nworked just great until the latest update: now I get 'COPY BINARY is not\nsupported to stdout or from stdin'. In the context of psql there may be\ngood reasons for this, but from a programme it works well, so why\ndisallow it? Now I've got to go and convert everything to a load of\nprintf's, which are then converted back into binary -- seems kind-of\nsilly. And I'm not comfortable with these types of conversions for\nfloats at all.\n\nOn a separate note: what would be really great for loading large amounts\nof data into a database is if one could have a serial field that\nupdated anyway. Some databases give you the next serial value if you try\nto insert null -- which seems reasonable in the context of a SERIAL,\nwhich is NOT NULL. I can see that this may cause problems in the case of\nfields that do a default nextval('sequence'), and which consequently can\nbe null, but the benefits would be large.\n\nOften simulations or sensors (we have both) generate large streams of\nbinary data and a way of loading this data into the database rapidly is\nvery, very useful.\n\nAdriaan\n\n", "msg_date": "Fri, 30 Jun 2000 11:04:11 +0300", "msg_from": "Adriaan Joubert <[email protected]>", "msg_from_op": true, "msg_subject": "copy binary from stdin; Why is it broken?" }, { "msg_contents": "Hi everybody...\nI need a clue...Maybe someone of you has something to suggest :)\nI'm developing a server with mico and postgresql in c++... The access to\npostgresql\nis done using libpq, not via corba...The compiler I use is gcc-2.95-2\nThe operating systems are Linux and Solaris 2.7.\nOn Solaris I'm trying to use Purify 5.1 but it founds a memory segment\nerror\nduring mico initialization and my program dumps core. Then I've get out\nmico and even Xerces (an XML parser) out of the server but then purify\ncomplains about an error in a an fstream function, like:\n IPR: Invalid pointer read\n This is occurring while in:\n ifstream::~ifstream() [fstream.cc:110]\n {\n return \"w\";\n }\n => #endif\n Configuration::loadConfig(const char*) [Configuration.o]\n DSControl_impl::start() [ds_sds_nocorba_impl.o]\n main [ds_sds_nocorba_main.o]\n _start [crt1.o]\n Reading 4 bytes from 0xff1927a4 between the heap and the stack.\n\nI've been forced to use libpq from version 6.5.3 because if I link the\n7.0.2\nversion in then I get an error from a function inside PostgreSQL setdb\nThe error is always the same: the program is accessing an area between the\nheap and the stack......but just migrates around corba, ifstream,\npostgres.....\nWell... Before I go squashing my head on the wall again, dou you have an\nidea on\nhow to get more info on this ??????????\nI know this is a bit OT here, because I don't think postgres is really\ninvolved in this,\nsimply I don't know what to do with this... Could it be simply a purify\nproblem ????\nThe system seems to work fine without it....(seems......)\n\nThanks for your patience....\n Fabrizio\n\n--\nFabrizio Sciarra ! Tel +39 050 545 111\nIntecs sistemi spa ! Fax +39 050 545 200\nvia Gereschi 32-34, 56100 Pisa ITALY. ! [email protected]\n-- My opinions are mine, not necessarly of my employers --\n * Linux. The choice of a GNU generation\n * Java Lobby Member * Ada community member.\n\n\n\n\nHi everybody...\nI need a clue...Maybe someone of you has something to suggest :)\nI'm developing a server with mico and postgresql in c++... The access\nto postgresql\nis done using libpq, not via corba...The compiler I use is gcc-2.95-2\nThe operating systems are Linux and Solaris 2.7.\nOn Solaris I'm trying to use Purify 5.1 but it founds a memory segment\nerror\nduring mico initialization and my program dumps core. Then I've get\nout mico and even Xerces (an XML parser) out of the server but then purify\ncomplains about an error in a an fstream function, like:\n      IPR: Invalid pointer read\n      This is occurring while in:\n           \nifstream::~ifstream() [fstream.cc:110]\n              \n{\n                  \nreturn \"w\";\n              \n}\n           \n=> #endif\n           \nConfiguration::loadConfig(const char*) [Configuration.o]\n           \nDSControl_impl::start() [ds_sds_nocorba_impl.o]\n           \nmain           [ds_sds_nocorba_main.o]\n           \n_start         [crt1.o]\n      Reading 4 bytes from 0xff1927a4 between\nthe heap and the stack.\nI've been forced to use libpq from version 6.5.3 because if I link the\n7.0.2\nversion in then I get an error from a function inside PostgreSQL setdb\nThe error is always the same: the program is accessing an area between\nthe\nheap and the stack......but just migrates around corba, ifstream, postgres.....\nWell... Before I go squashing my head on the wall again, dou you have\nan idea on\nhow to get more info on this ??????????\nI know this is a bit OT here, because I don't think postgres is\nreally involved in this,\nsimply I don't know what to do with this... Could it be simply a purify\nproblem ????\nThe system seems to work fine without it....(seems......)\nThanks for your patience....\n  Fabrizio\n-- \nFabrizio Sciarra                      ! Tel +39 050 545 111\nIntecs sistemi spa                    ! Fax +39 050 545 200\nvia Gereschi 32-34, 56100 Pisa ITALY. ! [email protected]\n-- My opinions are mine, not necessarly of my employers --\n * Linux. The choice of a GNU generation\n * Java Lobby Member * Ada community member.", "msg_date": "Fri, 30 Jun 2000 10:40:56 -0400", "msg_from": "Fabrizio Sciarra <[email protected]>", "msg_from_op": false, "msg_subject": "PostgreSQL + Solaris + Mico + Purify = ????" } ]
[ { "msg_contents": "Thus spake Jakub Bartosz Bielecki\n> I'm not really into PostgreSQL development, first of all.\n> However once I have needed to use INET type\n> in my database and (sadly) I noticed 2 very serious bugs:\n> \n> 1. all functions which return text (host, network etc) should\n> \treturn pascal-like ASCII string not C-like ASCIIZ\n> example:\n> \tselect host('10.0.0.1')='10.0.0.1'; -- this returns false!\n> \n> 2. inet comparison routines (and thus operators) work in a strange way \n> if I set netmask length to a non-default value (other than 32).\n> example:\n> \tselect '10.0.0.1/27'::inet='10.0.0.2/27'::inet; -- returns true\n> I guess that this behaviour is different from described in manual.\n> And if it's right, then in surce code is 1 function which behaves\n> in an opposite way, messing the whole thing up...\n> \n> I noticed these in Pg 6.5.3, however when I checked apropriate source file\n> in Pg 7.0, to my suprise it was not fixed...\n> \n> OK. Now my questions:\n> - am I right?\n> - are you working on it?\n> - if not, should I fix it myself and send a patch to developers?\n> (I really dont feel like duplicating someones work)\n\nI am working on different things right now so go ahead but do discuss your\nproposed changes on hackers first ([email protected]) as there\nwas quite a lot of discussion at the time about how it should work.\n\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n", "msg_date": "Fri, 30 Jun 2000 07:54:41 -0400 (EDT)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: postgres - development of inet/cidr" }, { "msg_contents": "\"D'Arcy J.M. Cain\" wrote:\n> \n> Thus spake Jakub Bartosz Bielecki\n> > 1. all functions which return text (host, network etc) should\n> > return pascal-like ASCII string not C-like ASCIIZ\n> > example:\n> > select host('10.0.0.1')='10.0.0.1'; -- this returns false!\n\nWhich it should not. In a text-to-text comparison, both have to be\nequal.\n\n> > 2. inet comparison routines (and thus operators) work in a strange way\n> > if I set netmask length to a non-default value (other than 32).\n> > example:\n> > select '10.0.0.1/27'::inet='10.0.0.2/27'::inet; -- returns true\n> > I guess that this behaviour is different from described in manual.\n\nThis would be proper behaviour for the cidr datatype, which describes a\nnetwork. \"select '10.0.0.1/27'::cidr='10.0.0.2/27'::cidr;\" has to return\ntrue, as both define the same network, the mask putting the 1 vs. 2\noutside the comparison scope. \n\nOn inet, I consider the above broken - going by the documentation,\nhaving a netmask on a inet datatype does not define a network address\nbut rather supplies additional information on the cidr network the host\nas specified by the address is in. Accordingly, it should only truncate\nif the comparison casts to cidr. \n\nThe big question is whether comparisons that only work on a cidr data\ntype (contains/contained) or have a cidr type on one side can safely\ncast the inet type to cidr implicitly. For: \n\n\"select '10.0.0.1/27'::inet = '10.0.0.2/27'::inet;\" FALSE \n\"select '10.0.0.1/27'::cidr = '10.0.0.2/27'::cidr;\" TRUE\n\"select '10.0.0.1/27'::cidr = '10.0.0.2/27'::inet;\" FALSE \n\"select '10.0.0.1/27'::cidr >> '10.0.0.2/27'::inet;\" TRUE \n\"select '10.0.0.1/27'::cidr << '10.0.0.2/27'::inet;\" ERROR \n\nit looks sane, IMHO. But we need to reach an agreement on the proper\nbehaviour on greater/smaller comparisons. Should:\n\n\"select '10.0.0.1/27'::inet > '10.0.0.2/27'::cidr;\" \n\nbe true or false? Casting to cidr prior to comparison would make it\nequivalent to \"select '10.0.0.0/27'::cidr > '10.0.0.0/27'::cidr;\", which\nis false, both networks being equal. But we have at least three possible\ncomparisons of host to network - besides comparing net to net, the host\nmight be compared to the the base or top address of the network. That\nsituation can only be resolved doing explicit casts, so throwing an\nerror in these cases is IMHO indicated.\n\nSevo\n\n-- \[email protected]\n", "msg_date": "Mon, 03 Jul 2000 16:41:02 +0200", "msg_from": "Sevo Stille <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: postgres - development of inet/cidr" }, { "msg_contents": "\n\nOn Mon, 3 Jul 2000, Sevo Stille wrote:\n> \n> This would be proper behaviour for the cidr datatype, which describes a\n> network. \"select '10.0.0.1/27'::cidr='10.0.0.2/27'::cidr;\" has to return\n> true, as both define the same network, the mask putting the 1 vs. 2\n> outside the comparison scope. \n> \n> On inet, I consider the above broken - going by the documentation,\n> having a netmask on a inet datatype does not define a network address\n> but rather supplies additional information on the cidr network the host\n> as specified by the address is in. Accordingly, it should only truncate\n> if the comparison casts to cidr. \n\nOK. After some inspection in list's archives I found the following\nstatement (http://www.postgresql.org/mhonarc/pgsql-hackers/1998-07):\n> It does not work that way. /24 is\n> not a shorthand for specifying a netmask -- in CIDR, it's a \"prefix\n> length\".\n> That means \"192.7.34.21/24\" is either (a) a syntax error or\n> (b) equivilent to \"192.7.34/24\".\n\nEverybody seemed to agree with the above opinion at that time.\n\nThis is obviously _not_ the way that CIDR is handled at this moment.\n\"select '1.2.3.4/24'\" returns \"1.2.3/24\" only because the _output_ routine\nsilently cuts host bits. Input routine stores it exactly as '1.2.3.4/24'.\n\nSince IMHO it's wrong I prepared a patch (I'm sending it to pgsql-patch).\nIt fixes the CIDR input routine to zero host bits (ie beyond-prefix bits).\nPlease note that I didn't change the INET input routine.\n\nEventually I had to change a bit comparison functions.\nTo this moment they worked in a CIDR way (didn't compare host bits at all)\nalthough they were used by both INET and CIDR.\nSince CIDR is zero-padded now, whole 32 bits are compared by > = <\noperators.\nSubnet operators <<, >> are still the same, don't compare host bits.\n\n> The big question is whether comparisons that only work on a cidr data\n> type (contains/contained) or have a cidr type on one side can safely\n> cast the inet type to cidr implicitly. For: \n> \"select '10.0.0.1/27'::inet = '10.0.0.2/27'::inet;\" FALSE\n> \"select '10.0.0.1/27'::cidr = '10.0.0.2/27'::cidr;\" TRUE\n> \"select '10.0.0.1/27'::cidr = '10.0.0.2/27'::inet;\" FALSE\n> \"select '10.0.0.1/27'::cidr >> '10.0.0.2/27'::inet;\" TRUE\nOK.\n> \"select '10.0.0.1/27'::cidr << '10.0.0.2/27'::inet;\" ERROR\n\nCurrently it's not an error... There is no way (and no reason) to\ndistinguish between INET and CIDR. Above example is exactly\nequivalent to:\n\tselect '10.0.0.0/27'::inet << '10.0.0.2/27'::inet; -- FALSE\nbut:\n\tselect '10.0.0.0/27'::inet <<= '10.0.0.2/27'::inet; -- TRUE\n\n> But we need to reach an agreement on the proper\n> behaviour on greater/smaller comparisons. Should:\n> \n> \"select '10.0.0.1/27'::inet > '10.0.0.2/27'::cidr;\" \n> \n> be true or false? Casting to cidr prior to comparison would make it\n> equivalent to \"select '10.0.0.0/27'::cidr > '10.0.0.0/27'::cidr;\", which\n> is false, both networks being equal. \n\nIt should be (and is!) true... Since second argument is\nreally '10.0.0.0/27'.\n\n", "msg_date": "Wed, 5 Jul 2000 14:43:49 +0200 (MET DST)", "msg_from": "Jakub Bartosz Bielecki <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: postgres - development of inet/cidr" }, { "msg_contents": "Jakub Bartosz Bielecki wrote:\n\n> > \"select '10.0.0.1/27'::cidr << '10.0.0.2/27'::inet;\" ERROR\n> \n> Currently it's not an error... There is no way (and no reason) to\n> distinguish between INET and CIDR. \n\nYes, there is. CIDR is defined as the network 10.0.0.1 & /27, while INET\nis defined as host 10.0.0.1 within network 10.0.0.1 & /27. You can do\nalmost every network and host calculation both in CIDR and INET, but\nyou need implicit knowledge for it. Two columns are necessary to define\na host and its network in CIDR, and a network cannot be specified\nwithout a host using INET - except for ugly in-band hacks like using\n10.0.0.0/27 for the network which would prevent you from specifying a\nbase address.\n\n> Above example is exactly\n> equivalent to:\n> select '10.0.0.0/27'::inet << '10.0.0.2/27'::inet; -- FALSE\n\nNope. If the right hand side is automatically propagated to a network,\nit is true. If not, the above IMHO should better raise an error, as a\nhost can never contain a host. \n\n> but:\n> select '10.0.0.0/27'::inet <<= '10.0.0.2/27'::inet; -- TRUE\n\nWell, you might argue that a host could contain-or-equal a host, but as\nonly the equals part could ever be true, that is a redundant operator\nwithout any meaning beyond equals, and accordingly it should not be\nvalid for that case.\n \n> > But we need to reach an agreement on the proper\n> > behaviour on greater/smaller comparisons. Should:\n> >\n> > \"select '10.0.0.1/27'::inet > '10.0.0.2/27'::cidr;\"\n> >\n> > be true or false? Casting to cidr prior to comparison would make it\n> > equivalent to \"select '10.0.0.0/27'::cidr > '10.0.0.0/27'::cidr;\", which\n> > is false, both networks being equal.\n> \n> It should be (and is!) true... Since second argument is\n> really '10.0.0.0/27'.\n\nYes, but that does not make it any truer. CIDR 10.0.0.0/27 is\ndefinitively not 10.0.0.0 but [10.0.0.0 .. 10.0.0.31]. A CIDR address is\nnever synonymous to a plain host address. You'll see the problem if you\ntry to calculate the inverse - any zeroed CIDR address in the entire\nrange from 10.0/8 to 10.0.0.0/32 would mask to 10.0.0.0. Accordingly,\nthere is no simple answer to a \"host bigger/smaller than network\"\nquestion. For many applications, it may be useful to define that to mean\nthat the host is smaller than the network bottom address respectively\nbigger than the top address, but any of the other possible views would\nbe perfectly legal as well.\n \nSevo\n\n-- \[email protected]\n", "msg_date": "Wed, 05 Jul 2000 15:48:09 +0200", "msg_from": "Sevo Stille <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: postgres - development of inet/cidr" }, { "msg_contents": "\n\nOn Wed, 5 Jul 2000, Sevo Stille wrote:\n> \n> > > \"select '10.0.0.1/27'::cidr << '10.0.0.2/27'::inet;\" ERROR\n> > \n> > Currently it's not an error... There is no way (and no reason) to\n> > distinguish between INET and CIDR. \n> \n> Yes, there is. CIDR is defined as the network 10.0.0.1 & /27, while INET\n> is defined as host 10.0.0.1 within network 10.0.0.1 & /27. You can do\n> almost every network and host calculation both in CIDR and INET, but\n> you need implicit knowledge for it.\n\nI was talking about *current* implementation of INET/CIDR (which IMHO \nis very ill). \nThere is INET for users that want simply to store IP's and don't care\nabout all the technical jargon.\nThere is CIDR for advanced users who want to store network data.\n\nCurrently these 2 types are handled by 1 implementation, moreover despite\nINET netmask and CIDR prefix-length are something completely different,\nboth are stored in the same field of inet structure (yuck).\n\nAt the moment it works fine. But that's only a hack.\nI guess the purpose was to prevent duplication of code... Blah...\n\n> > select '10.0.0.0/27'::inet << '10.0.0.2/27'::inet; -- FALSE\n> \n> Nope. If the right hand side is automatically propagated to a network,\n> it is true. If not, the above IMHO should better raise an error, as a\n> host can never contain a host. \n> \n> > select '10.0.0.0/27'::inet <<= '10.0.0.2/27'::inet; -- TRUE\n> \n> Well, you might argue that a host could contain-or-equal a host, but as\n> only the equals part could ever be true, that is a redundant operator\n> without any meaning beyond equals, and accordingly it should not be\n> valid for that case.\n> \n> > > \"select '10.0.0.1/27'::inet > '10.0.0.2/27'::cidr;\"\n> > It should be (and is!) true... Since second argument is\n> > really '10.0.0.0/27'.\n> \n> Yes, but that does not make it any truer. CIDR 10.0.0.0/27 is\n> definitively not 10.0.0.0 but [10.0.0.0 .. 10.0.0.31].\n\nSame as above... You are perfectly right.\n\nEverything works until user starts messing with _both_ INET and CIDR\nat the same time.\n\nThe possible solution is:\n- inhibit cidr-to-inet cast (and maybe also inet-to-cidr, because\n it would throw away netmask),\n- CIDR operators: > = < << >>\n- INET operators: > = < (and why not & | if it would be useful???)\n functions:\tcidr network(inet); \t// '10.0.0.0/27'\n\t\t\ttext host(inet); \t// '10.0.0.1'\n\t\t\tint masklen(inet); \t// 27\n- write an usable manual.\n\nComments?\nI *might* work on it if I find some spare time. But it's unlikely :(\n\n", "msg_date": "Thu, 6 Jul 2000 14:52:17 +0200 (MET DST)", "msg_from": "Jakub Bartosz Bielecki <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: postgres - development of inet/cidr" } ]
[ { "msg_contents": "\nA minor nasty error I got when trying to improve the query used to disable\ntriggers:\n\ncreate table t1(f1 int4, f2 int4);\ncreate table t2(f1 int4, f2 int4);\n\ninsert into t1 values(1, 0);\ninsert into t1 values(2, 0);\n\ninsert into t2 values(1, 0);\n\nupdate t1 set f2=count(*) from t2 where t1.f1=1 and t2.f1=t1.f1 ;\nUPDATE 1\n\nupdate t1 set f2=count(*) from t2 where t1.f1=2 and t2.f1=t1.f1 ;\nERROR: ExecutePlan: (junk) `ctid' is NULL!\n\nI would have expected no update to occur since no rows match.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Fri, 30 Jun 2000 21:57:46 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Minor buglet in update...from (I think)" }, { "msg_contents": "I can confirm this is still a bug.\n\n\n> \n> A minor nasty error I got when trying to improve the query used to disable\n> triggers:\n> \n> create table t1(f1 int4, f2 int4);\n> create table t2(f1 int4, f2 int4);\n> \n> insert into t1 values(1, 0);\n> insert into t1 values(2, 0);\n> \n> insert into t2 values(1, 0);\n> \n> update t1 set f2=count(*) from t2 where t1.f1=1 and t2.f1=t1.f1 ;\n> UPDATE 1\n> \n> update t1 set f2=count(*) from t2 where t1.f1=2 and t2.f1=t1.f1 ;\n> ERROR: ExecutePlan: (junk) `ctid' is NULL!\n> \n> I would have expected no update to occur since no rows match.\n> \n> \n> ----------------------------------------------------------------\n> Philip Warner | __---_____\n> Albatross Consulting Pty. Ltd. |----/ - \\\n> (A.C.N. 008 659 498) | /(@) ______---_\n> Tel: (+61) 0500 83 82 81 | _________ \\\n> Fax: (+61) 0500 83 82 82 | ___________ |\n> Http://www.rhyme.com.au | / \\|\n> | --________--\n> PGP key available upon request, | /\n> and from pgp5.ai.mit.edu:11371 |/\n> \n\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 10 Oct 2000 21:04:04 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Minor buglet in update...from (I think)" }, { "msg_contents": "\nThis still throws a strange error. Someone want to fix it?\n\n> \n> A minor nasty error I got when trying to improve the query used to disable\n> triggers:\n> \n> create table t1(f1 int4, f2 int4);\n> create table t2(f1 int4, f2 int4);\n> \n> insert into t1 values(1, 0);\n> insert into t1 values(2, 0);\n> \n> insert into t2 values(1, 0);\n> \n> update t1 set f2=count(*) from t2 where t1.f1=1 and t2.f1=t1.f1 ;\n> UPDATE 1\n> \n> update t1 set f2=count(*) from t2 where t1.f1=2 and t2.f1=t1.f1 ;\n> ERROR: ExecutePlan: (junk) `ctid' is NULL!\n> \n> I would have expected no update to occur since no rows match.\n> \n> \n> ----------------------------------------------------------------\n> Philip Warner | __---_____\n> Albatross Consulting Pty. Ltd. |----/ - \\\n> (A.C.N. 008 659 498) | /(@) ______---_\n> Tel: (+61) 0500 83 82 81 | _________ \\\n> Fax: (+61) 0500 83 82 82 | ___________ |\n> Http://www.rhyme.com.au | / \\|\n> | --________--\n> PGP key available upon request, | /\n> and from pgp5.ai.mit.edu:11371 |/\n> \n\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 24 Jan 2001 08:47:20 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Minor buglet in update...from (I think)" }, { "msg_contents": "\nCan anyone explain this failure? It still exists in CVS.\n\n---------------------------------------------------------------------------\n\n> \n> A minor nasty error I got when trying to improve the query used to disable\n> triggers:\n> \n> create table t1(f1 int4, f2 int4);\n> create table t2(f1 int4, f2 int4);\n> \n> insert into t1 values(1, 0);\n> insert into t1 values(2, 0);\n> \n> insert into t2 values(1, 0);\n> \n> update t1 set f2=count(*) from t2 where t1.f1=1 and t2.f1=t1.f1 ;\n> UPDATE 1\n> \n> update t1 set f2=count(*) from t2 where t1.f1=2 and t2.f1=t1.f1 ;\n> ERROR: ExecutePlan: (junk) `ctid' is NULL!\n> \n> I would have expected no update to occur since no rows match.\n> \n> \n> ----------------------------------------------------------------\n> Philip Warner | __---_____\n> Albatross Consulting Pty. Ltd. |----/ - \\\n> (A.C.N. 008 659 498) | /(@) ______---_\n> Tel: (+61) 0500 83 82 81 | _________ \\\n> Fax: (+61) 0500 83 82 82 | ___________ |\n> Http://www.rhyme.com.au | / \\|\n> | --________--\n> PGP key available upon request, | /\n> and from pgp5.ai.mit.edu:11371 |/\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Mon, 26 Nov 2001 15:22:55 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Minor buglet in update...from (I think)" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n> Can anyone explain this failure? It still exists in CVS.\n\n>> update t1 set f2=count(*) from t2 where t1.f1=2 and t2.f1=t1.f1 ;\n>> ERROR: ExecutePlan: (junk) `ctid' is NULL!\n\nAs I recall, discussion about fixing that problem trailed off because\nno one could explain what an aggregate means in UPDATE. My thought\nis we should probably forbid the construct entirely (SQL does).\nSee previous discussion around 7/7/00.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 26 Nov 2001 18:28:17 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Minor buglet in update...from (I think) " }, { "msg_contents": "> Bruce Momjian <[email protected]> writes:\n> > Can anyone explain this failure? It still exists in CVS.\n> \n> >> update t1 set f2=count(*) from t2 where t1.f1=2 and t2.f1=t1.f1 ;\n> >> ERROR: ExecutePlan: (junk) `ctid' is NULL!\n> \n> As I recall, discussion about fixing that problem trailed off because\n> no one could explain what an aggregate means in UPDATE. My thought\n> is we should probably forbid the construct entirely (SQL does).\n> See previous discussion around 7/7/00.\n\nOh, so it is the aggregate. What threw me off is that both parts of the\nWHERE clause are required to cause the failure, so I thought it was\nsomething else.\n\nI don't see a problem with aggregates in UPDATE, except when the updated\nfield is part of the WHERE clause, but even then, transaction semantics\nshould make it matter. I see the mailist thread now.\n\nI will try and get it added to the TODO list so it is documented.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Mon, 26 Nov 2001 19:23:17 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Minor buglet in update...from (I think)" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n> Oh, so it is the aggregate. What threw me off is that both parts of the\n> WHERE clause are required to cause the failure,\n\nNot necessarily; I think it's got more to do with a null aggregate\nresult:\n\nregression=# create table t1 (f1 datetime);\nCREATE\nregression=# create table t2 (f2 datetime);\nCREATE\nregression=# update t2 set f2 = min(f1) from t1;\nERROR: ExecutePlan: (junk) `ctid' is NULL!\nregression=# insert into t1 values ('now');\nINSERT 400577 1\nregression=# update t2 set f2 = min(f1) from t1;\nERROR: ExecutePlan: (junk) `ctid' is NULL!\nregression=# insert into t2 values ('now');\nINSERT 400578 1\nregression=# update t2 set f2 = min(f1) from t1;\nUPDATE 1\nregression=#\n\nHowever the ERROR is only one symptom. The real problem is that the\ncalculation that's being done is useless/nonsensical.\n\n> I don't see a problem with aggregates in UPDATE,\n\nThink harder ... what is the aggregate being taken over, and how do you\nassociate the aggregate's single result row with any particular row in\nthe UPDATE's target table?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 26 Nov 2001 19:28:31 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Minor buglet in update...from (I think) " }, { "msg_contents": "> Bruce Momjian <[email protected]> writes:\n> > Oh, so it is the aggregate. What threw me off is that both parts of the\n> > WHERE clause are required to cause the failure,\n> \n> Not necessarily; I think it's got more to do with a null aggregate\n> result:\n> \n> regression=# create table t1 (f1 datetime);\n> CREATE\n> regression=# create table t2 (f2 datetime);\n> CREATE\n> regression=# update t2 set f2 = min(f1) from t1;\n> ERROR: ExecutePlan: (junk) `ctid' is NULL!\n> regression=# insert into t1 values ('now');\n> INSERT 400577 1\n> regression=# update t2 set f2 = min(f1) from t1;\n> ERROR: ExecutePlan: (junk) `ctid' is NULL!\n> regression=# insert into t2 values ('now');\n> INSERT 400578 1\n> regression=# update t2 set f2 = min(f1) from t1;\n> UPDATE 1\n> regression=#\n> \n> However the ERROR is only one symptom. The real problem is that the\n> calculation that's being done is useless/nonsensical.\n> \n> > I don't see a problem with aggregates in UPDATE,\n> \n> Think harder ... what is the aggregate being taken over, and how do you\n> associate the aggregate's single result row with any particular row in\n> the UPDATE's target table?\n\nI thought the aggregate would be generated on all rows in the table in\nthe pre-transaction version of the table, so in this example:\n\n\tregression=# update t2 set f2 = min(f1) from t1;\n\nIt places the minimum value of t1.f1 in all t2.f2 rows. Is there\nanother way to look at it?\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Mon, 26 Nov 2001 19:39:06 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Minor buglet in update...from (I think)" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n> I thought the aggregate would be generated on all rows in the table in\n> the pre-transaction version of the table, so in this example:\n> \tregression=# update t2 set f2 = min(f1) from t1;\n> It places the minimum value of t1.f1 in all t2.f2 rows.\n\nThis actually is not the most interesting example, because the aggregate\ndoesn't depend at all on t2. Try this instead:\n\nregression=# create table t1(f1 int);\nCREATE\nregression=# create table t2(f1 int);\nCREATE\nregression=# insert into t1 values(-1);\nINSERT 400599 1\nregression=# insert into t1 values(-2);\nINSERT 400600 1\nregression=# insert into t1 values(-3);\nINSERT 400601 1\nregression=# insert into t2 values(-1);\nINSERT 400602 1\nregression=# insert into t2 values(-2);\nINSERT 400603 1\nregression=# insert into t2 values(-3);\nINSERT 400604 1\nregression=# update t2 set f1 = count(*) from t1;\nUPDATE 1\nregression=# select * from t2;\n f1\n----\n -2\n -3\n 9\n(3 rows)\n\nregression=#\n\nThis is certainly broken, but what's the correct behavior?\nOr how about this, which doesn't even use an aggregate:\n\nregression=# update t2 set f1 = t1.f1 from t1;\nUPDATE 3\nregression=# select * from t2;\n f1\n----\n -1\n -1\n -1\n(3 rows)\n\nregression=#\n\nThat's surprising too, perhaps, but what would you have expected\nand why?\n\nThere's a reason why SQL99 forbids joins and aggregates in UPDATE ...\nthey're not always well-defined.\n\nI had a proposal (GROUP BY ctid) in the older thread for fixing the\naggregate misbehavior, but it doesn't solve the more general problem\nof a join that produces multiple matches for the same target row.\nSeems like that probably ought to draw an error.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 26 Nov 2001 19:51:13 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Minor buglet in update...from (I think) " }, { "msg_contents": "> Bruce Momjian <[email protected]> writes:\n> > I thought the aggregate would be generated on all rows in the table in\n> > the pre-transaction version of the table, so in this example:\n> > \tregression=# update t2 set f2 = min(f1) from t1;\n> > It places the minimum value of t1.f1 in all t2.f2 rows.\n> \n> This actually is not the most interesting example, because the aggregate\n> doesn't depend at all on t2. Try this instead:\n> \n> regression=# create table t1(f1 int);\n> CREATE\n> regression=# create table t2(f1 int);\n> CREATE\n> regression=# insert into t1 values(-1);\n> INSERT 400599 1\n> regression=# insert into t1 values(-2);\n> INSERT 400600 1\n> regression=# insert into t1 values(-3);\n> INSERT 400601 1\n> regression=# insert into t2 values(-1);\n> INSERT 400602 1\n> regression=# insert into t2 values(-2);\n> INSERT 400603 1\n> regression=# insert into t2 values(-3);\n> INSERT 400604 1\n> regression=# update t2 set f1 = count(*) from t1;\n> UPDATE 1\n> regression=# select * from t2;\n> f1\n> ----\n> -2\n> -3\n> 9\n> (3 rows)\n> \n> regression=#\n> \n> This is certainly broken, but what's the correct behavior?\n\nShouldn't it be 9 because there is no join of t1 and t2?\nI can also see 3 as a valid answer.\n\n> Or how about this, which doesn't even use an aggregate:\n> \n> regression=# update t2 set f1 = t1.f1 from t1;\n> UPDATE 3\n> regression=# select * from t2;\n> f1\n> ----\n> -1\n> -1\n> -1\n> (3 rows)\n> \n> regression=#\n> \n> That's surprising too, perhaps, but what would you have expected\n> and why?\n\nSo it grabs the first match. Seems reasonable because t1 returns more\nthan one row.\n\n> \n> There's a reason why SQL99 forbids joins and aggregates in UPDATE ...\n> they're not always well-defined.\n\nYes, I see that now.\n\n> I had a proposal (GROUP BY ctid) in the older thread for fixing the\n> aggregate misbehavior, but it doesn't solve the more general problem\n> of a join that produces multiple matches for the same target row.\n> Seems like that probably ought to draw an error.\n\nOr a NOTICE stating a random row was chosen.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Mon, 26 Nov 2001 19:58:06 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Minor buglet in update...from (I think)" }, { "msg_contents": "\nThread added to TODO.detail.\n\n---------------------------------------------------------------------------\n\n> Bruce Momjian <[email protected]> writes:\n> > I thought the aggregate would be generated on all rows in the table in\n> > the pre-transaction version of the table, so in this example:\n> > \tregression=# update t2 set f2 = min(f1) from t1;\n> > It places the minimum value of t1.f1 in all t2.f2 rows.\n> \n> This actually is not the most interesting example, because the aggregate\n> doesn't depend at all on t2. Try this instead:\n> \n> regression=# create table t1(f1 int);\n> CREATE\n> regression=# create table t2(f1 int);\n> CREATE\n> regression=# insert into t1 values(-1);\n> INSERT 400599 1\n> regression=# insert into t1 values(-2);\n> INSERT 400600 1\n> regression=# insert into t1 values(-3);\n> INSERT 400601 1\n> regression=# insert into t2 values(-1);\n> INSERT 400602 1\n> regression=# insert into t2 values(-2);\n> INSERT 400603 1\n> regression=# insert into t2 values(-3);\n> INSERT 400604 1\n> regression=# update t2 set f1 = count(*) from t1;\n> UPDATE 1\n> regression=# select * from t2;\n> f1\n> ----\n> -2\n> -3\n> 9\n> (3 rows)\n> \n> regression=#\n> \n> This is certainly broken, but what's the correct behavior?\n> Or how about this, which doesn't even use an aggregate:\n> \n> regression=# update t2 set f1 = t1.f1 from t1;\n> UPDATE 3\n> regression=# select * from t2;\n> f1\n> ----\n> -1\n> -1\n> -1\n> (3 rows)\n> \n> regression=#\n> \n> That's surprising too, perhaps, but what would you have expected\n> and why?\n> \n> There's a reason why SQL99 forbids joins and aggregates in UPDATE ...\n> they're not always well-defined.\n> \n> I had a proposal (GROUP BY ctid) in the older thread for fixing the\n> aggregate misbehavior, but it doesn't solve the more general problem\n> of a join that produces multiple matches for the same target row.\n> Seems like that probably ought to draw an error.\n> \n> \t\t\tregards, tom lane\n> \n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 5 Dec 2001 16:04:32 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Minor buglet in update...from (I think)" } ]
[ { "msg_contents": "Hi!\n\nAttached is the tableoid patch. If there are no objections can I have\nthis applied?\n\nThanks!\n\n-- \nChris Bitmead\nmailto:[email protected]", "msg_date": "Sat, 01 Jul 2000 15:41:22 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": true, "msg_subject": "TABLEOID patch" }, { "msg_contents": "Chris Bitmead <[email protected]> writes:\n> Attached is the tableoid patch. If there are no objections can I have\n> this applied?\n\nI definitely object to the tests of tableoid that you have attached to\nthe inherit regress test. Don't you realize that anything that depends\non another installation matching the exact same OIDs you have is pretty\nmuch guaranteed to fail?\n\n\t\t\tregards, tom lane\n", "msg_date": "Sat, 01 Jul 2000 13:55:56 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PATCHES] TABLEOID patch " }, { "msg_contents": "Attached is a new patch which addresses this problem.\n\n-- \nChris Bitmead\nmailto:[email protected]", "msg_date": "Sun, 02 Jul 2000 19:32:59 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [PATCHES] TABLEOID patch" }, { "msg_contents": "Applied. Thanks.\n\n> \n> Attached is a new patch which addresses this problem.\n> \n> -- \n> Chris Bitmead\n> mailto:[email protected]\n\n> ? config.log\n> ? config.cache\n> ? config.status\n> ? nohup.out\n> ? GNUmakefile\n> ? src/GNUmakefile\n> ? src/Makefile.global\n> ? src/ID\n> ? src/nohup.out\n> ? src/backend/postgres\n> ? src/backend/fmgr.h\n> ? src/backend/parse.h\n> ? src/backend/global1.bki.source\n> ? src/backend/local1_template1.bki.source\n> ? src/backend/global1.description\n> ? src/backend/local1_template1.description\n> ? src/backend/1\n> ? src/backend/catalog/genbki.sh\n> ? src/backend/catalog/global1.bki.source\n> ? src/backend/catalog/global1.description\n> ? src/backend/catalog/local1_template1.bki.source\n> ? src/backend/catalog/local1_template1.description\n> ? src/backend/parser/y.output\n> ? src/backend/parser/y.output.gz\n> ? src/backend/parser/gram.y.works\n> ? src/backend/parser/gram.y.works.try\n> ? src/backend/parser/y.output.noerror\n> ? src/backend/parser/gram.y.gz\n> ? src/backend/port/Makefile\n> ? src/backend/utils/fmgr.h\n> ? src/backend/utils/fmgrstamp-h\n> ? src/bin/initdb/initdb\n> ? src/bin/initlocation/initlocation\n> ? src/bin/ipcclean/ipcclean\n> ? src/bin/pg_ctl/pg_ctl\n> ? src/bin/pg_dump/pg_dump\n> ? src/bin/pg_id/pg_id\n> ? src/bin/pg_passwd/pg_passwd\n> ? src/bin/pg_version/pg_version\n> ? src/bin/pgtclsh/mkMakefile.tkdefs.sh\n> ? src/bin/pgtclsh/mkMakefile.tcldefs.sh\n> ? src/bin/psql/psql\n> ? src/bin/scripts/createlang\n> ? src/include/version.h\n> ? src/include/config.h\n> ? src/include/parser/parse.h\n> ? src/include/utils/fmgroids.h\n> ? src/interfaces/ecpg/lib/libecpg.so.3.1.1\n> ? src/interfaces/ecpg/preproc/ecpg\n> ? src/interfaces/jdbc/postgresql.jar\n> ? src/interfaces/jdbc/example/psql.class\n> ? src/interfaces/jdbc/postgresql/DriverClass.java\n> ? src/interfaces/jdbc/postgresql/DriverClass.class\n> ? src/interfaces/jdbc/postgresql/Connection.class\n> ? src/interfaces/jdbc/postgresql/Field.class\n> ? src/interfaces/jdbc/postgresql/PG_Stream.class\n> ? src/interfaces/jdbc/postgresql/Driver.class\n> ? src/interfaces/jdbc/postgresql/ResultSet.class\n> ? src/interfaces/jdbc/postgresql/fastpath/Fastpath.class\n> ? src/interfaces/jdbc/postgresql/fastpath/FastpathArg.class\n> ? src/interfaces/jdbc/postgresql/geometric/PGbox.class\n> ? src/interfaces/jdbc/postgresql/geometric/PGpoint.class\n> ? src/interfaces/jdbc/postgresql/geometric/PGcircle.class\n> ? src/interfaces/jdbc/postgresql/geometric/PGline.class\n> ? src/interfaces/jdbc/postgresql/geometric/PGlseg.class\n> ? src/interfaces/jdbc/postgresql/geometric/PGpath.class\n> ? src/interfaces/jdbc/postgresql/geometric/PGpolygon.class\n> ? src/interfaces/jdbc/postgresql/jdbc2/ResultSet.class\n> ? src/interfaces/jdbc/postgresql/jdbc2/Connection.class\n> ? src/interfaces/jdbc/postgresql/jdbc2/ResultSetMetaData.class\n> ? src/interfaces/jdbc/postgresql/jdbc2/DatabaseMetaData.class\n> ? src/interfaces/jdbc/postgresql/jdbc2/Statement.class\n> ? src/interfaces/jdbc/postgresql/jdbc2/PreparedStatement.class\n> ? src/interfaces/jdbc/postgresql/jdbc2/CallableStatement.class\n> ? src/interfaces/jdbc/postgresql/largeobject/LargeObjectManager.class\n> ? src/interfaces/jdbc/postgresql/largeobject/LargeObject.class\n> ? src/interfaces/jdbc/postgresql/util/PSQLException.class\n> ? src/interfaces/jdbc/postgresql/util/UnixCrypt.class\n> ? src/interfaces/jdbc/postgresql/util/Serialize.class\n> ? src/interfaces/jdbc/postgresql/util/PGobject.class\n> ? src/interfaces/jdbc/postgresql/util/PGtokenizer.class\n> ? src/interfaces/jdbc/postgresql/util/PGmoney.class\n> ? src/interfaces/libpgeasy/libpgeasy.so.2.1\n> ? src/interfaces/libpgtcl/mkMakefile.tcldefs.sh\n> ? src/interfaces/libpgtcl/mkMakefile.tkdefs.sh\n> ? src/interfaces/libpq/libpq.so.2.1\n> ? src/interfaces/libpq++/libpq++.so.3.1\n> ? src/interfaces/odbc/Makefile.global\n> ? src/pl/plpgsql/src/libplpgsql.so.1.0\n> ? src/test/regress/GNUmakefile\n> ? src/test/regress/regress.out\n> ? src/test/regress/x.x\n> ? src/test/regress/nohup.out\n> ? src/test/regress/regression.diffs\n> ? src/test/regress/expected/copy.out\n> ? src/test/regress/expected/misc.out\n> ? src/test/regress/expected/create_function_1.out\n> ? src/test/regress/expected/create_function_2.out\n> ? src/test/regress/expected/constraints.out\n> ? src/test/regress/results/boolean.out\n> ? src/test/regress/results/char.out\n> ? src/test/regress/results/name.out\n> ? src/test/regress/results/varchar.out\n> ? src/test/regress/results/text.out\n> ? src/test/regress/results/int2.out\n> ? src/test/regress/results/int4.out\n> ? src/test/regress/results/int8.out\n> ? src/test/regress/results/oid.out\n> ? src/test/regress/results/float4.out\n> ? src/test/regress/results/float8.out\n> ? src/test/regress/results/numeric.out\n> ? src/test/regress/results/strings.out\n> ? src/test/regress/results/numerology.out\n> ? src/test/regress/results/point.out\n> ? src/test/regress/results/lseg.out\n> ? src/test/regress/results/box.out\n> ? src/test/regress/results/path.out\n> ? src/test/regress/results/polygon.out\n> ? src/test/regress/results/circle.out\n> ? src/test/regress/results/interval.out\n> ? src/test/regress/results/timestamp.out\n> ? src/test/regress/results/reltime.out\n> ? src/test/regress/results/tinterval.out\n> ? src/test/regress/results/inet.out\n> ? src/test/regress/results/comments.out\n> ? src/test/regress/results/oidjoins.out\n> ? src/test/regress/results/type_sanity.out\n> ? src/test/regress/results/opr_sanity.out\n> ? src/test/regress/results/abstime.out\n> ? src/test/regress/results/geometry.out\n> ? src/test/regress/results/horology.out\n> ? src/test/regress/results/create_function_1.out\n> ? src/test/regress/results/create_type.out\n> ? src/test/regress/results/create_table.out\n> ? src/test/regress/results/create_function_2.out\n> ? src/test/regress/results/copy.out\n> ? src/test/regress/results/onek.data\n> ? src/test/regress/results/constraints.out\n> ? src/test/regress/results/triggers.out\n> ? src/test/regress/results/create_misc.out\n> ? src/test/regress/results/create_aggregate.out\n> ? src/test/regress/results/create_operator.out\n> ? src/test/regress/results/create_index.out\n> ? src/test/regress/results/inherit.out\n> ? src/test/regress/results/create_view.out\n> ? src/test/regress/results/sanity_check.out\n> ? src/test/regress/results/errors.out\n> ? src/test/regress/results/select.out\n> ? src/test/regress/results/select_into.out\n> ? src/test/regress/results/select_distinct.out\n> ? src/test/regress/results/select_distinct_on.out\n> ? src/test/regress/results/select_implicit.out\n> ? src/test/regress/results/select_having.out\n> ? src/test/regress/results/subselect.out\n> ? src/test/regress/results/union.out\n> ? src/test/regress/results/case.out\n> ? src/test/regress/results/join.out\n> ? src/test/regress/results/aggregates.out\n> ? src/test/regress/results/transactions.out\n> ? src/test/regress/results/random.out\n> ? src/test/regress/results/portals.out\n> ? src/test/regress/results/arrays.out\n> ? src/test/regress/results/btree_index.out\n> ? src/test/regress/results/hash_index.out\n> ? src/test/regress/results/misc.out\n> ? src/test/regress/results/select_views.out\n> ? src/test/regress/results/alter_table.out\n> ? src/test/regress/results/portals_p2.out\n> ? src/test/regress/results/rules.out\n> ? src/test/regress/results/foreign_key.out\n> ? src/test/regress/results/limit.out\n> ? src/test/regress/results/plpgsql.out\n> ? src/test/regress/results/temp.out\n> ? src/test/regress/results/stud_emp.data\n> ? src/test/regress/sql/copy.sql\n> ? src/test/regress/sql/misc.sql\n> ? src/test/regress/sql/create_function_1.sql\n> ? src/test/regress/sql/create_function_2.sql\n> ? src/test/regress/sql/constraints.sql\n> Index: doc/src/sgml/inherit.sgml\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/doc/src/sgml/inherit.sgml,v\n> retrieving revision 1.10\n> diff -c -r1.10 inherit.sgml\n> *** doc/src/sgml/inherit.sgml\t2000/06/22 22:31:15\t1.10\n> --- doc/src/sgml/inherit.sgml\t2000/07/02 09:29:43\n> ***************\n> *** 96,101 ****\n> --- 96,152 ----\n> <command>UPDATE</command> and <command>DELETE</command> --\n> support this <quote>ONLY</quote> notation.\n> </para>\n> + \n> + <para>\n> + In some cases you may wish to know which table a particular tuple\n> + originated from. There is a system attribute called\n> + <quote>TABLEOID</quote> in each table which can tell you the\n> + originating table:\n> + \n> + <programlisting>\n> + SELECT c.tableoid, c.name, c.altitude\n> + FROM cities c\n> + WHERE c.altitude > 500;\n> + </programlisting>\n> + \n> + which returns:\n> + \n> + <programlisting>\n> + +---------+----------+----------+\n> + |tableoid |name | altitude |\n> + +---------+----------+----------+\n> + |37292 |Las Vegas | 2174 |\n> + +---------+----------+----------+\n> + |37280 |Mariposa | 1953 |\n> + +---------+----------+----------+\n> + |37280 |Madison | 845 |\n> + +---------+----------+----------+\n> + </programlisting>\n> + \n> + If you do a join with pg_class you can see the actual table name:\n> + \n> + <programlisting>\n> + SELECT p.relname, c.name, c.altitude\n> + FROM cities c, pg_class p\n> + WHERE c.altitude > 500 and c.tableoid = p.oid;\n> + </programlisting>\n> + \n> + which returns:\n> + \n> + <programlisting>\n> + +---------+----------+----------+\n> + |relname |name | altitude |\n> + +---------+----------+----------+\n> + |capitals |Las Vegas | 2174 |\n> + +---------+----------+----------+\n> + |cities |Mariposa | 1953 |\n> + +---------+----------+----------+\n> + |cities |Madison | 845 |\n> + +---------+----------+----------+\n> + </programlisting>\n> + \n> + </para>\n> + \n> <note>\n> <title>Deprecated</title> \n> <para>\n> Index: src/backend/access/common/heaptuple.c\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/backend/access/common/heaptuple.c,v\n> retrieving revision 1.62\n> diff -c -r1.62 heaptuple.c\n> *** src/backend/access/common/heaptuple.c\t2000/04/12 17:14:36\t1.62\n> --- src/backend/access/common/heaptuple.c\t2000/07/02 09:29:46\n> ***************\n> *** 9,15 ****\n> *\n> *\n> * IDENTIFICATION\n> ! *\t $Header: /home/projects/pgsql/cvsroot/pgsql/src/backend/access/common/heaptuple.c,v 1.62 2000/04/12 17:14:36 momjian Exp $\n> *\n> * NOTES\n> *\t The old interface functions have been converted to macros\n> --- 9,15 ----\n> *\n> *\n> * IDENTIFICATION\n> ! *\t $Header: /usr/local/cvsroot/pgsql/src/backend/access/common/heaptuple.c,v 1.62 2000/04/12 17:14:36 momjian Exp $\n> *\n> * NOTES\n> *\t The old interface functions have been converted to macros\n> ***************\n> *** 169,174 ****\n> --- 169,175 ----\n> \telse\n> \t\tswitch (attnum)\n> \t\t{\n> + \t\t\tcase TableOidAttributeNumber:\n> \t\t\tcase SelfItemPointerAttributeNumber:\n> \t\t\tcase ObjectIdAttributeNumber:\n> \t\t\tcase MinTransactionIdAttributeNumber:\n> ***************\n> *** 205,210 ****\n> --- 206,213 ----\n> \n> \tswitch (attno)\n> \t{\n> + \t\tcase TableOidAttributeNumber:\n> + \t\t\treturn sizeof f->t_oid;\n> \t\tcase SelfItemPointerAttributeNumber:\n> \t\t\treturn sizeof f->t_ctid;\n> \t\tcase ObjectIdAttributeNumber:\n> ***************\n> *** 237,242 ****\n> --- 240,248 ----\n> \n> \tswitch (attno)\n> \t{\n> + \t\tcase TableOidAttributeNumber:\n> + \t\t\tbyval = true;\n> + \t\t\tbreak;\n> \t\tcase SelfItemPointerAttributeNumber:\n> \t\t\tbyval = false;\n> \t\t\tbreak;\n> ***************\n> *** 275,281 ****\n> {\n> \tswitch (attnum)\n> \t{\n> ! \t\t\tcase SelfItemPointerAttributeNumber:\n> \t\t\treturn (Datum) &tup->t_ctid;\n> \t\tcase ObjectIdAttributeNumber:\n> \t\t\treturn (Datum) (long) tup->t_oid;\n> --- 281,289 ----\n> {\n> \tswitch (attnum)\n> \t{\n> ! \t\tcase TableOidAttributeNumber:\n> ! \t\t\treturn (Datum) &tup->t_tableoid;\n> ! \t\tcase SelfItemPointerAttributeNumber:\n> \t\t\treturn (Datum) &tup->t_ctid;\n> \t\tcase ObjectIdAttributeNumber:\n> \t\t\treturn (Datum) (long) tup->t_oid;\n> Index: src/backend/access/heap/heapam.c\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/backend/access/heap/heapam.c,v\n> retrieving revision 1.73\n> diff -c -r1.73 heapam.c\n> *** src/backend/access/heap/heapam.c\t2000/06/30 16:10:40\t1.73\n> --- src/backend/access/heap/heapam.c\t2000/07/02 09:29:49\n> ***************\n> *** 235,240 ****\n> --- 235,242 ----\n> \tint\t\t\tlinesleft;\n> \tItemPointer tid = (tuple->t_data == NULL) ?\n> \t(ItemPointer) NULL : &(tuple->t_self);\n> + \n> + \ttuple->tableOid = relation->rd_id;\n> \n> \t/* ----------------\n> \t *\tincrement access statistics\n> ***************\n> *** 621,626 ****\n> --- 623,629 ----\n> \n> \tAssert(lockmode >= NoLock && lockmode < MAX_LOCKMODES);\n> \n> + \n> \t/* ----------------\n> \t *\tincrement access statistics\n> \t * ----------------\n> ***************\n> *** 1084,1089 ****\n> --- 1087,1093 ----\n> \tItemPointer tid = &(tuple->t_self);\n> \tOffsetNumber offnum;\n> \n> + \ttuple->tableOid = relation->rd_id;\n> \t/* ----------------\n> \t *\tincrement access statistics\n> \t * ----------------\n> ***************\n> *** 1178,1183 ****\n> --- 1182,1188 ----\n> \tbool\t\tinvalidBlock,\n> \t\t\t\tlinkend;\n> \n> + \ttp.tableOid = relation->rd_id;\n> \t/* ----------------\n> \t *\tget the buffer from the relation descriptor\n> \t *\tNote that this does a buffer pin.\n> ***************\n> *** 1270,1275 ****\n> --- 1275,1281 ----\n> \t *\tincrement access statistics\n> \t * ----------------\n> \t */\n> + \ttup->tableOid = relation->rd_id;\n> \tIncrHeapAccessStat(local_insert);\n> \tIncrHeapAccessStat(global_insert);\n> \n> ***************\n> *** 1335,1340 ****\n> --- 1341,1347 ----\n> \tBuffer\t\tbuffer;\n> \tint\t\t\tresult;\n> \n> + \ttp.tableOid = relation->rd_id;\n> \t/* increment access statistics */\n> \tIncrHeapAccessStat(local_delete);\n> \tIncrHeapAccessStat(global_delete);\n> ***************\n> *** 1447,1452 ****\n> --- 1454,1460 ----\n> \tBuffer\t\tbuffer;\n> \tint\t\t\tresult;\n> \n> + \tnewtup->tableOid = relation->rd_id;\n> \t/* increment access statistics */\n> \tIncrHeapAccessStat(local_replace);\n> \tIncrHeapAccessStat(global_replace);\n> ***************\n> *** 1575,1580 ****\n> --- 1583,1589 ----\n> \tPageHeader\tdp;\n> \tint\t\t\tresult;\n> \n> + \ttuple->tableOid = relation->rd_id;\n> \t/* increment access statistics */\n> \tIncrHeapAccessStat(local_mark4update);\n> \tIncrHeapAccessStat(global_mark4update);\n> Index: src/backend/catalog/heap.c\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/backend/catalog/heap.c,v\n> retrieving revision 1.135\n> diff -c -r1.135 heap.c\n> *** src/backend/catalog/heap.c\t2000/07/02 04:46:09\t1.135\n> --- src/backend/catalog/heap.c\t2000/07/02 09:29:55\n> ***************\n> *** 131,142 ****\n> \tMaxCommandIdAttributeNumber, 0, -1, -1, '\\001', 'p', '\\0', 'i', '\\0', '\\0'\n> };\n> \n> ! static Form_pg_attribute HeapAtt[] = {&a1, &a2, &a3, &a4, &a5, &a6};\n> \n> /* ----------------------------------------------------------------\n> *\t\t\t\tXXX END OF UGLY HARD CODED BADNESS XXX\n> ! * ----------------------------------------------------------------\n> ! */\n> \n> \n> /* ----------------------------------------------------------------\n> --- 131,152 ----\n> \tMaxCommandIdAttributeNumber, 0, -1, -1, '\\001', 'p', '\\0', 'i', '\\0', '\\0'\n> };\n> \n> ! /* \n> ! We decide to call this attribute \"tableoid\" rather than say\n> ! \"classoid\" on the basis that in the future there may be more than one\n> ! table of a particular class/type. In any case table is still the word\n> ! used in SQL. \n> ! */\n> ! static FormData_pg_attribute a7 = {\n> ! \t0xffffffff, {\"tableoid\"}, OIDOID, 0, sizeof(Oid),\n> ! \tTableOidAttributeNumber, 0, -1, -1, '\\001', 'p', '\\0', 'i', '\\0', '\\0'\n> ! };\n> \n> + static Form_pg_attribute HeapAtt[] = {&a1, &a2, &a3, &a4, &a5, &a6, &a7};\n> + \n> /* ----------------------------------------------------------------\n> *\t\t\t\tXXX END OF UGLY HARD CODED BADNESS XXX\n> ! * ---------------------------------------------------------------- */\n> \n> \n> /* ----------------------------------------------------------------\n> Index: src/backend/parser/parse_relation.c\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/backend/parser/parse_relation.c,v\n> retrieving revision 1.44\n> diff -c -r1.44 parse_relation.c\n> *** src/backend/parser/parse_relation.c\t2000/06/20 01:41:21\t1.44\n> --- src/backend/parser/parse_relation.c\t2000/07/02 09:29:56\n> ***************\n> *** 57,62 ****\n> --- 57,65 ----\n> \t{\n> \t\t\"cmax\", MaxCommandIdAttributeNumber, CIDOID\n> \t},\n> + \t{\n> + \t\t\"tableoid\", TableOidAttributeNumber, OIDOID\n> + \t}\n> };\n> \n> #define SPECIALS ((int) (sizeof(special_attr)/sizeof(special_attr[0])))\n> Index: src/backend/utils/cache/lsyscache.c\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/backend/utils/cache/lsyscache.c,v\n> retrieving revision 1.42\n> diff -c -r1.42 lsyscache.c\n> *** src/backend/utils/cache/lsyscache.c\t2000/06/08 22:37:30\t1.42\n> --- src/backend/utils/cache/lsyscache.c\t2000/07/02 09:29:58\n> ***************\n> *** 249,254 ****\n> --- 249,256 ----\n> \tif (attnum == ObjectIdAttributeNumber ||\n> \t\tattnum == SelfItemPointerAttributeNumber)\n> \t\treturn 1.0 / (double) ntuples;\n> + \tif (attnum == TableOidAttributeNumber)\n> + \t\treturn 1.0;\n> \n> \t/*\n> \t * VACUUM ANALYZE has not been run for this table. Produce an estimate\n> Index: src/include/access/heapam.h\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/include/access/heapam.h,v\n> retrieving revision 1.54\n> diff -c -r1.54 heapam.h\n> *** src/include/access/heapam.h\t2000/06/30 16:10:49\t1.54\n> --- src/include/access/heapam.h\t2000/07/02 09:30:00\n> ***************\n> *** 165,200 ****\n> *\n> * ----------------\n> */\n> ! #define heap_getattr(tup, attnum, tupleDesc, isnull)\t\t\t\t\t\t\\\n> ! (\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \tAssertMacro((tup) != NULL &&\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t(attnum) > FirstLowInvalidHeapAttributeNumber &&\t\t\t\t\t\\\n> ! \t\t(attnum) != 0),\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t((attnum) > (int) (tup)->t_data->t_natts) ?\t\t\t\t\t\t\t\t\\\n> ! \t(\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t((isnull) ? (*(isnull) = true) : (dummyret)NULL),\t\t\t\t\t\\\n> ! \t\t(Datum)NULL\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t)\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t:\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t(\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t((attnum) > 0) ?\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t(\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t\tfastgetattr((tup), (attnum), (tupleDesc), (isnull))\t\t\t\t\\\n> ! \t\t)\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t:\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t(\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t\t((isnull) ? (*(isnull) = false) : (dummyret)NULL),\t\t\t\t\\\n> ! \t\t\t((attnum) == SelfItemPointerAttributeNumber) ?\t\t\t\t\t\\\n> ! \t\t\t(\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t\t\t(Datum)((char *)&((tup)->t_self))\t\t\t\t\t\t\t\\\n> ! \t\t\t)\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t\t:\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t\t(\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t\t\t(Datum)*(unsigned int *)\t\t\t\t\t\t\t\t\t\\\n> ! \t\t\t\t\t((char *)(tup)->t_data + heap_sysoffset[-(attnum)-1])\t\\\n> ! \t\t\t)\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t\t)\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> ! \t)\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\\\n> )\n> \n> extern HeapAccessStatistics heap_access_stats;\t/* in stats.c */\n> --- 165,205 ----\n> *\n> * ----------------\n> */\n> ! #define heap_getattr(tup, attnum, tupleDesc, isnull) \\\n> ! ( \\\n> ! \tAssertMacro((tup) != NULL && \\\n> ! \t\t(attnum) > FirstLowInvalidHeapAttributeNumber && \\\n> ! \t\t(attnum) != 0), \\\n> ! \t((attnum) > (int) (tup)->t_data->t_natts) ? \\\n> ! \t( \\\n> ! \t\t((isnull) ? (*(isnull) = true) : (dummyret)NULL), \\\n> ! \t\t(Datum)NULL \\\n> ! \t) \\\n> ! \t: \\\n> ! \t( \\\n> ! \t\t((attnum) > 0) ? \\\n> ! \t\t( \\\n> ! \t\t\tfastgetattr((tup), (attnum), (tupleDesc), (isnull)) \\\n> ! \t\t) \\\n> ! \t\t: \\\n> ! \t\t( \\\n> ! \t\t\t((isnull) ? (*(isnull) = false) : (dummyret)NULL), \\\n> ! \t\t\t((attnum) == SelfItemPointerAttributeNumber) ? \\\n> ! \t\t\t( \\\n> ! \t\t\t\t(Datum)((char *)&((tup)->t_self)) \\\n> ! \t\t\t) \\\n> ! \t\t\t: \\\n> ! \t\t\t(((attnum) == TableOidAttributeNumber) ? \\\n> ! \t\t\t( \\\n> ! \t\t\t\t(Datum)((tup)->tableOid) \\\n> ! \t\t\t) \\\n> ! : \\\n> ! \t\t\t( \\\n> ! \t\t\t\t(Datum)*(unsigned int *) \\\n> ! \t\t\t\t\t((char *)(tup)->t_data + heap_sysoffset[-(attnum)-1]) \\\n> ! \t\t\t)) \\\n> ! \t\t) \\\n> ! \t) \\\n> )\n> \n> extern HeapAccessStatistics heap_access_stats;\t/* in stats.c */\n> Index: src/include/access/htup.h\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/include/access/htup.h,v\n> retrieving revision 1.30\n> diff -c -r1.30 htup.h\n> *** src/include/access/htup.h\t2000/06/02 10:20:26\t1.30\n> --- src/include/access/htup.h\t2000/07/02 09:30:01\n> ***************\n> *** 133,139 ****\n> #define MinCommandIdAttributeNumber\t\t\t\t(-4)\n> #define MaxTransactionIdAttributeNumber\t\t\t(-5)\n> #define MaxCommandIdAttributeNumber\t\t\t\t(-6)\n> ! #define FirstLowInvalidHeapAttributeNumber\t\t(-7)\n> \n> /* If you make any changes above, the order off offsets in this must change */\n> extern long heap_sysoffset[];\n> --- 133,140 ----\n> #define MinCommandIdAttributeNumber\t\t\t\t(-4)\n> #define MaxTransactionIdAttributeNumber\t\t\t(-5)\n> #define MaxCommandIdAttributeNumber\t\t\t\t(-6)\n> ! #define TableOidAttributeNumber\t\t\t (-7)\n> ! #define FirstLowInvalidHeapAttributeNumber\t\t(-8)\n> \n> /* If you make any changes above, the order off offsets in this must change */\n> extern long heap_sysoffset[];\n> ***************\n> *** 156,161 ****\n> --- 157,163 ----\n> {\n> \tuint32\t\tt_len;\t\t\t/* length of *t_data */\n> \tItemPointerData t_self;\t\t/* SelfItemPointer */\n> + \tOid tableOid; /* */\n> \tMemoryContext t_datamcxt;\t/* */\n> \tHeapTupleHeader t_data;\t\t/* */\n> } HeapTupleData;\n> Index: src/include/catalog/pg_attribute.h\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/include/catalog/pg_attribute.h,v\n> retrieving revision 1.59\n> diff -c -r1.59 pg_attribute.h\n> *** src/include/catalog/pg_attribute.h\t2000/06/12 03:40:52\t1.59\n> --- src/include/catalog/pg_attribute.h\t2000/07/02 09:30:03\n> ***************\n> *** 267,272 ****\n> --- 267,273 ----\n> DATA(insert OID = 0 ( 1247 cmin\t\t\t\t29 0 4 -4 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1247 xmax\t\t\t\t28 0 4 -5 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1247 cmax\t\t\t\t29 0 4 -6 0 -1 -1 t p f i f f));\n> + DATA(insert OID = 0 ( 1247 tableoid\t\t\t26 0 4 -7 0 -1 -1 t p f i f f));\n> \n> /* ----------------\n> *\t\tpg_database\n> ***************\n> *** 282,287 ****\n> --- 283,289 ----\n> DATA(insert OID = 0 ( 1262 cmin\t\t\t\t29 0 4 -4 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1262 xmax\t\t\t\t28 0 4 -5 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1262 cmax\t\t\t\t29 0 4 -6 0 -1 -1 t p f i f f));\n> + DATA(insert OID = 0 ( 1262 tableoid\t\t\t26 0 4 -7 0 -1 -1 t p f i f f));\n> \n> /* ----------------\n> *\t\tpg_proc\n> ***************\n> *** 329,334 ****\n> --- 331,337 ----\n> DATA(insert OID = 0 ( 1255 cmin\t\t\t\t29 0 4 -4 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1255 xmax\t\t\t\t28 0 4 -5 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1255 cmax\t\t\t\t29 0 4 -6 0 -1 -1 t p f i f f));\n> + DATA(insert OID = 0 ( 1255 tableoid\t\t\t26 0 4 -7 0 -1 -1 t p f i f f));\n> \n> /* ----------------\n> *\t\tpg_shadow\n> ***************\n> *** 348,353 ****\n> --- 351,357 ----\n> DATA(insert OID = 0 ( 1260 cmin\t\t\t\t29 0 4 -4 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1260 xmax\t\t\t\t28 0 4 -5 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1260 cmax\t\t\t\t29 0 4 -6 0 -1 -1 t p f i f f));\n> + DATA(insert OID = 0 ( 1260 tableoid\t\t\t26 0 4 -7 0 -1 -1 t p f i f f));\n> \n> /* ----------------\n> *\t\tpg_group\n> ***************\n> *** 362,367 ****\n> --- 366,372 ----\n> DATA(insert OID = 0 ( 1261 cmin\t\t\t\t29 0 4 -4 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1261 xmax\t\t\t\t28 0 4 -5 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1261 cmax\t\t\t\t29 0 4 -6 0 -1 -1 t p f i f f));\n> + DATA(insert OID = 0 ( 1261 tableoid\t\t\t26 0 4 -7 0 -1 -1 t p f i f f));\n> \n> /* ----------------\n> *\t\tpg_attribute\n> ***************\n> *** 405,410 ****\n> --- 410,416 ----\n> DATA(insert OID = 0 ( 1249 cmin\t\t\t\t29 0 4 -4 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1249 xmax\t\t\t\t28 0 4 -5 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1249 cmax\t\t\t\t29 0 4 -6 0 -1 -1 t p f i f f));\n> + DATA(insert OID = 0 ( 1249 tableoid\t\t\t26 0 4 -7 0 -1 -1 t p f i f f));\n> \n> /* ----------------\n> *\t\tpg_class\n> ***************\n> *** 458,463 ****\n> --- 464,470 ----\n> DATA(insert OID = 0 ( 1259 cmin\t\t\t\t29 0 4 -4 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1259 xmax\t\t\t\t28 0 4 -5 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1259 cmax\t\t\t\t29 0 4 -6 0 -1 -1 t p f i f f));\n> + DATA(insert OID = 0 ( 1259 tableoid\t\t\t26 0 4 -7 0 -1 -1 t p f i f f));\n> \n> /* ----------------\n> *\t\tpg_attrdef\n> ***************\n> *** 473,478 ****\n> --- 480,486 ----\n> DATA(insert OID = 0 ( 1215 cmin\t\t\t\t29 0 4 -4 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1215 xmax\t\t\t\t28 0 4 -5 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1215 cmax\t\t\t\t29 0 4 -6 0 -1 -1 t p f i f f));\n> + DATA(insert OID = 0 ( 1215 tableoid\t\t\t26 0 4 -7 0 -1 -1 t p f i f f));\n> \n> /* ----------------\n> *\t\tpg_relcheck\n> ***************\n> *** 488,493 ****\n> --- 496,502 ----\n> DATA(insert OID = 0 ( 1216 cmin\t\t\t\t29 0 4 -4 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1216 xmax\t\t\t\t28 0 4 -5 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1216 cmax\t\t\t\t29 0 4 -6 0 -1 -1 t p f i f f));\n> + DATA(insert OID = 0 ( 1216 tableoid\t\t\t26 0 4 -7 0 -1 -1 t p f i f f));\n> \n> /* ----------------\n> *\t\tpg_trigger\n> ***************\n> *** 513,518 ****\n> --- 522,528 ----\n> DATA(insert OID = 0 ( 1219 cmin\t\t\t\t29 0 4 -4 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1219 xmax\t\t\t\t28 0 4 -5 0 -1 -1 t p f i f f));\n> DATA(insert OID = 0 ( 1219 cmax\t\t\t\t29 0 4 -6 0 -1 -1 t p f i f f));\n> + DATA(insert OID = 0 ( 1219 tableoid\t\t\t26 0 4 -7 0 -1 -1 t p f i f f));\n> \n> /* ----------------\n> *\t\tpg_variable - this relation is modified by special purpose access\n> Index: src/test/regress/expected/inherit.out\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/test/regress/expected/inherit.out,v\n> retrieving revision 1.1\n> diff -c -r1.1 inherit.out\n> *** src/test/regress/expected/inherit.out\t2000/06/09 11:12:37\t1.1\n> --- src/test/regress/expected/inherit.out\t2000/07/02 09:30:08\n> ***************\n> *** 29,150 ****\n> INSERT INTO d(aa) VALUES('dddddd');\n> INSERT INTO d(aa) VALUES('ddddddd');\n> INSERT INTO d(aa) VALUES('dddddddd');\n> ! SELECT * FROM a;\n> ! aa \n> ! ----------\n> ! aaa\n> ! aaaa\n> ! aaaaa\n> ! aaaaaa\n> ! aaaaaaa\n> ! aaaaaaaa\n> ! bbb\n> ! bbbb\n> ! bbbbb\n> ! bbbbbb\n> ! bbbbbbb\n> ! bbbbbbbb\n> ! ccc\n> ! cccc\n> ! ccccc\n> ! cccccc\n> ! ccccccc\n> ! cccccccc\n> ! ddd\n> ! dddd\n> ! ddddd\n> ! dddddd\n> ! ddddddd\n> ! dddddddd\n> (24 rows)\n> \n> ! SELECT * FROM b;\n> ! aa | bb \n> ! ----------+----\n> ! bbb | \n> ! bbbb | \n> ! bbbbb | \n> ! bbbbbb | \n> ! bbbbbbb | \n> ! bbbbbbbb | \n> ! ddd | \n> ! dddd | \n> ! ddddd | \n> ! dddddd | \n> ! ddddddd | \n> ! dddddddd | \n> (12 rows)\n> \n> ! SELECT * FROM c;\n> ! aa | cc \n> ! ----------+----\n> ! ccc | \n> ! cccc | \n> ! ccccc | \n> ! cccccc | \n> ! ccccccc | \n> ! cccccccc | \n> ! ddd | \n> ! dddd | \n> ! ddddd | \n> ! dddddd | \n> ! ddddddd | \n> ! dddddddd | \n> (12 rows)\n> \n> ! SELECT * FROM d;\n> ! aa | bb | cc | dd \n> ! ----------+----+----+----\n> ! ddd | | | \n> ! dddd | | | \n> ! ddddd | | | \n> ! dddddd | | | \n> ! ddddddd | | | \n> ! dddddddd | | | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY a;\n> ! aa \n> ! ----------\n> ! aaa\n> ! aaaa\n> ! aaaaa\n> ! aaaaaa\n> ! aaaaaaa\n> ! aaaaaaaa\n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY b;\n> ! aa | bb \n> ! ----------+----\n> ! bbb | \n> ! bbbb | \n> ! bbbbb | \n> ! bbbbbb | \n> ! bbbbbbb | \n> ! bbbbbbbb | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY c;\n> ! aa | cc \n> ! ----------+----\n> ! ccc | \n> ! cccc | \n> ! ccccc | \n> ! cccccc | \n> ! ccccccc | \n> ! cccccccc | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY d;\n> ! aa | bb | cc | dd \n> ! ----------+----+----+----\n> ! ddd | | | \n> ! dddd | | | \n> ! ddddd | | | \n> ! dddddd | | | \n> ! ddddddd | | | \n> ! dddddddd | | | \n> (6 rows)\n> \n> UPDATE a SET aa='zzzz' WHERE aa='aaaa';\n> --- 29,150 ----\n> INSERT INTO d(aa) VALUES('dddddd');\n> INSERT INTO d(aa) VALUES('ddddddd');\n> INSERT INTO d(aa) VALUES('dddddddd');\n> ! SELECT relname, a.* FROM a, pg_class where a.tableoid = pg_class.oid;\n> ! relname | aa \n> ! ---------+----------\n> ! a | aaa\n> ! a | aaaa\n> ! a | aaaaa\n> ! a | aaaaaa\n> ! a | aaaaaaa\n> ! a | aaaaaaaa\n> ! b | bbb\n> ! b | bbbb\n> ! b | bbbbb\n> ! b | bbbbbb\n> ! b | bbbbbbb\n> ! b | bbbbbbbb\n> ! c | ccc\n> ! c | cccc\n> ! c | ccccc\n> ! c | cccccc\n> ! c | ccccccc\n> ! c | cccccccc\n> ! d | ddd\n> ! d | dddd\n> ! d | ddddd\n> ! d | dddddd\n> ! d | ddddddd\n> ! d | dddddddd\n> (24 rows)\n> \n> ! SELECT relname, b.* FROM b, pg_class where b.tableoid = pg_class.oid;\n> ! relname | aa | bb \n> ! ---------+----------+----\n> ! b | bbb | \n> ! b | bbbb | \n> ! b | bbbbb | \n> ! b | bbbbbb | \n> ! b | bbbbbbb | \n> ! b | bbbbbbbb | \n> ! d | ddd | \n> ! d | dddd | \n> ! d | ddddd | \n> ! d | dddddd | \n> ! d | ddddddd | \n> ! d | dddddddd | \n> (12 rows)\n> \n> ! SELECT relname, c.* FROM c, pg_class where c.tableoid = pg_class.oid;\n> ! relname | aa | cc \n> ! ---------+----------+----\n> ! c | ccc | \n> ! c | cccc | \n> ! c | ccccc | \n> ! c | cccccc | \n> ! c | ccccccc | \n> ! c | cccccccc | \n> ! d | ddd | \n> ! d | dddd | \n> ! d | ddddd | \n> ! d | dddddd | \n> ! d | ddddddd | \n> ! d | dddddddd | \n> (12 rows)\n> \n> ! SELECT relname, d.* FROM d, pg_class where d.tableoid = pg_class.oid;\n> ! relname | aa | bb | cc | dd \n> ! ---------+----------+----+----+----\n> ! d | ddd | | | \n> ! d | dddd | | | \n> ! d | ddddd | | | \n> ! d | dddddd | | | \n> ! d | ddddddd | | | \n> ! d | dddddddd | | | \n> ! (6 rows)\n> ! \n> ! SELECT relname, a.* FROM ONLY a, pg_class where a.tableoid = pg_class.oid;\n> ! relname | aa \n> ! ---------+----------\n> ! a | aaa\n> ! a | aaaa\n> ! a | aaaaa\n> ! a | aaaaaa\n> ! a | aaaaaaa\n> ! a | aaaaaaaa\n> ! (6 rows)\n> ! \n> ! SELECT relname, b.* FROM ONLY b, pg_class where b.tableoid = pg_class.oid;\n> ! relname | aa | bb \n> ! ---------+----------+----\n> ! b | bbb | \n> ! b | bbbb | \n> ! b | bbbbb | \n> ! b | bbbbbb | \n> ! b | bbbbbbb | \n> ! b | bbbbbbbb | \n> ! (6 rows)\n> ! \n> ! SELECT relname, c.* FROM ONLY c, pg_class where c.tableoid = pg_class.oid;\n> ! relname | aa | cc \n> ! ---------+----------+----\n> ! c | ccc | \n> ! c | cccc | \n> ! c | ccccc | \n> ! c | cccccc | \n> ! c | ccccccc | \n> ! c | cccccccc | \n> ! (6 rows)\n> ! \n> ! SELECT relname, d.* FROM ONLY d, pg_class where d.tableoid = pg_class.oid;\n> ! relname | aa | bb | cc | dd \n> ! ---------+----------+----+----+----\n> ! d | ddd | | | \n> ! d | dddd | | | \n> ! d | ddddd | | | \n> ! d | dddddd | | | \n> ! d | ddddddd | | | \n> ! d | dddddddd | | | \n> (6 rows)\n> \n> UPDATE a SET aa='zzzz' WHERE aa='aaaa';\n> ***************\n> *** 152,534 ****\n> UPDATE b SET aa='zzz' WHERE aa='aaa';\n> UPDATE ONLY b SET aa='zzz' WHERE aa='aaa';\n> UPDATE a SET aa='zzzzzz' WHERE aa LIKE 'aaa%';\n> ! SELECT * FROM a;\n> ! aa \n> ! ----------\n> ! zzzz\n> ! zzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! bbb\n> ! bbbb\n> ! bbbbb\n> ! bbbbbb\n> ! bbbbbbb\n> ! bbbbbbbb\n> ! ccc\n> ! cccc\n> ! ccccc\n> ! cccccc\n> ! ccccccc\n> ! cccccccc\n> ! ddd\n> ! dddd\n> ! ddddd\n> ! dddddd\n> ! ddddddd\n> ! dddddddd\n> (24 rows)\n> \n> ! SELECT * FROM b;\n> ! aa | bb \n> ! ----------+----\n> ! bbb | \n> ! bbbb | \n> ! bbbbb | \n> ! bbbbbb | \n> ! bbbbbbb | \n> ! bbbbbbbb | \n> ! ddd | \n> ! dddd | \n> ! ddddd | \n> ! dddddd | \n> ! ddddddd | \n> ! dddddddd | \n> (12 rows)\n> \n> ! SELECT * FROM c;\n> ! aa | cc \n> ! ----------+----\n> ! ccc | \n> ! cccc | \n> ! ccccc | \n> ! cccccc | \n> ! ccccccc | \n> ! cccccccc | \n> ! ddd | \n> ! dddd | \n> ! ddddd | \n> ! dddddd | \n> ! ddddddd | \n> ! dddddddd | \n> (12 rows)\n> \n> ! SELECT * FROM d;\n> ! aa | bb | cc | dd \n> ! ----------+----+----+----\n> ! ddd | | | \n> ! dddd | | | \n> ! ddddd | | | \n> ! dddddd | | | \n> ! ddddddd | | | \n> ! dddddddd | | | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY a;\n> ! aa \n> ! --------\n> ! zzzz\n> ! zzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY b;\n> ! aa | bb \n> ! ----------+----\n> ! bbb | \n> ! bbbb | \n> ! bbbbb | \n> ! bbbbbb | \n> ! bbbbbbb | \n> ! bbbbbbbb | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY c;\n> ! aa | cc \n> ! ----------+----\n> ! ccc | \n> ! cccc | \n> ! ccccc | \n> ! cccccc | \n> ! ccccccc | \n> ! cccccccc | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY d;\n> ! aa | bb | cc | dd \n> ! ----------+----+----+----\n> ! ddd | | | \n> ! dddd | | | \n> ! ddddd | | | \n> ! dddddd | | | \n> ! ddddddd | | | \n> ! dddddddd | | | \n> (6 rows)\n> \n> UPDATE b SET aa='new';\n> ! SELECT * FROM a;\n> ! aa \n> ! ----------\n> ! zzzz\n> ! zzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! ccc\n> ! cccc\n> ! ccccc\n> ! cccccc\n> ! ccccccc\n> ! cccccccc\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> (24 rows)\n> \n> ! SELECT * FROM b;\n> ! aa | bb \n> ! -----+----\n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> (12 rows)\n> \n> ! SELECT * FROM c;\n> ! aa | cc \n> ! ----------+----\n> ! ccc | \n> ! cccc | \n> ! ccccc | \n> ! cccccc | \n> ! ccccccc | \n> ! cccccccc | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> (12 rows)\n> \n> ! SELECT * FROM d;\n> ! aa | bb | cc | dd \n> ! -----+----+----+----\n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY a;\n> ! aa \n> ! --------\n> ! zzzz\n> ! zzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! zzzzzz\n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY b;\n> ! aa | bb \n> ! -----+----\n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY c;\n> ! aa | cc \n> ! ----------+----\n> ! ccc | \n> ! cccc | \n> ! ccccc | \n> ! cccccc | \n> ! ccccccc | \n> ! cccccccc | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY d;\n> ! aa | bb | cc | dd \n> ! -----+----+----+----\n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> (6 rows)\n> \n> UPDATE a SET aa='new';\n> DELETE FROM ONLY c WHERE aa='new';\n> ! SELECT * FROM a;\n> ! aa \n> ! -----\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> (18 rows)\n> \n> ! SELECT * FROM b;\n> ! aa | bb \n> ! -----+----\n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> (12 rows)\n> \n> ! SELECT * FROM c;\n> ! aa | cc \n> ! -----+----\n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM d;\n> ! aa | bb | cc | dd \n> ! -----+----+----+----\n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY a;\n> ! aa \n> ! -----\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! new\n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY b;\n> ! aa | bb \n> ! -----+----\n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! new | \n> ! (6 rows)\n> ! \n> ! SELECT * FROM ONLY c;\n> ! aa | cc \n> ! ----+----\n> (0 rows)\n> \n> ! SELECT * FROM ONLY d;\n> ! aa | bb | cc | dd \n> ! -----+----+----+----\n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> ! new | | | \n> (6 rows)\n> \n> DELETE FROM a;\n> ! SELECT * FROM a;\n> ! aa \n> ! ----\n> (0 rows)\n> \n> ! SELECT * FROM b;\n> ! aa | bb \n> ! ----+----\n> (0 rows)\n> \n> ! SELECT * FROM c;\n> ! aa | cc \n> ! ----+----\n> (0 rows)\n> \n> ! SELECT * FROM d;\n> ! aa | bb | cc | dd \n> ! ----+----+----+----\n> (0 rows)\n> \n> ! SELECT * FROM ONLY a;\n> ! aa \n> ! ----\n> (0 rows)\n> \n> ! SELECT * FROM ONLY b;\n> ! aa | bb \n> ! ----+----\n> (0 rows)\n> \n> ! SELECT * FROM ONLY c;\n> ! aa | cc \n> ! ----+----\n> (0 rows)\n> \n> ! SELECT * FROM ONLY d;\n> ! aa | bb | cc | dd \n> ! ----+----+----+----\n> (0 rows)\n> \n> --- 152,534 ----\n> UPDATE b SET aa='zzz' WHERE aa='aaa';\n> UPDATE ONLY b SET aa='zzz' WHERE aa='aaa';\n> UPDATE a SET aa='zzzzzz' WHERE aa LIKE 'aaa%';\n> ! SELECT relname, a.* FROM a, pg_class where a.tableoid = pg_class.oid;\n> ! relname | aa \n> ! ---------+----------\n> ! a | zzzz\n> ! a | zzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! b | bbb\n> ! b | bbbb\n> ! b | bbbbb\n> ! b | bbbbbb\n> ! b | bbbbbbb\n> ! b | bbbbbbbb\n> ! c | ccc\n> ! c | cccc\n> ! c | ccccc\n> ! c | cccccc\n> ! c | ccccccc\n> ! c | cccccccc\n> ! d | ddd\n> ! d | dddd\n> ! d | ddddd\n> ! d | dddddd\n> ! d | ddddddd\n> ! d | dddddddd\n> (24 rows)\n> \n> ! SELECT relname, b.* FROM b, pg_class where b.tableoid = pg_class.oid;\n> ! relname | aa | bb \n> ! ---------+----------+----\n> ! b | bbb | \n> ! b | bbbb | \n> ! b | bbbbb | \n> ! b | bbbbbb | \n> ! b | bbbbbbb | \n> ! b | bbbbbbbb | \n> ! d | ddd | \n> ! d | dddd | \n> ! d | ddddd | \n> ! d | dddddd | \n> ! d | ddddddd | \n> ! d | dddddddd | \n> (12 rows)\n> \n> ! SELECT relname, c.* FROM c, pg_class where c.tableoid = pg_class.oid;\n> ! relname | aa | cc \n> ! ---------+----------+----\n> ! c | ccc | \n> ! c | cccc | \n> ! c | ccccc | \n> ! c | cccccc | \n> ! c | ccccccc | \n> ! c | cccccccc | \n> ! d | ddd | \n> ! d | dddd | \n> ! d | ddddd | \n> ! d | dddddd | \n> ! d | ddddddd | \n> ! d | dddddddd | \n> (12 rows)\n> \n> ! SELECT relname, d.* FROM d, pg_class where d.tableoid = pg_class.oid;\n> ! relname | aa | bb | cc | dd \n> ! ---------+----------+----+----+----\n> ! d | ddd | | | \n> ! d | dddd | | | \n> ! d | ddddd | | | \n> ! d | dddddd | | | \n> ! d | ddddddd | | | \n> ! d | dddddddd | | | \n> ! (6 rows)\n> ! \n> ! SELECT relname, a.* FROM ONLY a, pg_class where a.tableoid = pg_class.oid;\n> ! relname | aa \n> ! ---------+--------\n> ! a | zzzz\n> ! a | zzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! (6 rows)\n> ! \n> ! SELECT relname, b.* FROM ONLY b, pg_class where b.tableoid = pg_class.oid;\n> ! relname | aa | bb \n> ! ---------+----------+----\n> ! b | bbb | \n> ! b | bbbb | \n> ! b | bbbbb | \n> ! b | bbbbbb | \n> ! b | bbbbbbb | \n> ! b | bbbbbbbb | \n> ! (6 rows)\n> ! \n> ! SELECT relname, c.* FROM ONLY c, pg_class where c.tableoid = pg_class.oid;\n> ! relname | aa | cc \n> ! ---------+----------+----\n> ! c | ccc | \n> ! c | cccc | \n> ! c | ccccc | \n> ! c | cccccc | \n> ! c | ccccccc | \n> ! c | cccccccc | \n> ! (6 rows)\n> ! \n> ! SELECT relname, d.* FROM ONLY d, pg_class where d.tableoid = pg_class.oid;\n> ! relname | aa | bb | cc | dd \n> ! ---------+----------+----+----+----\n> ! d | ddd | | | \n> ! d | dddd | | | \n> ! d | ddddd | | | \n> ! d | dddddd | | | \n> ! d | ddddddd | | | \n> ! d | dddddddd | | | \n> (6 rows)\n> \n> UPDATE b SET aa='new';\n> ! SELECT relname, a.* FROM a, pg_class where a.tableoid = pg_class.oid;\n> ! relname | aa \n> ! ---------+----------\n> ! a | zzzz\n> ! a | zzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! b | new\n> ! b | new\n> ! b | new\n> ! b | new\n> ! b | new\n> ! b | new\n> ! c | ccc\n> ! c | cccc\n> ! c | ccccc\n> ! c | cccccc\n> ! c | ccccccc\n> ! c | cccccccc\n> ! d | new\n> ! d | new\n> ! d | new\n> ! d | new\n> ! d | new\n> ! d | new\n> (24 rows)\n> \n> ! SELECT relname, b.* FROM b, pg_class where b.tableoid = pg_class.oid;\n> ! relname | aa | bb \n> ! ---------+-----+----\n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> (12 rows)\n> \n> ! SELECT relname, c.* FROM c, pg_class where c.tableoid = pg_class.oid;\n> ! relname | aa | cc \n> ! ---------+----------+----\n> ! c | ccc | \n> ! c | cccc | \n> ! c | ccccc | \n> ! c | cccccc | \n> ! c | ccccccc | \n> ! c | cccccccc | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> (12 rows)\n> \n> ! SELECT relname, d.* FROM d, pg_class where d.tableoid = pg_class.oid;\n> ! relname | aa | bb | cc | dd \n> ! ---------+-----+----+----+----\n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! (6 rows)\n> ! \n> ! SELECT relname, a.* FROM ONLY a, pg_class where a.tableoid = pg_class.oid;\n> ! relname | aa \n> ! ---------+--------\n> ! a | zzzz\n> ! a | zzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! a | zzzzzz\n> ! (6 rows)\n> ! \n> ! SELECT relname, b.* FROM ONLY b, pg_class where b.tableoid = pg_class.oid;\n> ! relname | aa | bb \n> ! ---------+-----+----\n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! (6 rows)\n> ! \n> ! SELECT relname, c.* FROM ONLY c, pg_class where c.tableoid = pg_class.oid;\n> ! relname | aa | cc \n> ! ---------+----------+----\n> ! c | ccc | \n> ! c | cccc | \n> ! c | ccccc | \n> ! c | cccccc | \n> ! c | ccccccc | \n> ! c | cccccccc | \n> ! (6 rows)\n> ! \n> ! SELECT relname, d.* FROM ONLY d, pg_class where d.tableoid = pg_class.oid;\n> ! relname | aa | bb | cc | dd \n> ! ---------+-----+----+----+----\n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> (6 rows)\n> \n> UPDATE a SET aa='new';\n> DELETE FROM ONLY c WHERE aa='new';\n> ! SELECT relname, a.* FROM a, pg_class where a.tableoid = pg_class.oid;\n> ! relname | aa \n> ! ---------+-----\n> ! a | new\n> ! a | new\n> ! a | new\n> ! a | new\n> ! a | new\n> ! a | new\n> ! b | new\n> ! b | new\n> ! b | new\n> ! b | new\n> ! b | new\n> ! b | new\n> ! d | new\n> ! d | new\n> ! d | new\n> ! d | new\n> ! d | new\n> ! d | new\n> (18 rows)\n> \n> ! SELECT relname, b.* FROM b, pg_class where b.tableoid = pg_class.oid;\n> ! relname | aa | bb \n> ! ---------+-----+----\n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> (12 rows)\n> \n> ! SELECT relname, c.* FROM c, pg_class where c.tableoid = pg_class.oid;\n> ! relname | aa | cc \n> ! ---------+-----+----\n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! d | new | \n> ! (6 rows)\n> ! \n> ! SELECT relname, d.* FROM d, pg_class where d.tableoid = pg_class.oid;\n> ! relname | aa | bb | cc | dd \n> ! ---------+-----+----+----+----\n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! (6 rows)\n> ! \n> ! SELECT relname, a.* FROM ONLY a, pg_class where a.tableoid = pg_class.oid;\n> ! relname | aa \n> ! ---------+-----\n> ! a | new\n> ! a | new\n> ! a | new\n> ! a | new\n> ! a | new\n> ! a | new\n> ! (6 rows)\n> ! \n> ! SELECT relname, b.* FROM ONLY b, pg_class where b.tableoid = pg_class.oid;\n> ! relname | aa | bb \n> ! ---------+-----+----\n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! b | new | \n> ! (6 rows)\n> ! \n> ! SELECT relname, c.* FROM ONLY c, pg_class where c.tableoid = pg_class.oid;\n> ! relname | aa | cc \n> ! ---------+----+----\n> (0 rows)\n> \n> ! SELECT relname, d.* FROM ONLY d, pg_class where d.tableoid = pg_class.oid;\n> ! relname | aa | bb | cc | dd \n> ! ---------+-----+----+----+----\n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> ! d | new | | | \n> (6 rows)\n> \n> DELETE FROM a;\n> ! SELECT relname, a.* FROM a, pg_class where a.tableoid = pg_class.oid;\n> ! relname | aa \n> ! ---------+----\n> (0 rows)\n> \n> ! SELECT relname, b.* FROM b, pg_class where b.tableoid = pg_class.oid;\n> ! relname | aa | bb \n> ! ---------+----+----\n> (0 rows)\n> \n> ! SELECT relname, c.* FROM c, pg_class where c.tableoid = pg_class.oid;\n> ! relname | aa | cc \n> ! ---------+----+----\n> (0 rows)\n> \n> ! SELECT relname, d.* FROM d, pg_class where d.tableoid = pg_class.oid;\n> ! relname | aa | bb | cc | dd \n> ! ---------+----+----+----+----\n> (0 rows)\n> \n> ! SELECT relname, a.* FROM ONLY a, pg_class where a.tableoid = pg_class.oid;\n> ! relname | aa \n> ! ---------+----\n> (0 rows)\n> \n> ! SELECT relname, b.* FROM ONLY b, pg_class where b.tableoid = pg_class.oid;\n> ! relname | aa | bb \n> ! ---------+----+----\n> (0 rows)\n> \n> ! SELECT relname, c.* FROM ONLY c, pg_class where c.tableoid = pg_class.oid;\n> ! relname | aa | cc \n> ! ---------+----+----\n> (0 rows)\n> \n> ! SELECT relname, d.* FROM ONLY d, pg_class where d.tableoid = pg_class.oid;\n> ! relname | aa | bb | cc | dd \n> ! ---------+----+----+----+----\n> (0 rows)\n> \n> Index: src/test/regress/sql/inherit.sql\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/test/regress/sql/inherit.sql,v\n> retrieving revision 1.1\n> diff -c -r1.1 inherit.sql\n> *** src/test/regress/sql/inherit.sql\t2000/06/09 11:12:38\t1.1\n> --- src/test/regress/sql/inherit.sql\t2000/07/02 09:30:08\n> ***************\n> *** 34,47 ****\n> INSERT INTO d(aa) VALUES('ddddddd');\n> INSERT INTO d(aa) VALUES('dddddddd');\n> \n> ! SELECT * FROM a;\n> ! SELECT * FROM b;\n> ! SELECT * FROM c;\n> ! SELECT * FROM d;\n> ! SELECT * FROM ONLY a;\n> ! SELECT * FROM ONLY b;\n> ! SELECT * FROM ONLY c;\n> ! SELECT * FROM ONLY d;\n> \n> UPDATE a SET aa='zzzz' WHERE aa='aaaa';\n> UPDATE ONLY a SET aa='zzzzz' WHERE aa='aaaaa';\n> --- 34,47 ----\n> INSERT INTO d(aa) VALUES('ddddddd');\n> INSERT INTO d(aa) VALUES('dddddddd');\n> \n> ! SELECT relname, a.* FROM a, pg_class where a.tableoid = pg_class.oid;\n> ! SELECT relname, b.* FROM b, pg_class where b.tableoid = pg_class.oid;\n> ! SELECT relname, c.* FROM c, pg_class where c.tableoid = pg_class.oid;\n> ! SELECT relname, d.* FROM d, pg_class where d.tableoid = pg_class.oid;\n> ! SELECT relname, a.* FROM ONLY a, pg_class where a.tableoid = pg_class.oid;\n> ! SELECT relname, b.* FROM ONLY b, pg_class where b.tableoid = pg_class.oid;\n> ! SELECT relname, c.* FROM ONLY c, pg_class where c.tableoid = pg_class.oid;\n> ! SELECT relname, d.* FROM ONLY d, pg_class where d.tableoid = pg_class.oid;\n> \n> UPDATE a SET aa='zzzz' WHERE aa='aaaa';\n> UPDATE ONLY a SET aa='zzzzz' WHERE aa='aaaaa';\n> ***************\n> *** 49,94 ****\n> UPDATE ONLY b SET aa='zzz' WHERE aa='aaa';\n> UPDATE a SET aa='zzzzzz' WHERE aa LIKE 'aaa%';\n> \n> ! SELECT * FROM a;\n> ! SELECT * FROM b;\n> ! SELECT * FROM c;\n> ! SELECT * FROM d;\n> ! SELECT * FROM ONLY a;\n> ! SELECT * FROM ONLY b;\n> ! SELECT * FROM ONLY c;\n> ! SELECT * FROM ONLY d;\n> \n> UPDATE b SET aa='new';\n> \n> ! SELECT * FROM a;\n> ! SELECT * FROM b;\n> ! SELECT * FROM c;\n> ! SELECT * FROM d;\n> ! SELECT * FROM ONLY a;\n> ! SELECT * FROM ONLY b;\n> ! SELECT * FROM ONLY c;\n> ! SELECT * FROM ONLY d;\n> \n> UPDATE a SET aa='new';\n> \n> DELETE FROM ONLY c WHERE aa='new';\n> \n> ! SELECT * FROM a;\n> ! SELECT * FROM b;\n> ! SELECT * FROM c;\n> ! SELECT * FROM d;\n> ! SELECT * FROM ONLY a;\n> ! SELECT * FROM ONLY b;\n> ! SELECT * FROM ONLY c;\n> ! SELECT * FROM ONLY d;\n> \n> DELETE FROM a;\n> \n> ! SELECT * FROM a;\n> ! SELECT * FROM b;\n> ! SELECT * FROM c;\n> ! SELECT * FROM d;\n> ! SELECT * FROM ONLY a;\n> ! SELECT * FROM ONLY b;\n> ! SELECT * FROM ONLY c;\n> ! SELECT * FROM ONLY d;\n> --- 49,94 ----\n> UPDATE ONLY b SET aa='zzz' WHERE aa='aaa';\n> UPDATE a SET aa='zzzzzz' WHERE aa LIKE 'aaa%';\n> \n> ! SELECT relname, a.* FROM a, pg_class where a.tableoid = pg_class.oid;\n> ! SELECT relname, b.* FROM b, pg_class where b.tableoid = pg_class.oid;\n> ! SELECT relname, c.* FROM c, pg_class where c.tableoid = pg_class.oid;\n> ! SELECT relname, d.* FROM d, pg_class where d.tableoid = pg_class.oid;\n> ! SELECT relname, a.* FROM ONLY a, pg_class where a.tableoid = pg_class.oid;\n> ! SELECT relname, b.* FROM ONLY b, pg_class where b.tableoid = pg_class.oid;\n> ! SELECT relname, c.* FROM ONLY c, pg_class where c.tableoid = pg_class.oid;\n> ! SELECT relname, d.* FROM ONLY d, pg_class where d.tableoid = pg_class.oid;\n> \n> UPDATE b SET aa='new';\n> \n> ! SELECT relname, a.* FROM a, pg_class where a.tableoid = pg_class.oid;\n> ! SELECT relname, b.* FROM b, pg_class where b.tableoid = pg_class.oid;\n> ! SELECT relname, c.* FROM c, pg_class where c.tableoid = pg_class.oid;\n> ! SELECT relname, d.* FROM d, pg_class where d.tableoid = pg_class.oid;\n> ! SELECT relname, a.* FROM ONLY a, pg_class where a.tableoid = pg_class.oid;\n> ! SELECT relname, b.* FROM ONLY b, pg_class where b.tableoid = pg_class.oid;\n> ! SELECT relname, c.* FROM ONLY c, pg_class where c.tableoid = pg_class.oid;\n> ! SELECT relname, d.* FROM ONLY d, pg_class where d.tableoid = pg_class.oid;\n> \n> UPDATE a SET aa='new';\n> \n> DELETE FROM ONLY c WHERE aa='new';\n> \n> ! SELECT relname, a.* FROM a, pg_class where a.tableoid = pg_class.oid;\n> ! SELECT relname, b.* FROM b, pg_class where b.tableoid = pg_class.oid;\n> ! SELECT relname, c.* FROM c, pg_class where c.tableoid = pg_class.oid;\n> ! SELECT relname, d.* FROM d, pg_class where d.tableoid = pg_class.oid;\n> ! SELECT relname, a.* FROM ONLY a, pg_class where a.tableoid = pg_class.oid;\n> ! SELECT relname, b.* FROM ONLY b, pg_class where b.tableoid = pg_class.oid;\n> ! SELECT relname, c.* FROM ONLY c, pg_class where c.tableoid = pg_class.oid;\n> ! SELECT relname, d.* FROM ONLY d, pg_class where d.tableoid = pg_class.oid;\n> \n> DELETE FROM a;\n> \n> ! SELECT relname, a.* FROM a, pg_class where a.tableoid = pg_class.oid;\n> ! SELECT relname, b.* FROM b, pg_class where b.tableoid = pg_class.oid;\n> ! SELECT relname, c.* FROM c, pg_class where c.tableoid = pg_class.oid;\n> ! SELECT relname, d.* FROM d, pg_class where d.tableoid = pg_class.oid;\n> ! SELECT relname, a.* FROM ONLY a, pg_class where a.tableoid = pg_class.oid;\n> ! SELECT relname, b.* FROM ONLY b, pg_class where b.tableoid = pg_class.oid;\n> ! SELECT relname, c.* FROM ONLY c, pg_class where c.tableoid = pg_class.oid;\n> ! SELECT relname, d.* FROM ONLY d, pg_class where d.tableoid = pg_class.oid;\n> Index: src/tools/make_mkid\n> ===================================================================\n> RCS file: /home/projects/pgsql/cvsroot/pgsql/src/tools/make_mkid,v\n> retrieving revision 1.4\n> diff -c -r1.4 make_mkid\n> *** src/tools/make_mkid\t2000/03/31 01:41:27\t1.4\n> --- src/tools/make_mkid\t2000/07/02 09:30:08\n> ***************\n> *** 1,6 ****\n> #!/bin/sh\n> find `pwd`/ \\( -name _deadcode -a -prune \\) -o \\\n> ! \t-type f -name '*.[chyl]' -print|sed 's;//;/;g' | mkid\n> \n> find . -name 'CVS' -prune -o -type d -print |while read DIR\n> do\n> --- 1,6 ----\n> #!/bin/sh\n> find `pwd`/ \\( -name _deadcode -a -prune \\) -o \\\n> ! \t-type f -name '*.[chyl]' -print|sed 's;//;/;g' | mkid -\n> \n> find . -name 'CVS' -prune -o -type d -print |while read DIR\n> do\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sun, 2 Jul 2000 17:58:19 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PATCHES] TABLEOID patch" } ]
[ { "msg_contents": "At 13:53 30/06/00 -0300, The Hermit Hacker wrote:\n>\n>Sounds like something that is a suitable replacement for what we have now\n>... I particularly like the 'compressed output' *but* is there some way,\n>for the paranoid, of doing an uncompressed format\n>(--dont-compress?)? Then again, how does this handle someone wanting to\n>dump the database for reloading into a non-PgSQL database? \n>\n\nAs suggested I have now added a '--plain-text, -p' option to pg_dump to\ndump it into text. Needless to say the output can not be used by pg_restore\n(although, I suppose, it is feasible). This also supports the -Z flag, to\ndump in gzip format.\n\n\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Sat, 01 Jul 2000 23:10:33 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Modified pg_dump & new pg_restore need testing..." }, { "msg_contents": "[ Charset ISO-8859-1 unsupported, converting... ]\n> Philip Warner writes:\n> \n> > As suggested I have now added a '--plain-text, -p' option to pg_dump\n> > to dump it into text.\n> \n> The *default* action of pg_dump should be to write plain text and to\n> standard output. Changing that will certainly make a lot of people very\n> unhappy.\n\nAgreed. We need a enable-compress switch.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sat, 1 Jul 2000 17:15:39 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Modified pg_dump & new pg_restore need testing..." }, { "msg_contents": "Philip Warner writes:\n\n> As suggested I have now added a '--plain-text, -p' option to pg_dump\n> to dump it into text.\n\nThe *default* action of pg_dump should be to write plain text and to\nstandard output. Changing that will certainly make a lot of people very\nunhappy.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Sat, 1 Jul 2000 23:23:53 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Modified pg_dump & new pg_restore need testing..." }, { "msg_contents": "At 23:23 1/07/00 +0200, Peter Eisentraut wrote:\n>> As suggested I have now added a '--plain-text, -p' option to pg_dump\n>> to dump it into text.\n>\n>The *default* action of pg_dump should be to write plain text and to\n>standard output. Changing that will certainly make a lot of people very\n>unhappy.\n\nGood point; I'll make it the default. The -Z {0..9] switch and -F {p|c|f}\nswitches can be used to select output compression levels and output format.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Sun, 02 Jul 2000 11:34:26 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Modified pg_dump & new pg_restore need testing..." } ]
[ { "msg_contents": "The postmaster doesn't seem to want to shut down cleanly with active\nconnections anymore.\n\nStart up a server, start up a connection (say, psql), press Ctrl-C in the\nserver window.\n\n| peter ~$ pg-install/bin/postmaster -D /home/peter/pg-install/data -p6543\n| DEBUG: Data Base System is starting up at Sat Jul 1 00:43:17 2000\n| DEBUG: Data Base System was shut down at Sat Jul 1 00:41:32 2000\n| DEBUG: Data Base System is in production state at Sat Jul 1 00:43:17 2000\n| Fast Shutdown request at Sat Jul 1 00:43:24 2000\n| Aborting any active transaction...\n| FATAL 1: The system is shutting down\n| NOTICE: AbortTransaction and not in in-progress state\n| Server process (pid 29849) exited with status 139 at Sat Jul 1 00:43:24 2000\n| Terminating any active server processes...\n| Server processes were terminated at Sat Jul 1 00:43:24 2000\n| Reinitializing shared memory and semaphores\n| DEBUG: Data Base System is starting up at Sat Jul 1 00:43:24 2000\n| DEBUG: Data Base System was interrupted being in production at Sat Jul 1 00:43:17 2000\n| DEBUG: Data Base System is in production state at Sat Jul 1 00:43:24 2000\n| DEBUG: Data Base System shutting down at Sat Jul 1 00:43:24 2000\n| DEBUG: Data Base System shut down at Sat Jul 1 00:43:24 2000\n| peter ~$\n\nThe reason I'm saying \"bug in memory system\" is this backtrace:\n\n#0 0x8190f53 in MemoryContextDeleteChildren (context=0x0) at mcxt.c:201\n201 while (context->firstchild != NULL)\n(gdb) bt\n#0 0x8190f53 in MemoryContextDeleteChildren (context=0x0) at mcxt.c:201\n#1 0x8190f85 in MemoryContextResetAndDeleteChildren (context=0x0) at mcxt.c:218\n#2 0x8087e18 in AtAbort_Memory () at xact.c:838\n#3 0x8088092 in AbortTransaction () at xact.c:1108\n#4 0x8088248 in AbortCurrentTransaction () at xact.c:1339\n#5 0x812ff0b in PostgresMain (argc=4, argv=0xbffff278, real_argc=4, real_argv=0xbffffb34) at postgres.c:1455\n#6 0x810f2c4 in DoBackend (port=0x8213290) at postmaster.c:1967\n#7 0x810eca5 in BackendStartup (port=0x8213290) at postmaster.c:1740\n#8 0x810dabc in ServerLoop () at postmaster.c:985\n#9 0x810d353 in PostmasterMain (argc=4, argv=0xbffffb34) at postmaster.c:678\n#10 0x80cf988 in main (argc=4, argv=0xbffffb34) at main.c:97\n\n\nThere are no such problems when there are no connections open at the time\nof the shutdown.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Sat, 1 Jul 2000 17:03:02 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Bug in new memory system?" }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n> The reason I'm saying \"bug in memory system\" is this backtrace:\n\nGood catch. xact.c is being careless about the possibility that\nAbortTransaction() can be called when there's not actually any\ntransaction running. I'm on it.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sat, 01 Jul 2000 14:12:10 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Bug in new memory system? " } ]
[ { "msg_contents": "[ Charset ISO-8859-1 unsupported, converting... ]\n> We've got 7 linux templates all doing the same thing, and one more\n> (s390) coming up. How about putting them all into one?\n> \n> Any news on why linux_alpha defines CPU=alpha, which\n> is used in backend/storage/ipc/Makefile:\n> \n> ifeq ($(CPU),alpha)\n> ifeq ($(CC), gcc)\n> CFLAGS+= -fno-inline\n> endif\n> ifeq ($(CC), egcs)\n> CFLAGS+= -fno-inline\n> endif\n> endif\n\nFor some reason, alpha was not defined on some platforms, I think.\n\n\n> \n> (Certainly a compiler bug -- what's the latest on that?)\n> \n> and in backend/utils/adt/Makefile:\n> \n> ifeq ($(CPU),alpha)\n> ifeq ($(CC), gcc)\n> CFLAGS+= -mieee\n> endif\n> ifeq ($(CC), egcs)\n> CFLAGS+= -mieee\n> endif\n> endif\n> \n> (Reading the documentation on this, it seems that we should perhaps\n> compile the entire source with -mieee, rather than having certain parts\n> work more \"correctly\" than others.)\n\nThere was some goofy problem in the past with this flag, and making it\nhappen only in certain places helped some problem. I don't really\nremember the cause at this point. CVS log may show it.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sat, 1 Jul 2000 17:17:26 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Linux templates" }, { "msg_contents": "We've got 7 linux templates all doing the same thing, and one more\n(s390) coming up. How about putting them all into one?\n\nAny news on why linux_alpha defines CPU=alpha, which\nis used in backend/storage/ipc/Makefile:\n\nifeq ($(CPU),alpha)\nifeq ($(CC), gcc)\nCFLAGS+= -fno-inline\nendif\nifeq ($(CC), egcs)\nCFLAGS+= -fno-inline\nendif\nendif\n\n(Certainly a compiler bug -- what's the latest on that?)\n\nand in backend/utils/adt/Makefile:\n\nifeq ($(CPU),alpha)\nifeq ($(CC), gcc)\nCFLAGS+= -mieee\nendif\nifeq ($(CC), egcs)\nCFLAGS+= -mieee\nendif\nendif\n\n(Reading the documentation on this, it seems that we should perhaps\ncompile the entire source with -mieee, rather than having certain parts\nwork more \"correctly\" than others.)\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Sat, 1 Jul 2000 23:24:20 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Linux templates" }, { "msg_contents": "Actually, from compiling your latest version , the CPU= equate did a strange\nthing. I expected CPU=alpha, but it placed instead CPU=alpha_ev56 ( or\nsomething like that ) the ev56 is a particular alpha processor ( like\npentium II, III, 486 ... ).\nno-inline for the alpha/cc compiler was necessary because the in-lining of\nassembler code ( the test and set code ) did not work without getting lock\nfailures.\nThe -mieee flag tells the cc compiler to wait for FP traps to maybe happen.\nWithout this flag the FP errors will not be ieee754(?) comforment, and\ntherefor not produce the same results as the i386 port. But u pay for this\nfeature ( u gotta wait for the FP processing to complete before going on ).\nAt one point, having the -mieee switch on ment that the compiler would\ninsert waits, even where no FP processing has occured. I do not know of the\ncurrent state of the -mieee switch ( ie more carefull where stalls are\ninserted ). I limit my usage of -mieee to only the modules that do FP work.\n/gat\n\n\nBruce Momjian wrote:\n\n> [ Charset ISO-8859-1 unsupported, converting... ]\n> > We've got 7 linux templates all doing the same thing, and one more\n> > (s390) coming up. How about putting them all into one?\n> >\n> > Any news on why linux_alpha defines CPU=alpha, which\n> > is used in backend/storage/ipc/Makefile:\n> >\n> > ifeq ($(CPU),alpha)\n> > ifeq ($(CC), gcc)\n> > CFLAGS+= -fno-inline\n> > endif\n> > ifeq ($(CC), egcs)\n> > CFLAGS+= -fno-inline\n> > endif\n> > endif\n>\n> For some reason, alpha was not defined on some platforms, I think.\n>\n> >\n> > (Certainly a compiler bug -- what's the latest on that?)\n> >\n> > and in backend/utils/adt/Makefile:\n> >\n> > ifeq ($(CPU),alpha)\n> > ifeq ($(CC), gcc)\n> > CFLAGS+= -mieee\n> > endif\n> > ifeq ($(CC), egcs)\n> > CFLAGS+= -mieee\n> > endif\n> > endif\n> >\n> > (Reading the documentation on this, it seems that we should perhaps\n> > compile the entire source with -mieee, rather than having certain parts\n> > work more \"correctly\" than others.)\n>\n> There was some goofy problem in the past with this flag, and making it\n> happen only in certain places helped some problem. I don't really\n> remember the cause at this point. CVS log may show it.\n\n\n\n", "msg_date": "Sat, 01 Jul 2000 19:35:12 -0400", "msg_from": "Uncle George <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PORTS] Re: Linux templates" } ]
[ { "msg_contents": "\nIf anybody out there is comfortable with using make, has the PG sources on\ntheir machine, and is interested in testing a new version of pg_dump,\nplease let me know.\n\nDetails of the revised pg_dump, and new pg_restore are at the end of this\nmessage.\n\nThanks,\n\nPhilip Warner.\n\n-------------\nThe basic idea is to use pg_dump to dump an *entire* database, and then use\npg_restore to choose what is restored.\n\nThe salient features are as follows:\n\n- pg_dump still used to dump database; all output is via new interface\n(virtually all of the pg_dump code is changed, but not the logic). The\nchanges are relatively minor, all the same.\n\n- the '-c' option is not used in pg-dump: it now dumps the commands to\ndelete the schema, and it is up to the user of pg_restore to decide if they\nare output.\n\n- the default output file format is stll text, but a new custom format \nwith compressed sections (the data dumps) is available via the '-Fc' flag.\nIt is NOT a text file.\n\n- pg_restore reads the backup file and, depending on the options chosen,\nproduces a script (to stdout) that can be sent to psql.\n\n- by default pg_restore outputs the schema/data in the order it was sent\nfrom pg_dump, but the --oid flag will send the output in order of\nincreasing OID, and the --rearrange flag will put all 'non-parental' (??)\nitems at the end, after the data. (eg. indexes, acls, triggers etc).\nNeedless to say that the best results com from using both of these options.\n\n- If the -c (clear) option is chosen in pg_restore, it also dumps the\n'drop' commands in reverse order at the start of the script. This *should*\nmake it more reliable than dumping them when the item is defined. It also\nmeans that triggers can be dropped.\n\n- The --toc option shows a summary of the restore operation that would be\nperformed if the --toc were not there.\n\nPlease send me an email if you are interested and have the time to test them.\n\nThanks,\n\nPhilip Warner.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Sun, 02 Jul 2000 14:36:44 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Announce: Testers needed for revised pg_dump/pg_restore" }, { "msg_contents": "\nThis is also available via ftp from:\n\n ftp://ftp.rhyme.com.au/pub/pg_backup_110.tar.gz\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Sun, 02 Jul 2000 15:36:35 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Announce: Testers needed for revised\n pg_dump/pg_restore" }, { "msg_contents": "\nHi!\nI have a problem with vacuumdb on one of my tables.\n\n(spiral:3)-[~]$vacuumdb -t osoby -v dziekanat\nNOTICE: --Relation osoby--\nNOTICE: Pages 229: Changed 0, reaped 16, Empty 0, New 0; Tup 4427: Vac 5,\nKeep/VTL 0/0, Crash 0, UnUsed 70, MinLen 64, MaxLen\n616; Re-using: Free/Avail. Space 18176/2288; EndEmpty/Avail. Pages\n1/10. CPU 0.02s/0.00u sec.\nNOTICE: Index osoby_Imiďż˝_key: Pages 44; Tuples 4427: Deleted 0. CPU\n0.01s/0.01u sec.\nNOTICE: Index osoby_pkey: Pages 15; Tuples 4427: Deleted 0. CPU\n0.00s/0.00u sec.\nNOTICE: Rel osoby: Pages: 229 --> 228; Tuple(s) moved: 0. CPU 0.01s/0.00u\nsec.\nNOTICE: FlushRelationBuffers(osoby, 228): block 223 is referenced\n(private 0, global 1)\nFATAL 1: VACUUM (vc_repair_frag): FlushRelationBuffers returned -2\npqReadData() -- backend closed the channel unexpectedly.\n\tThis probably means the backend terminated abnormally\n\tbefore or while processing the request.\nconnection to server was lost\nvacuumdb: vacuum failed\n\n\nthis table is referenced in my db by a tree of FOREIGN KEYs. \n\nhowever my db seems to work and I am able to do pg_dump\nRescently, I dumped and restored it and for a few days I was able to\ndo vacuumdb. Today, the problem is here again.\n\nany ideas?\n\nthanks for any help.\n\n\n(I use Postgres ver 7.0.2 on Debian Potato/ dual PII box)\n\n", "msg_date": "Sun, 2 Jul 2000 11:01:37 +0200 (CEST)", "msg_from": "Marcin Inkielman <[email protected]>", "msg_from_op": false, "msg_subject": "vacuumdb problem" }, { "msg_contents": "\nIt's probably best to look in the directory:\n\n ftp://ftp.rhyme.com.au/pub/\n\nfor the latest version.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Sun, 02 Jul 2000 22:46:23 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Announce: Testers needed for revised\n pg_dump/pg_restore" }, { "msg_contents": "Marcin Inkielman <[email protected]> writes:\n> NOTICE: FlushRelationBuffers(osoby, 228): block 223 is referenced\n> (private 0, global 1)\n> FATAL 1: VACUUM (vc_repair_frag): FlushRelationBuffers returned -2\n\n> this table is referenced in my db by a tree of FOREIGN KEYs. \n\nHmm, I wonder whether there is a buffer-refcount leak in the foreign\nkey stuff.\n\n> however my db seems to work and I am able to do pg_dump\n> Rescently, I dumped and restored it and for a few days I was able to\n> do vacuumdb. Today, the problem is here again.\n\nYou will probably find that stopping and restarting the postmaster\nwill make the problem go away (until it comes back again). Might\nbe an acceptable workaround to let you vacuum, until we can find\nthe bug.\n\nDo you think you can extract a reproducible example? Short of that,\nit'd be helpful to at least see the declarations of \"osoby\" and all\nassociated tables.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 02 Jul 2000 12:47:55 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: vacuumdb problem " }, { "msg_contents": "On Sun, 2 Jul 2000, Tom Lane wrote:\n\n> Date: Sun, 02 Jul 2000 12:47:55 -0400\n> From: Tom Lane <[email protected]>\n> To: Marcin Inkielman <[email protected]>\n> Cc: [email protected], [email protected]\n> Subject: Re: [GENERAL] vacuumdb problem \n> \n> Marcin Inkielman <[email protected]> writes:\n> > NOTICE: FlushRelationBuffers(osoby, 228): block 223 is referenced\n> > (private 0, global 1)\n> > FATAL 1: VACUUM (vc_repair_frag): FlushRelationBuffers returned -2\n> \n> > this table is referenced in my db by a tree of FOREIGN KEYs. \n> \n> Hmm, I wonder whether there is a buffer-refcount leak in the foreign\n> key stuff.\n> \n> > however my db seems to work and I am able to do pg_dump\n> > Rescently, I dumped and restored it and for a few days I was able to\n> > do vacuumdb. Today, the problem is here again.\n> \n> You will probably find that stopping and restarting the postmaster\n> will make the problem go away (until it comes back again). Might\n> be an acceptable workaround to let you vacuum, until we can find\n> the bug.\n\nthanks, it worked.\n\n> \n> Do you think you can extract a reproducible example? Short of that,\n> it'd be helpful to at least see the declarations of \"osoby\" and all\n> associated tables.\n> \n> \t\t\tregards, tom lane\n> \n\nsure..\n\nCREATE TABLE \"osoby\" (\n\t\"Nazwisko\" text NOT NULL,\n\t\"Imiďż˝\" text NOT NULL,\n\t\"Drugie imiďż˝\" text,\n\t\"Data urodzenia\" date NOT NULL,\n\t\"Miejsce urodzenia\" text,\n\t\"Nazwisko rodowe\" text,\n\t\"Imiďż˝ ojca\" text,\n\t\"Imiďż˝ matki\" text,\n\t\"Nazwisko rodowe matki\" text,\n\t\"Pďż˝eďż˝\" int2,\n\t\"Stan cywilny\" int2,\n\t\"Numer dowodu osobistego\" text,\n\t\"PESEL\" text,\n\t\"NIP\" text,\n\t\"Nazwa szkoďż˝y\" text,\n\t\"Nr ďż˝wiadectwa\" text,\n\t\"Data wystawienia ďż˝wiad\" date,\n\t\"Miejsce wyst ďż˝waid\" text,\n\t\"ulica\" text,\n\t\"kod\" text,\n\t\"poczta\" text,\n\t\"miejscowo��\" text,\n\t\"gmina\" text,\n\t\"wojew\" text,\n\t\"telefon\" text,\n\t\"Adres koresp - ulica\" text,\n\t\"Adres koresp - kod\" text,\n\t\"Adres koresp - poczta\" text,\n\t\"Adres koresp - miejscowo��\" text,\n\t\"Adres koresp - gmina\" text,\n\t\"Adres koresp - wojew\" text,\n\t\"Adres koresp - telefon\" text,\n\t\"ob wojskowy\" int2,\n\t\"WKU\" int4,\n\t\"WKU - kod\" text,\n\t\"WKU - miejscowo��\" text,\n\t\"WKU - ulica\" text,\n\t\"osoba_id\" int4 NOT NULL);\n\n\nCREATE TABLE \"teczki_t\" (\n\t\"osoba_id\" int4 NOT NULL,\n\t\"tr_wdz\" int4 NOT NULL,\n\t\"dyplom_id\" int2 NOT NULL,\n\t\"termin_egzaminu_id\" int2 NOT NULL,\n\t\"teczka_id\" int4 NOT NULL,\n\t\"pracownik_id\" int4,\n\t\"stan_id\" int4,\n\t\"data\" datetime,\n\t\"pracownik_id_r\" int4,\n\t\"data_r\" datetime,\n\t\"uwagi\" text,\n\t\"poz_kurs\" int2,\n\t\"kurs_komp\" int2,\n\tFOREIGN KEY (osoba_id) REFERENCES \"osoby\"(osoba_id) ON DELETE CASCADE ON UPDATE CASCADE,\n\tFOREIGN KEY (dyplom_id) REFERENCES \"Dyplom\"(symbol) ON UPDATE CASCADE,\n\tFOREIGN KEY (stan_id) REFERENCES \"stany\"(stan_id) ON UPDATE CASCADE\n\t);\n\n\nCREATE TABLE \"studenci\" (\n\t\"numer_albumu\" int4 NOT NULL,\n\t\"osoba_id\" int4 NOT NULL,\n\t\"tr_wdz_id\" int4 NOT NULL,\n\t\"grupa_id\" int4,\n\t\"grupa_jez_id\" int4,\n\t\"semestr\" int4 NOT NULL,\n\t\"repeta\" bool,\n\t\"warunki\" int4,\n\t\"znacznik\" int4,\n\tFOREIGN KEY (tr_wdz_id) REFERENCES tr_wdz(tr_wdz_id) ON DELETE RESTRICT ON UPDATE CASCADE,\n\tFOREIGN KEY (osoba_id) REFERENCES osoby(osoba_id) ON DELETE RESTRICT ON UPDATE CASCADE,\n\tFOREIGN KEY (grupa_id) REFERENCES grupy_stud(grupa_id) ON DELETE RESTRICT ON UPDATE CASCADE,\n\tFOREIGN KEY (grupa_jez_id) REFERENCES grupy_jez_stud(grupa_jez_id) ON DELETE RESTRICT ON UPDATE CASCADE\n\t);\n\n\nCREATE TABLE \"rejestr_decyzji\" (\n\t\"decyzja_id\" int4 NOT NULL,\n\t\"osoba_id\" int4,\n\t\"numer_albumu\" int4,\n\t\"data_pocz\" date,\n\t\"data_konca\" date,\n\t\"typ_decyzji_id\" int4,\n\t\"podjeta_przez\" int4,\n\t\"podjeta_dnia\" datetime,\n\t\"anulowana_przez\" int4,\n\t\"anulowana_dnia\" datetime,\n\t\"uwagi\" text,\n\tFOREIGN KEY (osoba_id) REFERENCES \"osoby\"(osoba_id) ON DELETE CASCADE ON UPDATE CASCADE,\n\tFOREIGN KEY (numer_albumu) REFERENCES album(numer_albumu) ON DELETE CASCADE ON UPDATE CASCADE\n\t);\n\nCREATE TABLE \"konto\" (\n\t\"osoba_id\" int4 NOT NULL,\n\t\"login\" text,\n\t\"gr_id\" int4,\n\t\"alias\" text,\n\t\"pass\" text,\n\t\"homedir\" text,\n\t\"ma_konto\" bool,\n\tFOREIGN KEY (osoba_id) REFERENCES \"osoby\"(osoba_id) ON DELETE CASCADE ON UPDATE CASCADE\n\t);\n\n/*only 2 indexes on osoby*/\n\nCREATE UNIQUE INDEX \"osoby_pkey\" on \"osoby\" using btree (\"osoba_id\" \"int4_ops\" );\nCREATE UNIQUE INDEX \"osoby_Imiďż˝_key\" on \"osoby\" using btree (\n\"Imiďż˝\" \"text_ops\", \"Nazwisko\" \"text_ops\", \"Data urodzenia\" \"date_ops\" );\n\n\nalmost all this tables are referenced by other tables (however this\nreferences are not cyclic). the db is not very large from the point of\nview of data but it contains >200 tables.\n\nI hope it will help a little...\n\n regards\n\n-- \nmi\n\n", "msg_date": "Sun, 2 Jul 2000 22:23:02 +0200 (CEST)", "msg_from": "Marcin Inkielman <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] vacuumdb problem " } ]
[ { "msg_contents": "Hello,\n\nI tried to install postgresql-6.5.3 through ports to FreeBSD-3.4,\nbut couldn't be successful.\nIt stopped like below. It was the same error when I tried to install postgresql-6.5.2.\n\nCould anyone give me any advices? Thank you very much in advance.\n\n--------------------------------------------------begin----------------------------------\n---------------------------\n#make install\n..........\n..........\n..........\ngmake[3]: Entering directory\n`/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/pl/plpgsql/s\nrc'\n/usr/bin/install -c -o root -g wheel -m 644 libplpgsql.so.1\n/usr/local/pgsql/lib/plpgsql.so\ngmake[3]: Leaving directory\n`/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/pl/plpgsql/sr\nc'\ngmake[2]: Leaving directory\n`/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/pl/plpgsql'\ngmake[1]: Leaving directory\n`/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/pl'\ncat ../register.txt\n\n\n(1999-10-13)\nPostgreSQL has a Web site at http://www.postgresql.org/ which carries details\non the latest release, upcoming features, and other information to make your\nwork or play with PostgreSQL more productive.\n\nPlease check the following URL for a listing of the current user-support\nmailing lists:\n\n http://www.postgresql.org -> Info Central -> Mailing Lists\n\nAll of the mailing lists are currently archived and viewable at:\n\n http://www.postgresql.org -> Info Central -> Mailing Lists\n\nAnd, so that we have an idea of who is using what, please connect to the\nfollowing registration URL:\n\n http://www.postgresql.org -> Helping Us -> Survey/Register\n\n\nThank you for choosing PostgreSQL, the most advanced open source database\nengine.\n\n\ngmake -C man install\ngmake[1]: Entering directory\n`/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/man'\nmkdir /usr/local/pgsql/man\nmkdir: /usr/local/pgsql/man: File exists\ngmake[1]: [install] Error 1 (ignored)\nmkdir /usr/local/pgsql/man/man1\nmkdir /usr/local/pgsql/man/man3\nmkdir /usr/local/pgsql/man/man5\nmkdir /usr/local/pgsql/man/manl\ncp *.1* /usr/local/pgsql/man/man1\ncp *.3* /usr/local/pgsql/man/man3\ncp *.5* /usr/local/pgsql/man/man5\ncp *.l* /usr/local/pgsql/man/manl\ngmake[1]: Leaving directory\n`/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/man'\n===> Generating temporary packing list\nInitializing PostgreSQL Databases - this may take a few minutes...\nUnknown option: `-l'\nUsage: su [ -bcdefilmnqstvVxX ] [ argument ... ].\n*** Error code 1\n\nStop.\n*** Error code 1\n\nStop.\n*** Error code 1\n\nStop.\n*** Error code 1\n\nStop.\n\n---------------------end---------------------------------------\n\n\n\n------------------------//~----------\nBest Regards,\nShen Jianxiang\[email protected]\nNEC Informatec Systems Ltd.\nNetwork Service Platform Division\n----------------------------------^-^\n\n", "msg_date": "Mon, 3 Jul 2000 09:03:27 +0900", "msg_from": "\"Shen Jianxiang(Outlook)\" <[email protected]>", "msg_from_op": true, "msg_subject": "fail in installing postgresql-6.5.3 to FreeBSD" }, { "msg_contents": "\ndo the postgresql7 port ... \n\nOn Mon, 3 Jul 2000, Shen Jianxiang(Outlook) wrote:\n\n> Hello,\n> \n> I tried to install postgresql-6.5.3 through ports to FreeBSD-3.4,\n> but couldn't be successful.\n> It stopped like below. It was the same error when I tried to install postgresql-6.5.2.\n> \n> Could anyone give me any advices? Thank you very much in advance.\n> \n> --------------------------------------------------begin----------------------------------\n> ---------------------------\n> #make install\n> ..........\n> ..........\n> ..........\n> gmake[3]: Entering directory\n> `/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/pl/plpgsql/s\n> rc'\n> /usr/bin/install -c -o root -g wheel -m 644 libplpgsql.so.1\n> /usr/local/pgsql/lib/plpgsql.so\n> gmake[3]: Leaving directory\n> `/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/pl/plpgsql/sr\n> c'\n> gmake[2]: Leaving directory\n> `/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/pl/plpgsql'\n> gmake[1]: Leaving directory\n> `/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/pl'\n> cat ../register.txt\n> \n> \n> (1999-10-13)\n> PostgreSQL has a Web site at http://www.postgresql.org/ which carries details\n> on the latest release, upcoming features, and other information to make your\n> work or play with PostgreSQL more productive.\n> \n> Please check the following URL for a listing of the current user-support\n> mailing lists:\n> \n> http://www.postgresql.org -> Info Central -> Mailing Lists\n> \n> All of the mailing lists are currently archived and viewable at:\n> \n> http://www.postgresql.org -> Info Central -> Mailing Lists\n> \n> And, so that we have an idea of who is using what, please connect to the\n> following registration URL:\n> \n> http://www.postgresql.org -> Helping Us -> Survey/Register\n> \n> \n> Thank you for choosing PostgreSQL, the most advanced open source database\n> engine.\n> \n> \n> gmake -C man install\n> gmake[1]: Entering directory\n> `/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/man'\n> mkdir /usr/local/pgsql/man\n> mkdir: /usr/local/pgsql/man: File exists\n> gmake[1]: [install] Error 1 (ignored)\n> mkdir /usr/local/pgsql/man/man1\n> mkdir /usr/local/pgsql/man/man3\n> mkdir /usr/local/pgsql/man/man5\n> mkdir /usr/local/pgsql/man/manl\n> cp *.1* /usr/local/pgsql/man/man1\n> cp *.3* /usr/local/pgsql/man/man3\n> cp *.5* /usr/local/pgsql/man/man5\n> cp *.l* /usr/local/pgsql/man/manl\n> gmake[1]: Leaving directory\n> `/usr/ports/databases/postgresql/work/postgresql-6.5.3/src/man'\n> ===> Generating temporary packing list\n> Initializing PostgreSQL Databases - this may take a few minutes...\n> Unknown option: `-l'\n> Usage: su [ -bcdefilmnqstvVxX ] [ argument ... ].\n> *** Error code 1\n> \n> Stop.\n> *** Error code 1\n> \n> Stop.\n> *** Error code 1\n> \n> Stop.\n> *** Error code 1\n> \n> Stop.\n> \n> ---------------------end---------------------------------------\n> \n> \n> \n> ------------------------//~----------\n> Best Regards,\n> Shen Jianxiang\n> [email protected]\n> NEC Informatec Systems Ltd.\n> Network Service Platform Division\n> ----------------------------------^-^\n> \n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n", "msg_date": "Wed, 5 Jul 2000 19:30:02 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: fail in installing postgresql-6.5.3 to FreeBSD" }, { "msg_contents": ">> ===> Generating temporary packing list\n>> Initializing PostgreSQL Databases - this may take a few minutes...\n>> Unknown option: `-l'\n>> Usage: su [ -bcdefilmnqstvVxX ] [ argument ... ].\n>> *** Error code 1\n\nI don't know what these messages are from, but they sure are not from\nany code that's executed by the standard Postgres install sequence.\nPostgres does not invoke su, ever.\n\nMy guess is you are trying to use some sort of platform-specific\ninstallation script (an RPM or something like one) with the wrong kind\nof platform. Try getting an unmodified Postgres tarball and installing\nit per our instructions, instead. (While you're at it, I recommend\nusing 7.0.2, not 6.5.anything...)\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 05 Jul 2000 18:52:29 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: fail in installing postgresql-6.5.3 to FreeBSD " }, { "msg_contents": "On Wed, 5 Jul 2000, Tom Lane wrote:\n\n> >> ===> Generating temporary packing list\n> >> Initializing PostgreSQL Databases - this may take a few minutes...\n> >> Unknown option: `-l'\n> >> Usage: su [ -bcdefilmnqstvVxX ] [ argument ... ].\n> >> *** Error code 1\n> \n> I don't know what these messages are from, but they sure are not from\n> any code that's executed by the standard Postgres install sequence.\n> Postgres does not invoke su, ever.\n\nhe's trying to install it out of ports under FreeBSD ... the install runs\nas root, but does a 'su' to the postgres user to do the initdb ... my\nonly major guess is that his ports are out of sync with his operating\nsystem ...\n\n\n", "msg_date": "Wed, 5 Jul 2000 20:14:41 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: fail in installing postgresql-6.5.3 to FreeBSD " }, { "msg_contents": "Thanks for your replies.\n\nYes, I tried to install PostgreSQL to FreeBSD using port, as 'root'.\nAs your recomend, I tried to use port postgresql7, but I was stopped\nat the same error message.\nMay I run 'su [-bcdefilmnqstvVxX ] [ argument ... ]' at command line?\n\nThank you.\n\n::>-----Original Message-----\n::>From: [email protected]\n::>[mailto:[email protected]]On Behalf Of The Hermit Hacker\n::>Sent: Thursday, July 06, 2000 8:15 AM\n::>To: Tom Lane\n::>Cc: Shen Jianxiang(Outlook); [email protected];\n::>[email protected]\n::>Subject: Re: [HACKERS] fail in installing postgresql-6.5.3\n::>to FreeBSD\n::>\n::>\n::>On Wed, 5 Jul 2000, Tom Lane wrote:\n::>\n::>> >> ===> Generating temporary packing list\n::>> >> Initializing PostgreSQL Databases - this may take a\n::>few minutes...\n::>> >> Unknown option: `-l'\n::>> >> Usage: su [ -bcdefilmnqstvVxX ] [ argument ... ].\n::>> >> *** Error code 1\n::>>\n::>> I don't know what these messages are from, but they sure\n::>are not from\n::>> any code that's executed by the standard Postgres\n::>install sequence.\n::>> Postgres does not invoke su, ever.\n::>\n::>he's trying to install it out of ports under FreeBSD ...\n::>the install runs\n::>as root, but does a 'su' to the postgres user to do the\n::>initdb ... my\n::>only major guess is that his ports are out of sync with\n::>his operating\n::>system ...\n::>\n------------------------//~----------\nBest Regards,\nShen Jianxiang\[email protected]\nNEC Informatec Systems Ltd.\nNetwork Service Platform Division\n----------------------------------^-^\n\n", "msg_date": "Thu, 6 Jul 2000 11:38:01 +0900", "msg_from": "\"Shen Jianxiang(Outlook)\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: fail in installing postgresql-6.5.3 to FreeBSD " }, { "msg_contents": "On Thu, 6 Jul 2000, Shen Jianxiang(Outlook) wrote:\n\n> Thanks for your replies.\n> \n> Yes, I tried to install PostgreSQL to FreeBSD using port, as 'root'.\n> As your recomend, I tried to use port postgresql7, but I was stopped\n> at the same error message.\n> May I run 'su [-bcdefilmnqstvVxX ] [ argument ... ]' at command line?\n\ndo a 'which su'? are you, but some odd chance, running a non-FreeBSD su?\n\n\n> \n> Thank you.\n> \n> ::>-----Original Message-----\n> ::>From: [email protected]\n> ::>[mailto:[email protected]]On Behalf Of The Hermit Hacker\n> ::>Sent: Thursday, July 06, 2000 8:15 AM\n> ::>To: Tom Lane\n> ::>Cc: Shen Jianxiang(Outlook); [email protected];\n> ::>[email protected]\n> ::>Subject: Re: [HACKERS] fail in installing postgresql-6.5.3\n> ::>to FreeBSD\n> ::>\n> ::>\n> ::>On Wed, 5 Jul 2000, Tom Lane wrote:\n> ::>\n> ::>> >> ===> Generating temporary packing list\n> ::>> >> Initializing PostgreSQL Databases - this may take a\n> ::>few minutes...\n> ::>> >> Unknown option: `-l'\n> ::>> >> Usage: su [ -bcdefilmnqstvVxX ] [ argument ... ].\n> ::>> >> *** Error code 1\n> ::>>\n> ::>> I don't know what these messages are from, but they sure\n> ::>are not from\n> ::>> any code that's executed by the standard Postgres\n> ::>install sequence.\n> ::>> Postgres does not invoke su, ever.\n> ::>\n> ::>he's trying to install it out of ports under FreeBSD ...\n> ::>the install runs\n> ::>as root, but does a 'su' to the postgres user to do the\n> ::>initdb ... my\n> ::>only major guess is that his ports are out of sync with\n> ::>his operating\n> ::>system ...\n> ::>\n> ------------------------//~----------\n> Best Regards,\n> Shen Jianxiang\n> [email protected]\n> NEC Informatec Systems Ltd.\n> Network Service Platform Division\n> ----------------------------------^-^\n> \n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n", "msg_date": "Wed, 5 Jul 2000 23:48:38 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "RE: fail in installing postgresql-6.5.3 to FreeBSD " }, { "msg_contents": "\n::>On Thu, 6 Jul 2000, Shen Jianxiang(Outlook) wrote:\n::>\n::>> Thanks for your replies.\n::>> \n::>> Yes, I tried to install PostgreSQL to FreeBSD using \n::>port, as 'root'.\n::>> As your recomend, I tried to use port postgresql7, but I \n::>was stopped\n::>> at the same error message.\n::>> May I run 'su [-bcdefilmnqstvVxX ] [ argument ... ]' at \n::>command line?\n::>\n::>do a 'which su'? are you, but some odd chance, running a \n::>non-FreeBSD su?\n::>\n::>\nI am not sure I understand what's your meaning...\nWhen I install I have to login as 'root'. I don't know how can I \nbecome a \"FreeBSD su\". \n\n------------------------//~----------\nBest Regards,\nShen Jianxiang\[email protected]\nNEC Informatec Systems Ltd. \nNetwork Service Platform Division\n----------------------------------^-^\n\n", "msg_date": "Thu, 6 Jul 2000 12:00:59 +0900", "msg_from": "\"Shen Jianxiang(Outlook)\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: fail in installing postgresql-6.5.3 to FreeBSD " } ]
[ { "msg_contents": "\n> - the default output file format is a custom format with compressed\n> sections (the data dumps). It is NOT a text file.\n\nCan this be turned off, or made to be feature, that you can turn on ? \nImho most dumps will be piped to a locally \noptimized compressor (a tape, a storage manager, lzop, ...) anyway, \nthus most of the time a backup would compress twice.\n\nAndreas\n", "msg_date": "Mon, 3 Jul 2000 09:56:36 +0200 ", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: Modified pg_dump & new pg_restore need testing..." } ]
[ { "msg_contents": "At 09:56 3/07/00 +0200, Zeugswetter Andreas SB wrote:\n>\n>> - the default output file format is a custom format with compressed\n>> sections (the data dumps). It is NOT a text file.\n>\n>Can this be turned off, or made to be feature, that you can turn on ? \n>Imho most dumps will be piped to a locally \n>optimized compressor (a tape, a storage manager, lzop, ...) anyway, \n>thus most of the time a backup would compress twice.\n\nYes. The default is now plain text for compatibility with the original\npg_dump (and with pg_dumpall). But by going to plain text output you gain\nnone of the features of pg_restore (reordering, selection etc).\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Mon, 03 Jul 2000 18:02:10 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: AW: Modified pg_dump & new pg_restore need\n testing..." } ]
[ { "msg_contents": "\n> > As suggested I have now added a '--plain-text, -p' option to pg_dump\n> > to dump it into text.\n> \n> The *default* action of pg_dump should be to write plain text and to\n> standard output. Changing that will certainly make a lot of \n> people very unhappy.\n\nImho the default should be text for anything except data. The data should\nimho \nbe in a format similar to a binary cursor. I say similar, because this\nformat should \nprobably be converted to a network byte order, so you can restore on another\nhardware. Imho calling type output and input functions during backup/restore\nis a substantial overhead that would best be avoided.\n\nI agree, that the output should not be compressed by default.\n\nAndreas\n", "msg_date": "Mon, 3 Jul 2000 10:18:19 +0200 ", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: Modified pg_dump & new pg_restore need testing..." }, { "msg_contents": "Zeugswetter Andreas SB <[email protected]> writes:\n> Imho the default should be text for anything except data. The data\n> should imho be in a format similar to a binary cursor. I say similar,\n> because this format should probably be converted to a network byte\n> order, so you can restore on another hardware. Imho calling type\n> output and input functions during backup/restore is a substantial\n> overhead that would best be avoided.\n\nI think this would be an extremely *bad* idea. One of the principal\nfunctions of pg_dump is to provide a portable representation of data;\nthat is, portable across machines and across Postgres versions (where\nthe internal representation of data may change!).\n\nNot only should binary data representation not be the default, IMHO\nwe shouldn't offer it as an option either. Otherwise dbadmins will\npromptly shoot themselves in the foot with it. Do you want to field\nthe support calls saying \"help, I already blew away my 7.n installation\nbut 7.n+1 won't read my pg_dump backup\"?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 10:57:43 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: Modified pg_dump & new pg_restore need testing... " } ]
[ { "msg_contents": "\n> > > > > In my mind the point of the \"database\" concept is to \n> > > provide a domain\n> > > > > within which custom datatypes and functions are available.\n> > > >\n> > > \n> > > AFAIK few users understand it and many users have wondered\n> > > why we couldn't issue cross \"database\" queries.\n> > \n> > Imho the same issue is access to tables on another machine.\n> > If we \"fix\" that, access to another db on the same instance is just\n> > a variant of the above. \n> >\n> \n> What is a difference between SCHAMA and your \"database\" ?\n> I myself am confused about them.\n\n\"my *database*\" corresponds to the current database, which is created with\n\"create database\" in postgresql. It corresponds to the catalog concept in\nSQL99.\n\nThe schema is below the database. Access to different schemas with one\nconnection\nis mandatory. Access to different catalogs (databases) with one connection\nis not mandatory,\nbut should imho be solved analogous to access to another catalog on a\ndifferent \n(SQL99) cluster. This would be a very nifty feature.\n\nAndreas \n", "msg_date": "Mon, 3 Jul 2000 10:28:05 +0200 ", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: Big 7.1 open items " } ]
[ { "msg_contents": "At 10:18 3/07/00 +0200, Zeugswetter Andreas SB wrote:\n>\n>Imho the default should be text for anything except data. The data should\n>imho \n>be in a format similar to a binary cursor. I say similar, because this\n>format should \n>probably be converted to a network byte order, so you can restore on another\n>hardware. Imho calling type output and input functions during backup/restore\n>is a substantial overhead that would best be avoided.\n>\n\nHave a look at the utilities. I have attempted to make the custom format\ntransportable (at least as transportable as pg_dump). If anyone has a 64\nbit machine with weird byte order I'd love them to test restoring a backup\nmade on another machine. If it does not work, I am happy to make it work\n(at least as well as pg_dump works).\n\nIt just occurred to me that I may be missing something...do you mean you\nwould prefer to avoid the 'COPY' commands as the backup technique? If so, I\nthink that's best left to somebody else (or me, but not now, and probably\nnot until the WAL is implemented).\n\nAlso, defining your own output format is pretty easy; the new code defines\na top level interface (for pg_dump and pg_restore) and an archiver\ninterface (for output file formats), if you really want a different format.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Mon, 03 Jul 2000 18:33:21 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: AW: Modified pg_dump & new pg_restore need\n testing..." } ]
[ { "msg_contents": "\n> At 09:56 3/07/00 +0200, Zeugswetter Andreas SB wrote:\n> >\n> >> - the default output file format is a custom format with compressed\n> >> sections (the data dumps). It is NOT a text file.\n> >\n> >Can this be turned off, or made to be feature, that you can \n> turn on ? \n> >Imho most dumps will be piped to a locally \n> >optimized compressor (a tape, a storage manager, lzop, ...) anyway, \n> >thus most of the time a backup would compress twice.\n> \n> Yes. The default is now plain text for compatibility with the original\n> pg_dump (and with pg_dumpall). But by going to plain text \n> output you gain\n> none of the features of pg_restore (reordering, selection etc).\n\nCould the text format be changed in a compatible way, \nthat would allow pg_restore's features ? I am thinking of inserted comment \nlines that describe the needed sections. \n\nImho the new default format does not need to be compatible \nwith pg_dump's output. I would still not compress by default.\n\nAndreas\n", "msg_date": "Mon, 3 Jul 2000 10:39:12 +0200 ", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: Modified pg_dump & new pg_restore need testing.\n\t.." } ]
[ { "msg_contents": "\n> It just occurred to me that I may be missing something...do you mean you\n> would prefer to avoid the 'COPY' commands as the backup technique? If so,\nI\n> think that's best left to somebody else (or me, but not now, and probably\n> not until the WAL is implemented).\n\nYes, that was what I was thinking. I don't see a conflicting area with WAL.\nI would see this as an extension to libpq and how it handles binary cursors\nacross different platforms.\n\n> \n> Also, defining your own output format is pretty easy; the new \n> code defines\n> a top level interface (for pg_dump and pg_restore) and an archiver\n> interface (for output file formats), if you really want a \n> different format.\n\nSounds like you are doing a great job :-)\n\nMy favorite defaults would probably be -Fc -Z0 as you noted in another\nposting.\n\nAndreas\n", "msg_date": "Mon, 3 Jul 2000 10:49:21 +0200 ", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: Modified pg_dump & new pg_restore need testing.\n\t.." } ]
[ { "msg_contents": "At 10:39 3/07/00 +0200, Zeugswetter Andreas SB wrote:\n>\n>Could the text format be changed in a compatible way, \n>that would allow pg_restore's features ? I am thinking of inserted comment \n>lines that describe the needed sections. \n\nQuite possibly; although part of my motivation in writing to the new format\nwas to avoid writing a text parser. There is no reason why it couldn't use\na text file by doing a pass through the file to construct a TOC, but this\nseems like a bad idea for large backups, and not really necessary for small\nbackups (especially with the -Z option).\n\nThat said, I'm happy to give it a go - the output file is already pretty\n'parsable'.\n\n\n>Imho the new default format does not need to be compatible \n>with pg_dump's output. I would still not compress by default.\n\nI'm happy to do this; at the moment it uses zlib default compression by\ndefault for the custom format, and no compression for the plain text\noutput. I'm happy to make uncompressed output in all cases, and require the\n'-Z' qualifier before I compress.\n\nDoes anybody else have a preference on this?\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Mon, 03 Jul 2000 18:50:04 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: AW: AW: Modified pg_dump & new pg_restore need\n testing..." } ]
[ { "msg_contents": "\nPhilip Warner [mailto:[email protected]] wrote:\n> At 10:39 3/07/00 +0200, Zeugswetter Andreas SB wrote:\n> >\n> >Could the text format be changed in a compatible way, \n> >that would allow pg_restore's features ? I am thinking of \n> inserted comment \n> >lines that describe the needed sections. \n> \n> Quite possibly; although part of my motivation in writing to \n> the new format\n> was to avoid writing a text parser. \n\nYes, you are probably right.\n\n> There is no reason why it couldn't use\n> a text file by doing a pass through the file to construct a \n> TOC, but this\n> seems like a bad idea for large backups, and not really \n> necessary for small\n> backups (especially with the -Z option).\n> \n> That said, I'm happy to give it a go - the output file is \n> already pretty\n> 'parsable'.\n\nI guess it is probably really not worth the trouble.\n\n> \n> \n> >Imho the new default format does not need to be compatible \n> >with pg_dump's output. I would still not compress by default.\n> \n> I'm happy to do this; at the moment it uses zlib default \n> compression by\n\nDo we already check for zlib ? It is probably not available everywhere.\n\nAndreas\n", "msg_date": "Mon, 3 Jul 2000 10:56:27 +0200 ", "msg_from": "Zeugswetter Andreas SB <[email protected]>", "msg_from_op": true, "msg_subject": "AW: AW: AW: Modified pg_dump & new pg_restore need test\n\ting..." } ]
[ { "msg_contents": "\n\nHi. I just came accross the big trouble.\n\" ERROR: Procedures cannot take more than 8 arguments\"\nReally?!? Does anybody know what to do to make it take 9 and more argments.\ngrishick\[email protected] \n\n", "msg_date": "Mon, 3 Jul 2000 13:15:07 +0400", "msg_from": "\"Grigori Soloviov\" <[email protected]>", "msg_from_op": true, "msg_subject": "Fw: PLPGSQL - does not support more than 8 args?" } ]
[ { "msg_contents": "At 10:56 3/07/00 +0200, Zeugswetter Andreas SB wrote:\n>\n>Do we already check for zlib ? It is probably not available everywhere.\n\nThat may be true - although I thought it was pretty universal. \n\nI will have a look at using something 'HAVE_ZLIB' setting to enable/disable\ncompressed output...I presume 'configure' needs to be changed? Is that right?\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Mon, 03 Jul 2000 19:18:49 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: AW: AW: AW: Modified pg_dump & new pg_restore\n need testing..." }, { "msg_contents": "On Mon, 3 Jul 2000, Philip Warner wrote:\n\n> At 10:56 3/07/00 +0200, Zeugswetter Andreas SB wrote:\n> >\n> >Do we already check for zlib ? It is probably not available everywhere.\n> \n> That may be true - although I thought it was pretty universal. \n> \n> I will have a look at using something 'HAVE_ZLIB' setting to enable/disable\n> compressed output...I presume 'configure' needs to be changed? Is that right?\n\nthat will be correct, but its a quick/easy change ... I'll add the code in\ntonight so that its something we check for and set in config.h, so as long\nas you include that, you'll have that setting available ...\n\n\n", "msg_date": "Mon, 3 Jul 2000 22:39:04 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: AW: AW: Modified pg_dump & new pg_restore need\n testing..." }, { "msg_contents": "At 22:39 3/07/00 -0300, The Hermit Hacker wrote:\n>\n>that will be correct, but its a quick/easy change ... I'll add the code in\n>tonight so that its something we check for and set in config.h, so as long\n>as you include that, you'll have that setting available ...\n>\n\nOK. Heres a patch against CVS (pre new pg_dump). This is a complete patch,\nso you can throw away the last two...\n\nBye for now,\n\nPhilip\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/", "msg_date": "Tue, 04 Jul 2000 15:19:43 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: AW: AW: AW: Modified pg_dump & new pg_restore\n need testing..." }, { "msg_contents": "Do you want this applied?\n\n> At 22:39 3/07/00 -0300, The Hermit Hacker wrote:\n> >\n> >that will be correct, but its a quick/easy change ... I'll add the code in\n> >tonight so that its something we check for and set in config.h, so as long\n> >as you include that, you'll have that setting available ...\n> >\n> \n> OK. Heres a patch against CVS (pre new pg_dump). This is a complete patch,\n> so you can throw away the last two...\n> \n> Bye for now,\n> \n> Philip\n\n[ Attachment, skipping... ]\n\n> \n> ----------------------------------------------------------------\n> Philip Warner | __---_____\n> Albatross Consulting Pty. Ltd. |----/ - \\\n> (A.C.N. 008 659 498) | /(@) ______---_\n> Tel: (+61) 0500 83 82 81 | _________ \\\n> Fax: (+61) 0500 83 82 82 | ___________ |\n> Http://www.rhyme.com.au | / \\|\n> | --________--\n> PGP key available upon request, | /\n> and from pgp5.ai.mit.edu:11371 |/\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 4 Jul 2000 01:27:22 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: AW: AW: Modified pg_dump & new pg_restore need testing..." } ]
[ { "msg_contents": "At 10:49 3/07/00 +0200, Zeugswetter Andreas SB wrote:\n>> It just occurred to me that I may be missing something...do you mean you\n>> would prefer to avoid the 'COPY' commands as the backup technique? If so,\n>I\n>> think that's best left to somebody else (or me, but not now, and probably\n>> not until the WAL is implemented).\n>\n>Yes, that was what I was thinking. I don't see a conflicting area with WAL.\n>I would see this as an extension to libpq and how it handles binary cursors\n>across different platforms.\n\nOK: I know nothing about this, but I'd be interested to learn.\n\nThe reason I mentioned the WAL is that I think that any more advanced\nbackup strategy needs to be integrated into the journaling system (ie. the\nWAL). Ideally, we should be able to snapshot a consistent view of raw data\nand apply a copy of the journal to this snapshot. In other systems I use\n(Dec/Rdb), this is done via a page-based backup (I think) and an\nafter-image journal (both of which I *think* are at a lower level than\ntables & rows - more like pages & slots.\n\nAs a result, I had planned to wait and see what the WAL did for backup\nstrategies...\n\nBut, in the mean time, if you think a binary pg_dump is possible (an\nrelatively easy), I'd be interested to try to integrate it...\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Mon, 03 Jul 2000 19:49:58 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: AW: AW: Modified pg_dump & new pg_restore need\n testing..." } ]
[ { "msg_contents": "\nCan anyone tell me what the concept of a \"group\" is in\nbackend/libpq/portal.c? There are functions like PQntuplesGroup,\nPQfnumberGroup, PQfsizeGroup, PQftypeGroup etc etc.\n", "msg_date": "Mon, 03 Jul 2000 23:10:25 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": true, "msg_subject": "Backend Question" }, { "msg_contents": "Chris Bitmead <[email protected]> writes:\n> Can anyone tell me what the concept of a \"group\" is in\n> backend/libpq/portal.c? There are functions like PQntuplesGroup,\n> PQfnumberGroup, PQfsizeGroup, PQftypeGroup etc etc.\n\nOnce upon a time it probably supported something like the ability to\nmerge tuples of different types into a SELECT result. It does not\nnow: a little scanning shows there is no way to get no_groups above 1\nin the current logic.\n\nYou should realize, however, that all that backend-PQexec code is dead\ncode and has been for a long time. It's unlikely that it works at all\nany more, considering all the changes that have gone on while it sits\nuntested on the sidelines. I proposed ripping it out a few months ago,\nand no one objected ... but I haven't got round to it yet.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 10:31:49 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Backend Question " }, { "msg_contents": "Tom Lane wrote:\n> \n> Chris Bitmead <[email protected]> writes:\n> > Can anyone tell me what the concept of a \"group\" is in\n> > backend/libpq/portal.c? There are functions like PQntuplesGroup,\n> > PQfnumberGroup, PQfsizeGroup, PQftypeGroup etc etc.\n> \n> Once upon a time it probably supported something like the ability to\n> merge tuples of different types into a SELECT result. It does not\n> now: a little scanning shows there is no way to get no_groups above 1\n> in the current logic.\n> \n> You should realize, however, that all that backend-PQexec code is dead\n> code and has been for a long time. It's unlikely that it works at all\n> any more, considering all the changes that have gone on while it sits\n> untested on the sidelines. I proposed ripping it out a few months ago,\n> and no one objected ... but I haven't got round to it yet.\n\nWhy is it dead code? Why is it no longer necessary?\n", "msg_date": "Tue, 04 Jul 2000 00:38:38 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Backend Question" }, { "msg_contents": "Chris Bitmead <[email protected]> writes:\n>> You should realize, however, that all that backend-PQexec code is dead\n>> code and has been for a long time.\n\n> Why is it dead code? Why is it no longer necessary?\n\nIt's dead code because it's not called from anywhere.\n\nOnce upon a time it was intended to be called by dynamically-loaded\nC functions, but the SPI interface superseded it a long time ago.\n\nThe only reason for not ripping it out is the faint possibility that\nsomewhere out there is someone with hoary old C-function code that\nstill does queries via the backend-PQexec interface instead of SPI.\nBut considering that backend-PQexec has been undocumented at least\nsince Berkeley days, I rather doubt it.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 10:47:07 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Backend Question " }, { "msg_contents": "Tom Lane wrote:\n\n> You should realize, however, that all that backend-PQexec code is dead\n> code and has been for a long time. It's unlikely that it works at all\n> any more, considering all the changes that have gone on while it sits\n> untested on the sidelines. I proposed ripping it out a few months ago,\n> and no one objected ... but I haven't got round to it yet.\n\nOh does it allow you to execute queries from within the backend? Sounds\nlike a good thing to resurrect at some stage.\n", "msg_date": "Tue, 04 Jul 2000 00:47:49 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Backend Question" } ]
[ { "msg_contents": "There's a broken configure test for fcntl(F_SETLK). (Hint: Of course it's\ngoing to fail if you don't include the header file that defines\nF_SETLK.) Apparently, it's been broken for quite a while so I'm afraid to\nfix it, considering the problems with file locking.\n\nReading up on TODO.detail/flock, it seems this mechanism is made obsolete\nby the pid file, so I'm inclined to remove it. At least those using Linux\nand Solaris shouldn't feel a difference because the locking code wasn't\nbeing used for said reasons. :)\n\n-- \nPeter Eisentraut Sernanders vaeg 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Mon, 3 Jul 2000 10:04:08 -0400 (EDT)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "fcntl(F_SETLK)" }, { "msg_contents": "> There's a broken configure test for fcntl(F_SETLK). (Hint: Of course it's\n> going to fail if you don't include the header file that defines\n> F_SETLK.) Apparently, it's been broken for quite a while so I'm afraid to\n> fix it, considering the problems with file locking.\n> \n> Reading up on TODO.detail/flock, it seems this mechanism is made obsolete\n> by the pid file, so I'm inclined to remove it. At least those using Linux\n> and Solaris shouldn't feel a difference because the locking code wasn't\n> being used for said reasons. :)\n\nI just applied a patch from an IA64 user that fixes that. Seems Tom\nLane thinks protecting multiple postmasters on the same port is still an\nissue. Let's hash it out and get it done, maybe.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 5 Jul 2000 12:21:22 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: fcntl(F_SETLK)" }, { "msg_contents": "At 12:21 5/07/00 -0400, Bruce Momjian wrote:\n>> There's a broken configure test for fcntl(F_SETLK). (Hint: Of course it's\n>> going to fail if you don't include the header file that defines\n>> F_SETLK.) Apparently, it's been broken for quite a while so I'm afraid to\n>> fix it, considering the problems with file locking.\n>> \n>> Reading up on TODO.detail/flock, it seems this mechanism is made obsolete\n>> by the pid file, so I'm inclined to remove it. At least those using Linux\n>> and Solaris shouldn't feel a difference because the locking code wasn't\n>> being used for said reasons. :)\n>\n>I just applied a patch from an IA64 user that fixes that. Seems Tom\n>Lane thinks protecting multiple postmasters on the same port is still an\n>issue. Let's hash it out and get it done, maybe.\n\nBuilding the current CVS sources for SuSE Linux 6.1 works, but when running\nit I get crashes:\n\nopen(\"/var/lib/pgsql7.0.2d/data/postgresql.conf\", O_RDONLY) = 3\nfstat(3, {st_mode=0, st_size=0, ...}) = 0\nbrk(0x823e000) = 0x823e000\nioctl(3, TTY_MAGIC, 0xbffff4cc) = -1 ENOTTY (Inappropriate ioctl\nfor device)\nfstat(3, {st_mode=0, st_size=0, ...}) = 0\nmmap(0, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) =\n0x4000d000\nread(3, \"#\\n# PostgreSQL configuration fi\"..., 4096) = 592\nread(3, \"\", 4096) = 0\nread(3, \"\", 4096) = 0\nioctl(3, TTY_MAGIC, 0xbffff4a0) = -1 ENOTTY (Inappropriate ioctl\nfor device)\nclose(3) = 0\nmunmap(0x4000d000, 4096) = 0\nsocket(PF_UNIX, SOCK_STREAM, 0) = 3\nopen(\"/tmp/.s.PGSQL.5432\", O_WRONLY|O_NONBLOCK) = 4\nfcntl(4, F_SETLK, {type=F_WRLCK, whence=SEEK_SET, start=0, len=0}\n<unfinished ...>\n+++ killed by SIGSEGV +++\n\nI *presume* this relates to your recent discussions re flock?\n\nBy causing the configure test to fail, the resulting code works...\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Thu, 06 Jul 2000 22:17:31 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: fcntl(F_SETLK)" }, { "msg_contents": "> >> There's a broken configure test for fcntl(F_SETLK). (Hint: Of course it's\n> >> going to fail if you don't include the header file that defines\n> >> F_SETLK.) Apparently, it's been broken for quite a while so I'm afraid to\n> >> fix it, considering the problems with file locking.\n> >> Reading up on TODO.detail/flock, it seems this mechanism is made obsolete\n> >> by the pid file, so I'm inclined to remove it. At least those using Linux\n> >> and Solaris shouldn't feel a difference because the locking code wasn't\n> >> being used for said reasons. :)\n> I *presume* this relates to your recent discussions re flock?\n> By causing the configure test to fail, the resulting code works...\n\nCurrent CVS sources are hard-broken for Linux :((\n\nOn my laptop with a 2.2.14 kernel, postmaster exits with a segfault, and\ngdb fails to keep ahold of the image when running:\n\n Program terminated with signal SIGSEGV, Segmentation fault.\n The program no longer exists.\n\nwhich will make it difficult for me to debug.\n\nThe tree worked yesterday. Philip, what exactly did you do to \"cause the\nconfigure test to fail\"? Can I just go in a muck with Makefile.global?\n\nI'd like to get up and running so I can finish testing and submit some\npatches!\n\n - Thomas\n", "msg_date": "Thu, 06 Jul 2000 13:41:52 +0000", "msg_from": "Thomas Lockhart <[email protected]>", "msg_from_op": false, "msg_subject": "Re: fcntl(F_SETLK)" }, { "msg_contents": "At 13:41 6/07/00 +0000, Thomas Lockhart wrote:\n>\n>The tree worked yesterday. Philip, what exactly did you do to \"cause the\n>configure test to fail\"? Can I just go in a muck with Makefile.global?\n>\n\nI tried that, but it seemed not to work - rerunning configure seems to be\nnecessary. The solution I adopted is quite embarrasing, so if anyone wants\nto give me a lesson in the best way to do this, feel free!\n\nIn 'configure' you will find a section:\n\necho $ac_n \"checking for fcntl(F_SETLK)\"\"... $ac_c\" 1>&6\necho \"configure:5598: checking for fcntl(F_SETLK)\" >&5\ncat > conftest.$ac_ext <<EOF\n#line 5600 \"configure\"\n#include \"confdefs.h\"\n#include <tdio.h>\n#include <fcntl.h>\n\nI changed the last two lines to:\n\n#include <pjwpjwstdio.h>\n#include <pjwpjwfcntl.h>\n\n(ie. two file names that I knew did not exist) - hence the test will fail.\nIt'd probably work if you just deleted them, but I like to leave\nfootprints, at least...\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Thu, 06 Jul 2000 23:49:43 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: fcntl(F_SETLK)" }, { "msg_contents": "Thomas Lockhart <[email protected]> writes:\n> Current CVS sources are hard-broken for Linux :((\n\nAs a workaround, undef HAVE_FCNTL_SETLK in config.h before building.\n\nI think what we'd better do is turn the does-it-compile configure test\nfor fcntl() into a full-blown does-it-execute test... ain't working\naround OS bugs fun?\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 06 Jul 2000 17:43:05 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: fcntl(F_SETLK) " }, { "msg_contents": "> > Current CVS sources are hard-broken for Linux :((\n> As a workaround, undef HAVE_FCNTL_SETLK in config.h before building.\n> I think what we'd better do is turn the does-it-compile configure test\n> for fcntl() into a full-blown does-it-execute test... ain't working\n> around OS bugs fun?\n\nPer your recent go-round, \"that ain't a bug, that's a misunderstanding\"\n;)\n\n - Thomas\n", "msg_date": "Fri, 07 Jul 2000 05:23:21 +0000", "msg_from": "Thomas Lockhart <[email protected]>", "msg_from_op": false, "msg_subject": "Re: fcntl(F_SETLK)" } ]
[ { "msg_contents": "\nWhat was the reason that PQprint was abandoned and psql went in its own\ndirection?\n\n-- \nChris Bitmead\nmailto:[email protected]\n", "msg_date": "Tue, 04 Jul 2000 00:19:31 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": true, "msg_subject": "PQprint" }, { "msg_contents": "On Tue, 4 Jul 2000, Chris Bitmead wrote:\n\n> What was the reason that PQprint was abandoned and psql went in its own\n> direction?\n\nThe printing routines had to be rewritten, in particular to make them more\nmodular. The options were to add a fourth(!) generation of printing\nroutines to libpq and live with them until the end of days, or make them\ninternal to psql and don't tell anyone when changes happen.\n\nIf you want to use psql's printing routines you can just use print.c with\nno changes. I've made an effort to keep it independent of psql.\n\n-- \nPeter Eisentraut Sernanders vaeg 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Mon, 3 Jul 2000 10:33:56 -0400 (EDT)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: PQprint" } ]
[ { "msg_contents": "\nMy work on pg_dump/pg_restore has left me wondering about also supporting\nBLOBs. I realize that TOAST will replace BLOBs, but the ability to dump &\nrestore a DB with BLOBs will make conversion a lot easier, I would guess.\n\nThe question then is, for the purpose of pg_dump/restore, how difficult\nwould it be to create a 'lo_create_with_oid' function?\n\nThis technique would obviously only work properly for full restores of the\ndatabase, and if the max oid was set prior to the restore.\n\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 04 Jul 2000 00:29:55 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Specifying OID for BLOBs?" } ]
[ { "msg_contents": "Greetings all,\n\nI'm sending this to -announce, -hackers, and -general; apologies for\nthe cross-post. Replies should automatically go to -general, which\nis the best forum for this discussion. But I wanted to make sure\nthe largest possible audience of PostgreSQL users had a chance to\ncomment...\n\nSeveral weeks ago, we announced the formation of a new company\ncalled Great Bridge, which will professionally market and support\nopen source software solutions based on PostgreSQL. As we said at\nthe time, we've been working since late last year to understand the\nsoftware itself, the needs of business customers who might be\ninclined to use PostgreSQL, and of course the dynamics of the\nPostgreSQL development community. Our team met with the six members\nof the PostgreSQL core steering group in March, and had a good and\nopen exchange about each other's plans for the future.\n\nOne issue that has always been a source of uncertainty - I think for\nall of us - has been the license under which PostgreSQL is\ndistributed. As we've said publicly on Slashdot, ZDNet, and other\nforums, we're big fans of the current Berkeley license; we find it\nmore \"open\" than other open source licenses, in the sense that the\nuser/hacker has almost total freedom as to what he wants to do with\nthe code. We've also found, through some rather extensive market\nresearch, that the business community (to which we'll be selling\nproducts and services) vastly prefers it over GPL, or hybrids like\nMozilla, etc. I don't want to re-start that debate here - the\nconsensus in the PostgreSQL community over the past few years seems\nto be that the Berkeley style license is best suited for the\ncontinued development of PostgreSQL.\n\nWhat we'd like to propose is a general tightening up of what the\nexisting license is *supposed* to be doing in the first place -\nprotecting the developers who worked on the code, and ensuring that\nthe code stays open source in perpetuity. Rusty Friddell, the\ngeneral counsel of Great Bridge's parent company Landmark\nCommunications, explained our views on this to the core group in\nMarch, and they expressed an interest in our going ahead with some\nresearch on how the license might be improved. So Great Bridge went\nahead and engaged two outside law firms to work on it, and we now\nhave something we're ready to present to the community. We've\nincluded the core group in early looks at some of the thinking, and\nwe all feel it's now ready for a full-fledged debate in the broader\nhacker world.\n\nI'm including below two things - 1) a note from Rusty explaining a\nbit more about what we're trying to do, and 2) the proposed text of\nthe license itself. I'll monitor the discussion on the -general\nlist, and if anyone has any specific questions for Rusty, I'll be\nhappy to channel them to him.\n\nThanks,\n\nNed Lilly\nVP Hacker Relations\nGreat Bridge, LLC\n\n-----\n\ntext of note from Rusty Friddell, general counsel of Landmark\nCommunications, Inc.:\n\nWhat follows is a suggested addition to the existing Berkeley\nlicense governing the use of PostgreSQL. The changes are suggested\nmainly to address and deal with the many contributions by the hacker\ncommunity to the work of the original licensor, the University of\nCalifornia. Unlike other open source licenses (GNU, Mozilla,\nInterbase), the original Berkeley license does not take into account\nthat over time a lot of different individual developers and perhaps\nsome corporate contributors, would have individual copyrights on\nsubstantial portions of the code.\n\nThis deficiency has two adverse affects. First, the contributing\ndevelopers are not afforded the protection of the exculpatory\nlanguage in \"bold face.\" Certainly, having given of their time and\ncreativity without compensation, the developers should be extended\nthis coverage as clearly as possible. Second, and admittedly of\nsignificant importance to Great Bridge, the commercial proliferation\nof PostgreSQL could be hindered if business users are concerned that\nthe license might not cover the substantial additions and\nimprovements made to the code over the last few years.\n\nIn developing the new language, the resources of two intellectual\nproperty law firms, one East Coast and one West Coast, were tapped.\nNo less important was my education by the core group as to what was\ngood about the existing license. As a result of the former, you\nwill see some minor language clean up. From the latter, you'll note\na pretty strict adherence to the \"short and sweet\" approach of the\noriginal Berkeley license, particularly as compared to the\nlawyer-friendly (that may be an oxymoron) GPL and Mozilla licenses.\n\nNo discussion of this type should be without controversy, so I throw\nyou the following red meat: the choice of state law has been\nselected to cause the application of the Uniform Computer\nInformation Transactions Act (UCITA) to the usage of the software.\n\n(Pause for outrage to subside.)\n\nNow, I ask that you suspend your initial reaction for a moment and\nconsider the following. As to this license, you are not on the\nreceiving end of an attempted screwing at the hands of some evil\nempire desiring to take advantage of defenseless consumers. You are\nthe volunteer creators and improvers of this product who, without\nremuneration, are providing sophisticated business users with an\nalternative to proprietary database software.\n\nI submit that your efforts should be exempted from any potential\nliability. The original Berkeley language sought that result, and\nwould likely suffice in most states under most circumstances. The\napplication of UCITA simply ensures the result that the original\nlanguage attempts to achieve. Much of the rest of the UCITA is\nsimply inapplicable to this product - the risk that big brother will\nreach into your computer and remove Postgres for failure to pay\nlicense fees is simply not present where we're dealing with a free\nproduct. End of manifesto.\n\nTwo states have adopted UCITA - Virginia and Maryland. Maryland has\nan October 1, 2000, effective date, but requires that its laws will\nonly apply if there is a reasonable connection with the state.\nVirginia has an effective date of July 1, 2001, but does not require\na connection with the state and thereby gives somewhat greater\nassurance that UCITA will apply to all Postgres-related dealings,\nwherever they occur. The fact that Great Bridge is based in\nVirginia is really a complete coincidence.\n\nThe revised license also anticipates a concern which many people\nraised in a comment to the Great Bridge announcement, and which came\nup again in the recent Berkeley DB discussion. It was the mutual\ndetermination of Great Bridge and the PGSQL core group that, while\nnone of us wanted to see the software taken private, an attempt to\nlegislate, through the license, that every modification anyone made\nhad to be contributed back to the public use, would be poorly\nreceived and would possibly discourage the use by businesses who\nmight want to make certain improvements for their internal use\nonly. Accordingly, the compromise is that no such \"poison pill\"\ntype language has been added to the original Berkeley text, although\nwe have provided that once a contribution to the PostgreSQL project\nhas been voluntarily made, it is subject to this license and is\nirrevocable - which seems to provide the open-source perpetuity that\neveryone is looking for.\n\n-----\n\n[To be integrated with the software in such a way that this license\nmust be seen before downloading can occur]\n\n\nPostgreSQL Data Base Management System (formerly known as\nPostgres95)\n\nThis directory contains the _______ release of PostgreSQL, as well\nas various post-release patches in the patches directory. See\nINSTALL for the installation notes and HISTORY for the changes.\n\nWe also have a WWW home page located at: http://www.postgreSQL.org\n\n-------------------------\n\nPostgreSQL is not public domain software. It is copyrighted by the\nUniversity of California but may be used according to the following\nlicensing terms:\n\nPOSTGRES95 Data Base Management System (formerly known as Postgres,\nthen as Postgres95).\n\nCopyright (c) 1994-8 Regents of the University of California\n\nPermission to use, copy, modify, and distribute this software and\nits documentation for any purpose, without fee, and without a\nwritten agreement is hereby granted, provided that the above\ncopyright notice and this paragraph and the following two paragraphs\nappear in all copies.\n\nIN NO EVENT SHALL THE UNIVERSITY OF CALIFORNIA BE LIABLE TO ANY\nPARTY FOR DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL\nDAMAGES, INCLUDING LOST PROFITS, ARISING OUT OF THE USE OF THIS\nSOFTWARE AND ITS DOCUMENTATION, EVEN IF THE UNIVERSITY OF CALIFORNIA\nHAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n\nTHE UNIVERSITY OF CALIFORNIA SPECIFICALLY DISCLAIMS ANY WARRANTIES,\nINCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF\nMERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE\nPROVIDED HEREUNDER IS ON AN \"AS IS\" BASIS, AND THE UNIVERSITY OF\nCALIFORNIA HAS NO OBLIGATIONS TO PROVIDE MAINTENANCE, SUPPORT,\nUPDATES, ENHANCEMENTS, OR MODIFICATIONS.\n\n-------------------------\n\nCopyright ( 1996, 1997, 1998, 1999, 2000 by various contributors (as\nidentified in HISTORY) (collectively \"Developers\") which may be used\naccording to the following licensing terms:\n\nWorldwide permission to use, copy, modify, and distribute this\nsoftware and its documentation for any purpose, without fee, and\nwithout a written agreement is hereby granted, on a non-exclusive\nbasis, provided that the above copyright notice, this paragraph and\nthe following paragraphs appear in all copies:\n\nAny person who contributes or submits any modification or other\nchange to the PostgreSQL software or documentation grants\nirrevocable, non-exclusive, worldwide permission, without charge, to\nuse, copy, further modify and distribute the same under the terms of\nthis license.\n\nIN NO EVENT SHALL ANY DEVELOPER BE LIABLE TO ANY PARTY FOR DIRECT,\nINDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING,\nWITHOUT LIMITATION, LOST PROFITS, ARISING OUT OF THE USE OF THIS\nSOFTWARE AND ITS DOCUMENTATION, EVEN IF THE DEVELOPER HAS BEEN\nADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n\nTHE DEVELOPERS SPECIFICALLY DISCLAIM ALL WARRANTIES, EXPRESS OR\nIMPLIED INCLUDING BUT NOT LIMITED TO IMPLIED WARRANTIES OF\nMERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, NEED, OR QUALITY,\nAND ANY IMPLIED WARRANTY FROM COURSE OF DEALING OR USAGE OF TRADE.\nIN ADDITION, THERE IS NO IMPLIED WARRANTY AGAINST INTERFERENCE WITH\nENJOYMENT OR AGAINST INFRINGEMENT. THE SOFTWARE AND DOCUMENTATION\nPROVIDED HEREUNDER IS ON AN \"AS IS\" BASIS. NO DEVELOPER HAS ANY\nOBLIGATION TO PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS OR\nMODIFICATIONS TO OR FOR THE SOFTWARE OR DOCUMENTATION.\n\nThe foregoing shall be governed by and construed under the laws of\nthe State of Virginia.\n\nBY USING THIS SOFTWARE YOU AGREE TO THESE TERMS AND CONDITIONS. IF\nYOU DO NOT AGREE TO THESE TERMS AND CONDITIONS, YOU SHOULD NOT USE\nTHIS SOFTWARE.\n\n", "msg_date": "Mon, 03 Jul 2000 11:09:21 -0400", "msg_from": "Ned Lilly <[email protected]>", "msg_from_op": true, "msg_subject": "proposed improvements to PostgreSQL license" }, { "msg_contents": "Ned Lilly writes:\n\n> What we'd like to propose is a general tightening up of what the\n> existing license is *supposed* to be doing in the first place -\n\nIn order to tighten up the license you'd need to get every developer past,\npresent, and future to sign paperwork that they agree to this change.\n\n> protecting the developers who worked on the code,\n\nI've said it before and I'll say it again: If you, as a developer, want to\n\"protect\" yourself, talk to a lawyer licensed in your jurisdiction, and\nnot \"one East coast\" and \"one West coast\". And don't trust any untested\nlicenses based on a law that will potentially be overturned in the Supreme\nCourt anyway.\n\n> and ensuring that the code stays open source in perpetuity.\n\nNo, that's what the GPL does.\n\n> Second, and admittedly of significant importance to Great Bridge, the\n> commercial proliferation of PostgreSQL could be hindered if business\n> users are concerned that the license might not cover the substantial\n> additions and improvements made to the code over the last few years.\n\nHow can you guarantee them that none of the code now is under a different\nlicense (because it was ripped off somewhere) or under patent\nrestrictions? Or for that matter, how can you guarantee it in the future?\nWhat do you say to anyone that comes along and says \"This is my code, it\nshouldn't be here\"? Will you go back to the logs and prove that he\n\"contributed or submitted\" it and is therefore bound to the license?\n\nChanging the license wording might satisfy the suits, but it doesn't\nchange reality a bit. It just creates more legal uncertainty for users at\nthat point. The least thing we need is Yet Another Open Source License.\n\nTo my knowledge, the BSD license has been used in one form or another for\nat least 20 years and neither has any contributor ever been sued for\nliability, nor was there any court case that concluded that the BSD\nlicense is worth anything at all, nor has the developer or commercial\nacceptance of any product ever been affected by this \"untight\" license.\n\n> [To be integrated with the software in such a way that this license\n> must be seen before downloading can occur]\n\nThat's funny...\n\n> The foregoing shall be governed by and construed under the laws of\n> the State of Virginia.\n\nThe recurring theme throughout this email was that Great Bridge has\napparently not appreciated that PostgreSQL land extends beyond the borders\nof the U.S. of A. Maybe your 32 focus groups in major U.S. cities wanted\nthe license changed like this, but I'll bet lunch that 32 out of 32 focus\ngroups in major European cities will look with extreme suspicion at\nanything with \"laws of the State of XXX\" attached to it.\n\nUntil they realize that the laws of Virginia don't apply to them. Or to\nCanada, where hub.org is located these days.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Mon, 3 Jul 2000 22:49:54 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "Ned Lilly <[email protected]> writes:\n\n\n> We've also found, through some rather extensive market\n> research, that the business community (to which we'll be selling\n> products and services) vastly prefers it over GPL, or hybrids like\n> Mozilla, etc. \n\nThat depends on what your market is - for businesses who wants to be\nable to hide source, yes. For businesses who use it, being sure the\nsource is available is the best - which the GPL guarantees. BSD gives\nthe middle man more freedom to screw the end user ;)\n\n> What we'd like to propose is a general tightening up of what the\n> existing license is *supposed* to be doing in the first place -\n> protecting the developers who worked on the code, and ensuring that\n> the code stays open source in perpetuity.\n\nGPL would solve this - the main advantage of BSDish licenses is you\ncan go closed source if you want to.\n\nNow, I don't advocate a change in license - my main consern is \"there\nare enough licenses in the world\". I think the \"each package one\nlicense\" is a bad trend.\n\n-- \nTrond Eivind Glomsr�d\nRed Hat, Inc.\n", "msg_date": "03 Jul 2000 17:51:46 -0400", "msg_from": "[email protected] (Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?=)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "\nI think this is a bad idea for the following reasons:\n\n1) It is trying to be a GPL in what it is trying to achieve without\nactually being well thought out. Any person who \"submits\" modifications\nmust do so under the same licence. Submits to what or whom?\n\n2) If the core team want to make sure modifications to the software are\nunder the same licence then they should merely insist that any patches\nare accompanied by that same licence (i.e. the current licence). End of\nstory end of problem. If you want to go any further than that you may as\nwell go GPL.\n\n3) You talk about how wonderful the BSD licence is, then you really\nchange the whole meaning of that licence.\n\n4) What is this stuff about \"tightening up of what the existing licence\nis supposed to do\"? What do you think it is supposed to do? I think it\nis basicly an annoying artifact of UCB's legal team that happens to make\nthe software virtually public domain. We might just as well get rid of\nall licences except that we're not allowed.\n\n5) This \"protection\" for developers is a straw-man. I don't see, say the\nfree-bsd developers worried about this. If Great bridge wants to\ndistribute with extra disclaimers then go ahead.\n\n6) This is a very US-centric view of the world. Most of the developers\nare not in the US if the postgresql.org home page is correct. We don't\ncare about the stinkin UCITA, we are not bound by and don't care about\nanything the State of Virginia may or may not say.\n\n7) I hope you're not thinking of bloating each and every source file\nwith all that legalese.\n\n8) \"To be integrated with the software in such a way that this license\nmust be seen before downloading can occur\".\nUmm, can all the laywers please just butt out? Every other open-source\npackage in the universe just relies on a licence file in the home\ndirectory. You going to try and stop people downloading with clicking a\nlicence agreement? How you going to handle mirrors? Or are you not\ngoing to mirror any more? What about Red Hat el al?\n\nPoint (8) makes me thing that this whole thing is the recommendation of\nsome lawyer who is totally out of touch with the free software community\nbut feels compelled to add a whole lot of disclaimers and so-forth\nbecause that's his job. Bottom line is it's not broke so leave it alone.\n\n\nNed Lilly wrote:\n> \n> Greetings all,\n> \n> I'm sending this to -announce, -hackers, and -general; apologies for\n> the cross-post. Replies should automatically go to -general, which\n> is the best forum for this discussion. But I wanted to make sure\n> the largest possible audience of PostgreSQL users had a chance to\n> comment...\n> \n> Several weeks ago, we announced the formation of a new company\n> called Great Bridge, which will professionally market and support\n> open source software solutions based on PostgreSQL. As we said at\n> the time, we've been working since late last year to understand the\n> software itself, the needs of business customers who might be\n> inclined to use PostgreSQL, and of course the dynamics of the\n> PostgreSQL development community. Our team met with the six members\n> of the PostgreSQL core steering group in March, and had a good and\n> open exchange about each other's plans for the future.\n> \n> One issue that has always been a source of uncertainty - I think for\n> all of us - has been the license under which PostgreSQL is\n> distributed. As we've said publicly on Slashdot, ZDNet, and other\n> forums, we're big fans of the current Berkeley license; we find it\n> more \"open\" than other open source licenses, in the sense that the\n> user/hacker has almost total freedom as to what he wants to do with\n> the code. We've also found, through some rather extensive market\n> research, that the business community (to which we'll be selling\n> products and services) vastly prefers it over GPL, or hybrids like\n> Mozilla, etc. I don't want to re-start that debate here - the\n> consensus in the PostgreSQL community over the past few years seems\n> to be that the Berkeley style license is best suited for the\n> continued development of PostgreSQL.\n> \n> What we'd like to propose is a general tightening up of what the\n> existing license is *supposed* to be doing in the first place -\n> protecting the developers who worked on the code, and ensuring that\n> the code stays open source in perpetuity. Rusty Friddell, the\n> general counsel of Great Bridge's parent company Landmark\n> Communications, explained our views on this to the core group in\n> March, and they expressed an interest in our going ahead with some\n> research on how the license might be improved. So Great Bridge went\n> ahead and engaged two outside law firms to work on it, and we now\n> have something we're ready to present to the community. We've\n> included the core group in early looks at some of the thinking, and\n> we all feel it's now ready for a full-fledged debate in the broader\n> hacker world.\n> \n> I'm including below two things - 1) a note from Rusty explaining a\n> bit more about what we're trying to do, and 2) the proposed text of\n> the license itself. I'll monitor the discussion on the -general\n> list, and if anyone has any specific questions for Rusty, I'll be\n> happy to channel them to him.\n> \n> Thanks,\n> \n> Ned Lilly\n> VP Hacker Relations\n> Great Bridge, LLC\n> \n> -----\n> \n> text of note from Rusty Friddell, general counsel of Landmark\n> Communications, Inc.:\n> \n> What follows is a suggested addition to the existing Berkeley\n> license governing the use of PostgreSQL. The changes are suggested\n> mainly to address and deal with the many contributions by the hacker\n> community to the work of the original licensor, the University of\n> California. Unlike other open source licenses (GNU, Mozilla,\n> Interbase), the original Berkeley license does not take into account\n> that over time a lot of different individual developers and perhaps\n> some corporate contributors, would have individual copyrights on\n> substantial portions of the code.\n> \n> This deficiency has two adverse affects. First, the contributing\n> developers are not afforded the protection of the exculpatory\n> language in \"bold face.\" Certainly, having given of their time and\n> creativity without compensation, the developers should be extended\n> this coverage as clearly as possible. Second, and admittedly of\n> significant importance to Great Bridge, the commercial proliferation\n> of PostgreSQL could be hindered if business users are concerned that\n> the license might not cover the substantial additions and\n> improvements made to the code over the last few years.\n> \n> In developing the new language, the resources of two intellectual\n> property law firms, one East Coast and one West Coast, were tapped.\n> No less important was my education by the core group as to what was\n> good about the existing license. As a result of the former, you\n> will see some minor language clean up. From the latter, you'll note\n> a pretty strict adherence to the \"short and sweet\" approach of the\n> original Berkeley license, particularly as compared to the\n> lawyer-friendly (that may be an oxymoron) GPL and Mozilla licenses.\n> \n> No discussion of this type should be without controversy, so I throw\n> you the following red meat: the choice of state law has been\n> selected to cause the application of the Uniform Computer\n> Information Transactions Act (UCITA) to the usage of the software.\n> \n> (Pause for outrage to subside.)\n> \n> Now, I ask that you suspend your initial reaction for a moment and\n> consider the following. As to this license, you are not on the\n> receiving end of an attempted screwing at the hands of some evil\n> empire desiring to take advantage of defenseless consumers. You are\n> the volunteer creators and improvers of this product who, without\n> remuneration, are providing sophisticated business users with an\n> alternative to proprietary database software.\n> \n> I submit that your efforts should be exempted from any potential\n> liability. The original Berkeley language sought that result, and\n> would likely suffice in most states under most circumstances. The\n> application of UCITA simply ensures the result that the original\n> language attempts to achieve. Much of the rest of the UCITA is\n> simply inapplicable to this product - the risk that big brother will\n> reach into your computer and remove Postgres for failure to pay\n> license fees is simply not present where we're dealing with a free\n> product. End of manifesto.\n> \n> Two states have adopted UCITA - Virginia and Maryland. Maryland has\n> an October 1, 2000, effective date, but requires that its laws will\n> only apply if there is a reasonable connection with the state.\n> Virginia has an effective date of July 1, 2001, but does not require\n> a connection with the state and thereby gives somewhat greater\n> assurance that UCITA will apply to all Postgres-related dealings,\n> wherever they occur. The fact that Great Bridge is based in\n> Virginia is really a complete coincidence.\n> \n> The revised license also anticipates a concern which many people\n> raised in a comment to the Great Bridge announcement, and which came\n> up again in the recent Berkeley DB discussion. It was the mutual\n> determination of Great Bridge and the PGSQL core group that, while\n> none of us wanted to see the software taken private, an attempt to\n> legislate, through the license, that every modification anyone made\n> had to be contributed back to the public use, would be poorly\n> received and would possibly discourage the use by businesses who\n> might want to make certain improvements for their internal use\n> only. Accordingly, the compromise is that no such \"poison pill\"\n> type language has been added to the original Berkeley text, although\n> we have provided that once a contribution to the PostgreSQL project\n> has been voluntarily made, it is subject to this license and is\n> irrevocable - which seems to provide the open-source perpetuity that\n> everyone is looking for.\n> \n> -----\n> \n> [To be integrated with the software in such a way that this license\n> must be seen before downloading can occur]\n> \n> PostgreSQL Data Base Management System (formerly known as\n> Postgres95)\n> \n> This directory contains the _______ release of PostgreSQL, as well\n> as various post-release patches in the patches directory. See\n> INSTALL for the installation notes and HISTORY for the changes.\n> \n> We also have a WWW home page located at: http://www.postgreSQL.org\n> \n> -------------------------\n> \n> PostgreSQL is not public domain software. It is copyrighted by the\n> University of California but may be used according to the following\n> licensing terms:\n> \n> POSTGRES95 Data Base Management System (formerly known as Postgres,\n> then as Postgres95).\n> \n> Copyright (c) 1994-8 Regents of the University of California\n> \n> Permission to use, copy, modify, and distribute this software and\n> its documentation for any purpose, without fee, and without a\n> written agreement is hereby granted, provided that the above\n> copyright notice and this paragraph and the following two paragraphs\n> appear in all copies.\n> \n> IN NO EVENT SHALL THE UNIVERSITY OF CALIFORNIA BE LIABLE TO ANY\n> PARTY FOR DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL\n> DAMAGES, INCLUDING LOST PROFITS, ARISING OUT OF THE USE OF THIS\n> SOFTWARE AND ITS DOCUMENTATION, EVEN IF THE UNIVERSITY OF CALIFORNIA\n> HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n> \n> THE UNIVERSITY OF CALIFORNIA SPECIFICALLY DISCLAIMS ANY WARRANTIES,\n> INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF\n> MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE\n> PROVIDED HEREUNDER IS ON AN \"AS IS\" BASIS, AND THE UNIVERSITY OF\n> CALIFORNIA HAS NO OBLIGATIONS TO PROVIDE MAINTENANCE, SUPPORT,\n> UPDATES, ENHANCEMENTS, OR MODIFICATIONS.\n> \n> -------------------------\n> \n> Copyright ( 1996, 1997, 1998, 1999, 2000 by various contributors (as\n> identified in HISTORY) (collectively \"Developers\") which may be used\n> according to the following licensing terms:\n> \n> Worldwide permission to use, copy, modify, and distribute this\n> software and its documentation for any purpose, without fee, and\n> without a written agreement is hereby granted, on a non-exclusive\n> basis, provided that the above copyright notice, this paragraph and\n> the following paragraphs appear in all copies:\n> \n> Any person who contributes or submits any modification or other\n> change to the PostgreSQL software or documentation grants\n> irrevocable, non-exclusive, worldwide permission, without charge, to\n> use, copy, further modify and distribute the same under the terms of\n> this license.\n> \n> IN NO EVENT SHALL ANY DEVELOPER BE LIABLE TO ANY PARTY FOR DIRECT,\n> INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING,\n> WITHOUT LIMITATION, LOST PROFITS, ARISING OUT OF THE USE OF THIS\n> SOFTWARE AND ITS DOCUMENTATION, EVEN IF THE DEVELOPER HAS BEEN\n> ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n> \n> THE DEVELOPERS SPECIFICALLY DISCLAIM ALL WARRANTIES, EXPRESS OR\n> IMPLIED INCLUDING BUT NOT LIMITED TO IMPLIED WARRANTIES OF\n> MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, NEED, OR QUALITY,\n> AND ANY IMPLIED WARRANTY FROM COURSE OF DEALING OR USAGE OF TRADE.\n> IN ADDITION, THERE IS NO IMPLIED WARRANTY AGAINST INTERFERENCE WITH\n> ENJOYMENT OR AGAINST INFRINGEMENT. THE SOFTWARE AND DOCUMENTATION\n> PROVIDED HEREUNDER IS ON AN \"AS IS\" BASIS. NO DEVELOPER HAS ANY\n> OBLIGATION TO PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS OR\n> MODIFICATIONS TO OR FOR THE SOFTWARE OR DOCUMENTATION.\n> \n> The foregoing shall be governed by and construed under the laws of\n> the State of Virginia.\n> \n> BY USING THIS SOFTWARE YOU AGREE TO THESE TERMS AND CONDITIONS. IF\n> YOU DO NOT AGREE TO THESE TERMS AND CONDITIONS, YOU SHOULD NOT USE\n> THIS SOFTWARE.\n", "msg_date": "Tue, 04 Jul 2000 11:13:55 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "\"We\" (the Postgres steering committee) have discussed these issues for\nmonths and months. We aren't trying to change anything, just reinforce\nwhat we believe to be already the case. However, the path to do this\nisn't perfectly clear to anyone; this is the first concrete proposal we\nhave had which does try to address the issues we believe are already\nhere whether we want them or not.\n\nI'll bring them up farther down (and will probably forget and leave some\nrelevant pieces out).\n\n> I think this is a bad idea for the following reasons:\n> 1) It is trying to be a GPL in what it is trying to achieve without\n> actually being well thought out. Any person who \"submits\" modifications\n> must do so under the same licence. Submits to what or whom?\n\nIt is *not* trying to be GPL. It is trying to be BSD, while extending\nliability protection to the current cast of developers, who are (I'm\npretty sure) not covered in any of the wording of the UCB-generated\nlicense.\n\n> 2) If the core team want to make sure modifications to the software are\n> under the same licence then they should merely insist that any patches\n> are accompanied by that same licence (i.e. the current licence). End of\n> story end of problem. If you want to go any further than that you may as\n> well go GPL.\n\nThe current license asks users to absolve the University of California\nof any liability involving use of the Postgres source code. It does not\n(currently) explicitly ask the same on behalf of the current developers\n(including yourself ;)\n\n> 3) You talk about how wonderful the BSD licence is, then you really\n> change the whole meaning of that licence.\n\nHow?\n\n> 4) What is this stuff about \"tightening up of what the existing licence\n> is supposed to do\"? What do you think it is supposed to do? I think it\n> is basicly an annoying artifact of UCB's legal team that happens to make\n> the software virtually public domain. We might just as well get rid of\n> all licences except that we're not allowed.\n\nI disagree, though we don't know UC's motivations for sure. imho the BSD\nlicense is intended to protect UC from \"deep pockets\" lawsuits, while\npreserving some credit for the original design team and the institution\nwhich made it possible.\n\nThe new wording is intended to continue to do exactly that, extending\nthe umbrella to cover developers with no connection to UC.\n\n> 5) This \"protection\" for developers is a straw-man. I don't see, say the\n> free-bsd developers worried about this. If Great bridge wants to\n> distribute with extra disclaimers then go ahead.\n\nIt is being proposed as an addition to the Postgres development effort.\nI'm sure that GB knows they could add anything they want to their own\nproduct.\n\n> 6) This is a very US-centric view of the world. Most of the developers\n> are not in the US if the postgresql.org home page is correct. We don't\n> care about the stinkin UCITA, we are not bound by and don't care about\n> anything the State of Virginia may or may not say.\n\nGood point. But the USA is the demon spawning ground for lawyers, and is\nat the leading edge of aggressive new legal territory. That may change\neventually, but since 90+% of our federal legislative representatives\nare lawyers (stats from memory, but it is a *high* number), that may not\nchange very quickly :(\n\n> 7) I hope you're not thinking of bloating each and every source file\n> with all that legalese.\n\nNo, afaik that is not considered necessary.\n\n> 8) \"To be integrated with the software in such a way that this license\n> must be seen before downloading can occur\".\n> Umm, can all the laywers please just butt out? Every other open-source\n> package in the universe just relies on a licence file in the home\n> directory. You going to try and stop people downloading with clicking a\n> licence agreement? How you going to handle mirrors? Or are you not\n> going to mirror any more? What about Red Hat el al?\n\nGood point. Not exactly sure why this was suggested, but the American\ncourts are *full* of cases where the plaintif said that they \"didn't\nreally know\" something that should have been obvious.\n\n> Point (8) makes me thing that this whole thing is the recommendation of\n> some lawyer who is totally out of touch with the free software community\n> but feels compelled to add a whole lot of disclaimers and so-forth\n> because that's his job. Bottom line is it's not broke so leave it alone.\n\nafaik \"it's not broken\" is true, for the free software community. And\npart of my pleasure in contributing to Postgres is exactly because of\nthat general distain for legaleze and idiot-speak commercial agreements.\n\nPostgres is starting to become a visible thing, and is going to be used\nby people who don't know much about the free software movement. And\n*I'm* within reach of the American court system, and *you* can\ncontribute code which could make me a target for a lawsuit. I'd rather\nshort-circuit that before the lawsuit, rather than asking for a donation\nfor my defense ;)\n\nSo the intent was, as stated, to *reinforce* what we already believe to\nbe true (including yourself). The recently-enacted UCITA law was\n(afaict) intended to protect, perhaps wrongly imho, commercial software\ncompanies from liability claims (I know that Oracle *claims* a whole lot\nmore for 8i than we do for Postgres, so why shouldn't they be held\naccountable for what they claim?). But UCITA is a sharp tool which we\ncan use to protect volunteer software developers such as myself, and\nyou.\n\nI (and *all* of the steering committee) had pretty much the same\nreaction as you did at first. But some of us are closer to the US legal\nsystem, and see what silliness it can generate, so came around to\nthinking that there was something to be gained by license additions.\n\nRegards.\n\n - Thomas\n", "msg_date": "Tue, 04 Jul 2000 02:10:05 +0000", "msg_from": "Thomas Lockhart <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": ">One issue that has always been a source of uncertainty - I think for\n>all of us - has been the license under which PostgreSQL is\n>distributed.\n\nOnly an issue of uncertainty when people talk about changing it.\n\nAs a company who wants PostgreSQL to remain in the public domain, I would\nprefer to see it go GPL; this effectively prevents another company coming\nalong and swallowing the major developers as a means of stifling further\ndevelopment. This latter tactic would probably never work, but it would be\nextermely disruptive.\n\nAs a developer who likes to be able to make a buck from their work, I like\nthe idea of licencing my code for free use within, and only within, the\nfree, open-source version of PostgreSQL.\n\nIt seems to me that if you want to do all that you say, this is probably\nthe way to go.\n\n\n>We've also found, through some rather extensive market\n>research\n\nOut of curiosity, who with & where?\n\n\n>I don't want to re-start that debate here\n\nSorry.\n\n> - the\n>consensus in the PostgreSQL community over the past few years seems\n>to be that the Berkeley style license is best suited for the\n>continued development of PostgreSQL.\n\nNews to me. Perhaps an informed survey of the current PostgreSQL user base\nmight be in order (perhaps each party - BSD, GPL, and other - can put their\ncase in 500 words or less?). I would be very interested to know the\noutcome. It seems to me that the camps are highly divided, perhaps the\nmajority prefer BSD, but I suspect that they all agree that the status-quo\nis acceptable. Once you seek to change the agreement to another\nnon-standard agreement, I think you will be opening a can of worms. You\nwill at least have to persuade all past contribitors that your new version\nis better than their favorite model and BSD.\n\n\n>protecting the developers who worked on the code, and ensuring that\n>the code stays open source in perpetuity.\n\nNo; this is what companies seeking to develop private versions of the\nsoftware want. As it stands, all developers who have contributed to PG over\nthe years still own the copyright of their code, and can withold the\nlicense to use it. Unless they have assigned that copyright somewhere else,\nor signed the sort of agreement you propose. Releasing code to the PG\ncommunity provides an implied license to the PG community. It does not\nnecessarily grant rights to others, unless specifically stated in the code.\nThis is why you have:\n\n\"Any person who contributes or submits any modification or other\nchange to the PostgreSQL software or documentation grants\nirrevocable, non-exclusive, worldwide permission, without charge, to\nuse, copy, further modify and distribute the same under the terms of\nthis license\"\n\nwhich just dimishes the rights of the developer make a buck if someone else\nuses their code as part of a commercial version of PG.\n\nI have just submitted a chunk of code to the project, and the notice in the\ncode says, basically, it is free to be used in any way. But I would have\n*much* preferred it to have been under the GPL; then when someone fixes my\ncode, or improves it, I know that I will get to see (and use) the changes.\n\n\n>Unlike other open source licenses (GNU, Mozilla,\n>Interbase), the original Berkeley license does not take into account\n>that over time a lot of different individual developers and perhaps\n>some corporate contributors, would have individual copyrights on\n>substantial portions of the code.\n\nIt does not need to in the sense that while it remains public and in it's\noriginal form, an implied license is granted. Otherwise the notice that the\ndeveloper placed in the code holds, and failing that, whatever the\ngoverning law for the country/state of origin applies. Generally this means\nthat the developer owns the copyright, but gives others rights to use it in\nthe public version of PG.\n\n\n>This deficiency has two adverse affects. First, the contributing\n>developers are not afforded the protection of the exculpatory\n>language in \"bold face.\"\n\nI agree this is a problem. Developers should be warned to always add some\nkind of text like this to all their public code - not just PG.\n\n\n>Second, and admittedly of\n>significant importance to Great Bridge, the commercial proliferation\n>of PostgreSQL could be hindered if business users are concerned that\n>the license might not cover the substantial additions and\n>improvements made to the code over the last few years.\n\nThey will be covered so long as they release their developments back into\nthe community, I think.\n\n\n>In developing the new language, the resources of two intellectual\n>property law firms, one East Coast and one West Coast, were tapped.\n\nWhat about European (east and west), Japanese, and Australian?\n\n\n>No discussion of this type should be without controversy, so I throw\n>you the following red meat: the choice of state law has been\n>selected to cause the application of the Uniform Computer\n>Information Transactions Act (UCITA) to the usage of the software.\n>\n>(Pause for outrage to subside.)\n\nWhat is it? Does it even apply to me? [I am awaiting advice on this from my\nIP lawyer]\n\n\n>I submit that your efforts should be exempted from any potential\n>liability.\n\nThis is good. We can add the extra <bold face> paragraph.\n\n\n>Two states have adopted UCITA - Virginia and Maryland. \n\nAs a matter of interest - do Virginia & Maryland have a reputation for\nbeing forward looking in their law making, and being protective of the\nright of individuals over companies and government? I ask this because I\nknow very little about individual US states...\n\n\n>had to be contributed back to the public use, would be poorly\n>received and would possibly discourage the use by businesses who\n>might want to make certain improvements for their internal use\n\nHow about allowing developers the choice, as they have now?\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 04 Jul 2000 12:13:12 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "On Tue, 4 Jul 2000, Thomas Lockhart wrote:\n\nI'm not going to comment on those points that Thomas said that I do agree\nwith, since it could become a very long email ...\n\n> > 6) This is a very US-centric view of the world. Most of the developers\n> > are not in the US if the postgresql.org home page is correct. We don't\n> > care about the stinkin UCITA, we are not bound by and don't care about\n> > anything the State of Virginia may or may not say.\n> \n> Good point. But the USA is the demon spawning ground for lawyers, and is\n> at the leading edge of aggressive new legal territory. That may change\n> eventually, but since 90+% of our federal legislative representatives\n> are lawyers (stats from memory, but it is a *high* number), that may not\n> change very quickly :(\n\nPoint 6 here is the one that prevents me from being able to back up this\nchange, and is the reason I'm against it. PostgreSQL, for 3+ years, has\nbeen a *Canadian* based project, yet now she's going to fall under US\nlaws? The whole 'juristiction of Virginia' point puts me on the\n\"anti-changes\" side of this issue ... and other then that point, (and\npending several more re-reads), I like the general wording of the\nadditions ...\n\n> > 8) \"To be integrated with the software in such a way that this license\n> > must be seen before downloading can occur\".\n> > Umm, can all the laywers please just butt out? Every other open-source\n> > package in the universe just relies on a licence file in the home\n> > directory. You going to try and stop people downloading with clicking a\n> > licence agreement? How you going to handle mirrors? Or are you not\n> > going to mirror any more? What about Red Hat el al?\n> \n> Good point. Not exactly sure why this was suggested, but the American\n> courts are *full* of cases where the plaintif said that they \"didn't\n> really know\" something that should have been obvious.\n\nPoint 8 here I'm against also ... god, could you imagine having to \"agree\nto an open source license\" each time you wanted to download it? \n\n> I (and *all* of the steering committee) had pretty much the same\n> reaction as you did at first. But some of us are closer to the US legal\n> system, and see what silliness it can generate, so came around to\n> thinking that there was something to be gained by license additions.\n\nFirst off, why are we trying to set a precedent here for the open source\ncommunity? Have no other open source projects out there not looked at the\nlegal ramifications of their softare? Why are we more special then, say,\nLinux(GPL), FreeBSD(Standard BSD), MySQL(GPL), KDE(GPL), etc as far as\nlicensing is concerned?\n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n\n", "msg_date": "Mon, 3 Jul 2000 23:41:29 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "On Tue, 4 Jul 2000, Philip Warner wrote:\n\n> >One issue that has always been a source of uncertainty - I think for\n> >all of us - has been the license under which PostgreSQL is\n> >distributed.\n> \n> Only an issue of uncertainty when people talk about changing it.\n> \n> As a company who wants PostgreSQL to remain in the public domain, I would\n> prefer to see it go GPL; this effectively prevents another company coming\n> along and swallowing the major developers as a means of stifling further\n> development. This latter tactic would probably never work, but it would be\n> extermely disruptive.\n\nActually, with the BSD license as it is now, that isn't an issue either\n... if someone where to come along and 'close the source', that license\nchange couldn't only be on future changes, not past ones ... as Vadim has\nstated previously, he'd just go off and branch off the code and continue\nopen source ...\n\n> >We've also found, through some rather extensive market\n> >research\n> \n> Out of curiosity, who with & where?\n\nAmericans ... I do not believe they've done any market research in any\ncountry out of the USofA, but I may be wrong here ...\n\n\n> >This deficiency has two adverse affects. First, the contributing\n> >developers are not afforded the protection of the exculpatory\n> >language in \"bold face.\"\n> \n> I agree this is a problem. Developers should be warned to always add some\n> kind of text like this to all their public code - not just PG.\n\nAnd any developer is more then welcome to add that to their patches when\nthey submit it ... just nobody has done it to date ...\n\n> >In developing the new language, the resources of two intellectual\n> >property law firms, one East Coast and one West Coast, were tapped.\n> \n> What about European (east and west), Japanese, and Australian?\n\nhey, what about Canadian, where this project operates out of? It isn't an\nAmerican project, it is \"Proudly Canadian\" with a crack team of developers\naround the world working on it ... by number(s), I would guess that the\nmajority of our developers are non-US citizens ...\n\n> What is it? Does it even apply to me? [I am awaiting advice on this\n> from my IP lawyer]\n\n From my undertanding, it only applies to those states (US) that have\npassed UCITA through legislature ... which, I believe, only accounts for 2\nstates right now out of 52, and zero other countries are even considering\nit (but on that point I might be mistaken) ...\n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n", "msg_date": "Mon, 3 Jul 2000 23:50:26 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: [HACKERS] proposed improvements to PostgreSQL\n license" }, { "msg_contents": "\n> Good point. But the USA is the demon spawning ground for lawyers, and is\n> at the leading edge of aggressive new legal territory. \n\nActually that is the exact reason you _don't_ want to be based in the\nUSA. Do you really want Postgres to be breaking new ground in the\ncourts? The USA is at the leading edge of lame new legislation. If the\npostgresql licence is locked into Virginia law forever, (because any\nlicence change will be forever), you are subject to that law forever no\nmatter how stupid it may get. \n\nFor that reason I don't think you should be naming a jurisdiction. You\ndon't know what that jurisdiction may do in the future. Now any normal\ncorporation in this event could just change their licence to\njurisdiction B which has more favourable laws. Open source can't change\nthe licence ever unless you assign the rights to every bit of submitted\ncode like RMS insists on for GNU code.\n\nIf you must pick a jurisdiction pick Australia. We are *much* less\nlitigious. :-) Actually, pick Sealand. They have no laws and no courts.\n", "msg_date": "Tue, 04 Jul 2000 13:05:37 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "\n\nThomas Lockhart wrote:\n> > 8) \"To be integrated with the software in such a way that this license\n> > must be seen before downloading can occur\".\n> > Umm, can all the laywers please just butt out? Every other open-source\n> > package in the universe just relies on a licence file in the home\n> > directory. You going to try and stop people downloading with clicking a\n> > licence agreement? How you going to handle mirrors? Or are you not\n> > going to mirror any more? What about Red Hat el al?\n> Good point. Not exactly sure why this was suggested, but the American\n> courts are *full* of cases where the plaintif said that they \"didn't\n> really know\" something that should have been obvious.\n\nMy dos centavos of an alternate solution:\nUpon a sucessful install, and/or when opening template1, spit this out on\nscreen. This means that to *use* the product, they must see the license\nat least once. \n\n> So the intent was, as stated, to *reinforce* what we already believe to\n> be true (including yourself). The recently-enacted UCITA law was\n> (afaict) intended to protect, perhaps wrongly imho, commercial software\n> companies from liability claims\n\nEven liability of their own making, and negligence... which might be why\nit was hotly contested, and possibly struck down soon.... is it written\nin such a way as to be enforcable if UTOCA is struck down?\n\n-Bop\n\n--\nBrought to you from iBop the iMac, a MacOS, Win95, Win98, LinuxPPC machine,\nwhich is currently in MacOS land. Your bopping may vary.\n", "msg_date": "Mon, 03 Jul 2000 20:23:07 -0700", "msg_from": "Ron Chmara <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] Re: [HACKERS] proposed improvements to PostgreSQL\n\tlicense" }, { "msg_contents": "Thomas Lockhart wrote:\n> However, the path to do this\n> isn't perfectly clear to anyone; this is the first concrete proposal we\n> have had which does try to address the issues we believe are already\n> here whether we want them or not.\n\nAs someone else mentioned, why does postgresql have to break new ground? \n\n> > I think this is a bad idea for the following reasons:\n> > 1) It is trying to be a GPL in what it is trying to achieve without\n> > actually being well thought out. Any person who \"submits\" modifications\n> > must do so under the same licence. Submits to what or whom?\n\n> It is *not* trying to be GPL. \n\nGPL is essentially \"You must make changes under the same licence\". As\nfar as I can see this licence is saying the same thing in a wishy-washy\nway.\n\n> It is trying to be BSD, while extending\n> liability protection to the current cast of developers,\n\nThe liability exclusion clause I don't really have a problem with. It's\nthe other bits that I'd concerned about. I ask again what does \"submits\"\nmean? Who does it mean to? The GPL has nailed down the definitions here.\nThis language is so wide I'd defy you to get the same meaning from most\npeople who read it.\n\n> The current license asks users to absolve the University of California\n> of any liability involving use of the Postgres source code. It does not\n> (currently) explicitly ask the same on behalf of the current developers\n> (including yourself ;)\n\nMy guess is that if anyone is going to be sued (which I just don't\nbelieve, but anyway....), it wouldn't be based on the software, it would\nbe based on what some developer has said on a mailing list. Given a\nworking compiler the source code will do exactly what the source code\nsays it should do. It's the statements the developers make in other\nforums which people will be relying on to know what the source code\nshould do.\n\n> > 3) You talk about how wonderful the BSD licence is, then you really\n> > change the whole meaning of that licence.\n> \n> How?\n\nBy changing what you are and aren't allowed to do with changes to the\ncode.\n\n> I disagree, though we don't know UC's motivations for sure. imho the BSD\n> license is intended to protect UC from \"deep pockets\" lawsuits, while\n> preserving some credit for the original design team and the institution\n> which made it possible.\n\nIf we accept the above, then why the restrictions on how you can change\nit?\n\n> Good point. But the USA is the demon spawning ground for lawyers, and is\n> at the leading edge of aggressive new legal territory. \n\nThe nice thing about the simple licence with no mention of legal\nterritory is that it can be sensibly interpreted in each independant\nlegal jurisdiction. If Virginia passes a law saying that any developer\nwho releases software with bugs with licencing subject to their laws,\nshall be hung until they are dead, then I am not affected, even if one\nday I want to visit Virginia. That's extreme I know, but what do I as an\nAustralian know abouth Virginia? For all I know they are a nazi regime.\nI just don't want anything I have to do with be in any way subject to\nthe laws of that state. Why would I?\n\n\nMy final statement would be this, YOU CAN ALWAYS MAKE AN OPEN-SOURCE\nLICENCE STRONGER. YOU CAN NEVER MAKE IT WEAKER EVER AGAIN.\n", "msg_date": "Tue, 04 Jul 2000 13:27:14 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "-On [20000704 08:00], Thomas Lockhart ([email protected]) wrote:\n>> I think this is a bad idea for the following reasons:\n>> 1) It is trying to be a GPL in what it is trying to achieve without\n>> actually being well thought out. Any person who \"submits\" modifications\n>> must do so under the same licence. Submits to what or whom?\n>\n>It is *not* trying to be GPL. It is trying to be BSD, while extending\n>liability protection to the current cast of developers, who are (I'm\n>pretty sure) not covered in any of the wording of the UCB-generated\n>license.\n\n * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND\n * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\n * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\n * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE\n * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL\n * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS\n * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)\n * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT\n * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY\n * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF\n * SUCH DAMAGE.\n\nSeems pretty clear to me. ``In no event shall the author or\ncontributors be liable for any...''\n\nAnyways, why do people always have to start whole threads on -announce?\nReply-to set. Please honour it.\n\n-- \nJeroen Ruigrok vd Werven/Asmodai asmodai@[wxs.nl|bart.nl|freebsd.org]\nDocumentation nutter/C-rated Coder BSD: Technical excellence at its best \nThe BSD Programmer's Documentation Project <http://home.wxs.nl/~asmodai>\nMalam bulan dipagar bintang makin indah jika dipandang bagai gadis beri\nsenyuman pada bujang idaman...\n", "msg_date": "Tue, 4 Jul 2000 08:19:34 +0200", "msg_from": "Jeroen Ruigrok/Asmodai <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "Thomas Lockhart <[email protected]> writes:\n> Postgres is starting to become a visible thing, and is going to be used\n> by people who don't know much about the free software movement. And\n> *I'm* within reach of the American court system, and *you* can\n> contribute code which could make me a target for a lawsuit.\n\nA further comment here: BSD and similar licenses have indeed been used\nsuccessfully for a couple of decades --- within a community of like-\nminded hackers who wouldn't dream of suing each other in the first\nplace. Postgres is starting to get out into a colder and harder world.\nTo name just one unpleasant scenario: if PG continues to be as\nsuccessful as it has been, sooner or later Oracle will decide that we\nare a threat to their continued world domination. Oracle have a\nlongstanding reputation for playing dirty pool when they feel it\nnecessary. It'd be awfully convenient for them if they could eliminate\nthe threat of Postgres with a couple of well-placed lawsuits hinging on\nthe weaknesses of the existing PG license. It'd hardly even cost them\nanything, if they can sue individual developers who have no funds for\na major court case.\n\nChris and Peter may not feel that they need to worry about the\nsillinesses of the American legal system, but those of us who are\nwithin its reach do need to worry about it.\n\nI'm not opining here about the merits or weaknesses of Great Bridge's\nproposal. (What I'd really like is to see some review from other\nlegal experts --- surely there are some people on these mailing lists\nwho can bring in their corporate legal departments to comment?) But\nwhat we have here is a well-qualified lawyer telling us that we've got\nsome problems in the existing license. IMHO we'd be damned fools to\nignore his advice completely. Sticking your head in the sand is not\na good defense mechanism.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 03:23:28 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] Re: [HACKERS] proposed improvements to PostgreSQL\n\tlicense" }, { "msg_contents": "At 03:23 4/07/00 -0400, Tom Lane wrote:\n>IMHO we'd be damned fools to\n>ignore his advice completely. Sticking your head in the sand is not\n>a good defense mechanism.\n\nI think virtually everybody is happy with the extra disclaimer. It the\nother parts that bother me.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 04 Jul 2000 17:46:35 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] Re: [HACKERS] proposed improvements to\n\tPostgreSQL license" }, { "msg_contents": "Tom Lane wrote:\n> \n> Thomas Lockhart <[email protected]> writes:\n> > Postgres is starting to become a visible thing, and is going to be used\n> > by people who don't know much about the free software movement. And\n> > *I'm* within reach of the American court system, and *you* can\n> > contribute code which could make me a target for a lawsuit.\n> \n> A further comment here: BSD and similar licenses have indeed been used\n> successfully for a couple of decades --- within a community of like-\n> minded hackers who wouldn't dream of suing each other in the first\n> place. Postgres is starting to get out into a colder and harder world.\n> To name just one unpleasant scenario: if PG continues to be as\n> successful as it has been, sooner or later Oracle will decide that we\n> are a threat to their continued world domination. Oracle have a\n> longstanding reputation for playing dirty pool when they feel it\n> necessary. \n\nDoes hiring private detectives to rifle through allies of\nMicrosoft's trash count as dirty pool? ;-) I personally feel that\nanalogies between PostgreSQL/Oracle and Linux/Windows NT are\nbecoming more realistic. You'll know PostgreSQL has reached Prime\nTime when a CNBC reporter asks Larry Ellison about it the same\nway they ask Bill Gates about Linux (sorry Marc).\n\n> It'd be awfully convenient for them if they could eliminate\n> the threat of Postgres with a couple of well-placed lawsuits hinging on\n> the weaknesses of the existing PG license. It'd hardly even cost them\n> anything, if they can sue individual developers who have no funds for\n> a major court case.\n> \n> Chris and Peter may not feel that they need to worry about the\n> sillinesses of the American legal system, but those of us who are\n> within its reach do need to worry about it.\n\n From a user's perspective, the only concern that I have is that\nit remains BSD-ish instead of GPL-ish. Commercial products built\naround database solutions often wander too vaguely into \"GPL vs.\nLGPL\" land to be safe, depending upon how \"wired\" they are in the\nproduct. For example, if PostgreSQL were GPL and libpq were LGPL,\nand I wanted to sell a product which required SPI or new types,\nwould I have to release such source? With pure BSD the ambiguity\nis gone. The \"intentions\" mentioned in the proposal seemed GPLish\neven though the agreement seemed BSDish. \n\n> \n> I'm not opining here about the merits or weaknesses of Great Bridge's\n> proposal. (What I'd really like is to see some review from other\n> legal experts --- surely there are some people on these mailing lists\n> who can bring in their corporate legal departments to comment?) But\n> what we have here is a well-qualified lawyer telling us that we've got\n> some problems in the existing license. IMHO we'd be damned fools to\n> ignore his advice completely. Sticking your head in the sand is not\n> a good defense mechanism.\n\nMy distaste for the profession grows with every day (just try and\nwade through corporate tax law). Its a pretty sorry state we're\n(Americans) in when guys who want to give out software *free*\nhave to worry about the legal consequences...But, for what its\nworth, I agree with your conclusions :-(\n\nMike Mascari\n", "msg_date": "Tue, 04 Jul 2000 04:06:29 -0400", "msg_from": "Mike Mascari <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] Re: [HACKERS] proposed improvements to PostgreSQL\n\tlicense" }, { "msg_contents": "Ned Lilly wrote:\n> \n> > Two states have adopted UCITA - Virginia and Maryland. Maryland has\n> > an October 1, 2000, effective date, but requires that its laws will\n> > only apply if there is a reasonable connection with the state.\n> > Virginia has an effective date of July 1, 2001, but does not require\n> > a connection with the state and thereby gives somewhat greater\n> > assurance that UCITA will apply to all Postgres-related dealings,\n> > wherever they occur.\n\nNot here in Scotland, they won't. If people in the United States feel\nthat United States law prevents them contributing to Open Source\nprojects, that is a local problem which should be addressed locally - by\nlobbying their representatives to change the law.\n\n> > The fact that Great Bridge is based in\n> > Virginia is really a complete coincidence.\n\nI was initially agnostic regarding Great Bridge's involvement. Now I am\nnot so sure.\n\nI would regard any variation from one of the Big Two open source\nlicences an extremely retrograde step -- the more different licences\nthere are out there, the more confusion there is, and the more room\nthere is for sleight of hand like the soi-disant 'open' Motif licence.\nThat's why my company uses the exact wording of the BSD licence for our\nproducts; if we were to tighten up at all it would be to adopt the GPL.\nIf there is to be any change to the BSD licence currently used for\nPostgres I would suggest it be limited to:\n\ns/\\(University of California\\)/\\1 and the developers listed in the\nHISTORY file/g\n\nSincerely\n\nSimon Brooke\n\n-- \nSimon Brooke, Technical Director, Weft Technology Ltd --\nhttp://www.weft.co.uk/\n\n\tthe weft is not just what binds the web: it is what makes it a web\n", "msg_date": "Tue, 04 Jul 2000 10:38:47 +0100", "msg_from": "Simon Brooke <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [ANNOUNCE] Re: [HACKERS] proposed improvements to PostgreSQL\n\tlicense" }, { "msg_contents": "On Tue, 4 Jul 2000, Tom Lane wrote:\n\n> Chris and Peter may not feel that they need to worry about the\n> sillinesses of the American legal system, but those of us who are\n> within its reach do need to worry about it.\n\nI grant you that, but as Chris pointed out the proposed change may\nactually have a net negative effect, namely bringing those outside the\nreach of the American legal system withing it, and at the same time not\ndoing anything for other silly legal systems.\n\nI, and I think most others, don't have a problem with repeating the\nexisting boilerplate with s/Regents of the University of\nCalifornia/various contributors/g.\n\n-- \nPeter Eisentraut Sernanders vaeg 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Tue, 4 Jul 2000 07:23:21 -0400 (EDT)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: Re: [HACKERS] proposed improvements to PostgreSQL\n license" }, { "msg_contents": "On Tue, 4 Jul 2000, Thomas Lockhart wrote:\n\n> It is being proposed as an addition to the Postgres development effort.\n\nThe Yet Another Open Source License issue is not to be played with. It\nwill have to go to the OSI and RMS, Slashdot, all the usual suspects. And\nyou know what it will say? \"PostgreSQL changes to hand-crafted license\ndictated by corporate interests\", no matter how much anyone denies\nthat.\n\nThere are enough open source licenses out there. Use one.\n\n-- \nPeter Eisentraut Sernanders vaeg 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Tue, 4 Jul 2000 07:30:31 -0400 (EDT)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: Re: [HACKERS] proposed improvements to PostgreSQL\n license" }, { "msg_contents": "\nNote that I have no issues at all with the addition of the three BOLD\nparagraphs ... it is the \"under juristiction of the state of\nVirginia\" part that I have an issue with, as I've noticed, do those other\ndevelopers outside of the USofA ...\n\n\n\nOn Tue, 4 Jul 2000, Tom Lane wrote:\n\n> Thomas Lockhart <[email protected]> writes:\n> > Postgres is starting to become a visible thing, and is going to be used\n> > by people who don't know much about the free software movement. And\n> > *I'm* within reach of the American court system, and *you* can\n> > contribute code which could make me a target for a lawsuit.\n> \n> A further comment here: BSD and similar licenses have indeed been used\n> successfully for a couple of decades --- within a community of like-\n> minded hackers who wouldn't dream of suing each other in the first\n> place. Postgres is starting to get out into a colder and harder world.\n> To name just one unpleasant scenario: if PG continues to be as\n> successful as it has been, sooner or later Oracle will decide that we\n> are a threat to their continued world domination. Oracle have a\n> longstanding reputation for playing dirty pool when they feel it\n> necessary. It'd be awfully convenient for them if they could eliminate\n> the threat of Postgres with a couple of well-placed lawsuits hinging on\n> the weaknesses of the existing PG license. It'd hardly even cost them\n> anything, if they can sue individual developers who have no funds for\n> a major court case.\n> \n> Chris and Peter may not feel that they need to worry about the\n> sillinesses of the American legal system, but those of us who are\n> within its reach do need to worry about it.\n> \n> I'm not opining here about the merits or weaknesses of Great Bridge's\n> proposal. (What I'd really like is to see some review from other\n> legal experts --- surely there are some people on these mailing lists\n> who can bring in their corporate legal departments to comment?) But\n> what we have here is a well-qualified lawyer telling us that we've got\n> some problems in the existing license. IMHO we'd be damned fools to\n> ignore his advice completely. Sticking your head in the sand is not\n> a good defense mechanism.\n> \n> \t\t\tregards, tom lane\n> \n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n", "msg_date": "Tue, 4 Jul 2000 08:54:50 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [ANNOUNCE] Re: Re: [HACKERS] proposed improvements to\n\tPostgreSQL license" }, { "msg_contents": "On Mon, 3 Jul 2000, Peter Eisentraut wrote:\n\n> > and ensuring that the code stays open source in perpetuity.\n> No, that's what the GPL does.\n\nThis is only an end user's reply but here goes...\n\nAnd I feel alot more comfortable with the GPL as an end user. I *trust*\nRichard Stallman...alot more than any johnny-come-lately. Peter's point\nabout the longevity of the Bersekeley licence is well taken.\n\n> To my knowledge, the BSD license has been used in one form or another for\n> at least 20 years and neither has any contributor ever been sued for\n> liability, nor was there any court case that concluded that the BSD\n> license is worth anything at all, nor has the developer or commercial\n> acceptance of any product ever been affected by this \"untight\" license.\n> \n> > [To be integrated with the software in such a way that this license\n> > must be seen before downloading can occur]\n> \n> That's funny...\n\nActually, that's frightening...more than a bit reminiscent of the old Bill.\nI've invested *alot* of time in writing code that wraps around Pg.\nBecause of its OSS licence and Berkeley lineage.\n\nPerhaps the end user should also have to enter a key to do the build.\nAnd subsequently be pestered to register online for 'free updates'...\nMaybe code could be worked in to reach out on the network to see if any \nunauthorized binaries are in use.\n\n> > The foregoing shall be governed by and construed under the laws of\n> > the State of Virginia.\n> \n> The recurring theme throughout this email was that Great Bridge has\n> apparently not appreciated that PostgreSQL land extends beyond the borders\n> of the U.S. of A. Maybe your 32 focus groups in major U.S. cities wanted\n> the license changed like this, but I'll bet lunch that 32 out of 32 focus\n> groups in major European cities will look with extreme suspicion at\n> anything with \"laws of the State of XXX\" attached to it.\n\n> Until they realize that the laws of Virginia don't apply to them. Or to\n> Canada, where hub.org is located these days.\n\nAh, The Old Dominion. In NYC we have some of the toughest gun laws in\nthe US. But they are largely ineffective (aside from blocking honest\ncitizens access to sporting firearms). You see all sorts of guns flow in\nillegally from states that don't enforce their laws. Like Virginia.\nThe end result is that hospital ERs continue to treat gunshot wounds.\n\nRewriting the GPL or BSD licence sounds like reinventing the wheel...\nUnless of course there is another agenda.\n\n--------------------------------------------------------------------\n SVCMC - Center for Behavioral Health \n--------------------------------------------------------------------\nThomas Good tomg@ { admin | q8 } .nrnet.org\nIS Coordinator / DBA Phone: 718-354-5528 \n Fax: 718-354-5056 \n--------------------------------------------------------------------\nPowered by: PostgreSQL s l a c k w a r e FreeBSD:\n RDBMS |---------- linux The Power To Serve\n--------------------------------------------------------------------\n\n", "msg_date": "Tue, 4 Jul 2000 09:00:44 -0400 (EDT)", "msg_from": "Thomas Good <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: [HACKERS] proposed improvements to PostgreSQL\n license" }, { "msg_contents": "Chris Bitmead wrote:\n\n> Actually that is the exact reason you _don't_ want to be based in the\n> USA. Do you really want Postgres to be breaking new ground in the\n> courts? The USA is at the leading edge of lame new legislation. If the\n> postgresql licence is locked into Virginia law forever, (because any\n> licence change will be forever), you are subject to that law forever no\n> matter how stupid it may get.\n\nBesides, it effectively reduces the rights of any non-US developers to\nzero for sheer cost reasons, as they'd have to defend them in a Virginia\n(or at any rate US) court. And liabilities issues are far more likely to\ncrop up in the US than anywhere else, where sueing for damages seems to\nbe a profitable business. \n\nSevo\n\n-- \[email protected]\n", "msg_date": "Tue, 04 Jul 2000 15:33:35 +0200", "msg_from": "Sevo Stille <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "Philip Warner <[email protected]> el d�a Tue, 04 Jul 2000 12:13:12 +1000, \nescribi�:\n\n>As a company who wants PostgreSQL to remain in the public domain, I would\n>prefer to see it go GPL; \n\nI agree with this.\n(altough is not public domain, it's copywrigth'ed, well copyleft'ed).\n\nbtw, if you change the license in this way, is =not= BSD anymore,\nhow you will call this new license ?\n\n PPL (Postgres Public License) ?\n PBML (Postgres BSD Modified License) ?\n\nsergio\n\n", "msg_date": "Tue, 4 Jul 2000 11:11:12 -0300", "msg_from": "\"Sergio A. Kessler\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "\nOne thing to keep in mind: for a very long time, PostgreSQL was the *only* free (\"free as in free speech, not free as in free beer\") DBMS. I told dozens of people to consider PostgreSQL instead of, say, MySQL, for that very reason. Whichever free software licence you preferred, there was no real choice.\n\nNow, it is no longer the case: as you have read here, MySQL is now fully GPL. The concurrency between the two DBMS will increase. Since MySQL has a licence which is more hacker-friendly (it cannot be turned into a proprietary product), PostgreSQL, which had (along with techincal strengthes) a big advantage with its licence, is now behind.\n\n\n\n", "msg_date": "Tue, 04 Jul 2000 16:38:21 +0200", "msg_from": "Stephane Bortzmeyer <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license " }, { "msg_contents": "> That depends on what your market is - for businesses who wants to be\n> able to hide source, yes. For businesses who use it, being sure the\n> source is available is the best - which the GPL guarantees. BSD gives\n> the middle man more freedom to screw the end user ;)\n\nWell, we all want more freedom, right? (please note sarcastic tone ;)\n\n> > What we'd like to propose is a general tightening up of what the\n> > existing license is *supposed* to be doing in the first place -\n> > protecting the developers who worked on the code, and ensuring that\n> > the code stays open source in perpetuity.\n> GPL would solve this - the main advantage of BSDish licenses is you\n> can go closed source if you want to.\n\nI imagine that RH has extensive ongoing internal discussions of\nlicenses. Is there a \"company opinion\" that the main advantage of BSD is\nthat you can go closed source? \n\nimho an advantage of BSD is that there is no question that you can use\nthe open source anywhere you want, at any time, mixed with any other\ncode you want. For some, that might be a \"main advantage\"; for others, a\n\"don't care\". Can't really see it as a negative from my PoV.\n\n> Now, I don't advocate a change in license - my main consern is \"there\n> are enough licenses in the world\". I think the \"each package one\n> license\" is a bad trend.\n\nMe too. PostgreSQL has been distributed with a plain-vanilla BSD license\nforever. We would like to keep it that way. But BSD doesn't say anything\nabout developers outside of the UC system, so in the long run we\nprobably need to do something to address that. And I don't know about\nany BSD licenses or existing offshoots which do that (though I haven't\nlooked much beyond the packages I already know). istm that in most cases\n\"companies with lawyers\" go for something much tighter and more\nrestrictive than BSD or the recently suggested modification.\n\nRegards.\n\n - Thomas\n", "msg_date": "Wed, 05 Jul 2000 05:24:32 +0000", "msg_from": "Thomas Lockhart <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: [HACKERS] proposed improvements to PostgreSQL license" }, { "msg_contents": "Simon Brooke wrote:\n\n> Ned Lilly wrote:\n> > \n> > > Two states have adopted UCITA - Virginia and Maryland. Maryland has\n> > > an October 1, 2000, effective date, but requires that its laws will\n> > > only apply if there is a reasonable connection with the state.\n> > > Virginia has an effective date of July 1, 2001, but does not require\n> > > a connection with the state and thereby gives somewhat greater\n> > > assurance that UCITA will apply to all Postgres-related dealings,\n> > > wherever they occur.\n> \n> Not here in Scotland, they won't. If people in the United States feel\n> that United States law prevents them contributing to Open Source\n> projects, that is a local problem which should be addressed locally - by\n> lobbying their representatives to change the law.\n\nLobbying is an art of influencing representatives to cater for someone\nelse's needs. This is how the UCITA proposal emerged in the first\npace: through an extensive lobbying by those badly reputed industries\nwho started losing in the marketplace to open source/free software, or\nbecame apprehensive of the possibility of such losses. Aside from the\nmoney and other forces involved in lobbying, the reasoning in support\nof UCITA has a lot of brainwashing potential: why should one allow\nfree software to exist at all, let alone without liability by default,\nin the same time when the free health care, free police and free money\nmarket are banned from existence. This is, I guess, the kind of\nuniformity sought by the UCITA -- what else does that \"U\" stand for?\n\nWith that said, how likely do you think is that all software\ndevelopers of Maryland together, even backed up by all law professors\nand practicing lawyers, will be able to overturn the decision to adopt\nUCITA by simply annoying their representatives? Considering that most\nlawmakers and lobbyists of the World live in a 20-mile wide area in\nVirginia and Maryland centered around the U.S. Capitol, and that the\neconomy of these states depends on the lobbying industry more than\nJamaica on tourism, I wouldn't bet my lunch.\n\nBTW, I don't think that UCITA has terribly good chances of becoming\nthe United States law. While many states are still evaluating the\nproposal, a few others including Illinois, where I happen to live,\nhave declined to even consider it for evaluation.\n\nNow picture this: in a not so distant future you visit a web site\nwhose title page says:\n\n\"Residents of Europe, Canada, Illinois and other free countries,\nplease go ahead. Residents of Virginia and Maryland, please kindly\nfollow <a>this link<a>. Your local law requires us to harass you some\nmore before you can download. Avoiding this harassment by making an\ninappropriate selection voids our responsibility.\n\nHow's that as an alternative to a pop-up license statement?\n\nI wonder what Dilbert has to say about this.\n\n--Gene\n", "msg_date": "Wed, 05 Jul 2000 13:43:37 -0500", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: Re: [ANNOUNCE] Re: [HACKERS] proposed improvements to PostgreSQL\n\tlicense" }, { "msg_contents": "Ned,\n\nThanks for inviting the community to participate in this discussion. I\nwonder, though, if you might like to invite the participation of a wider\naudience. While I'm sure the subscribers to this list are fervent about\nall matters related to PostgreSQL, perhaps the subject matter deserves\nthe scrutiny of a larger and more diverse community. I might suggest\nthat beloved cesspool of civil discord - Slashdot.\n\nAs for the particulars of your proposal, I'd like to suggest, and I see\nothers agree, that it would still be premature to table the discussion\nof GPL vs. BSD style licensing. If for no other reason than if not now,\nwhen?\n\nThere seem to be two primary objectives here: (1) protect contributers\nfrom liability. (2) maintain the code as open source.\n\nI don't really understand liability issues or how they relate to the GPL\n(or any other license for that matter). I'm certainly 100% in favor of\nprotecting PostgreSQL developers from court claims, of course. So I'm\nnot going to chime in about liability issues.\n\nOne objection to the use of the GPL has been that it has never been\ntested in court. That may soon change. See\nhttp://www.linuxplanet.com/linuxplanet/reports/2000/1/.\n\nThen of course there's the discussion about which license is really more\n\"free\". True, a BSD style license places no restrictions on how someone\nmay use the code. So you are \"free to innovate\", as it were. Isn't\nanyone worried that PostgreSQL might become it's own competition?\n\n> ...we're big fans of the current Berkeley license; we find it\n> more \"open\" than other open source licenses, in the sense that the\n> user/hacker has almost total freedom as to what he wants to do with\n> the code. \n\t-Ned Lilly\n\nTo me, it's the difference between the freedom of anarchy, and the\nfreedom afforded by good government. Licenses are inherently\nrestrictive. That's the whole point of having them. This is true even\nfor BSD style licenses. So the question is not \"do you ask the users of\nyour software to make any concessions?\". Of course you do. The\nquestion is just what concessions do you require before granting use of\nyour product. Any statement to the effect that BSD is \"really free\" is\njust navel gazing mumbo jumbo.\n\nI keep seeing mention of the \"fact\" that the \"business community\"\nprefers a BSD style license to the GPL. Might I ask for details on how\nthis conclusion was reached?\n\n> We've also found, through some rather extensive market\n> research, that the business community (to which we'll be selling\n> products and services) vastly prefers it over GPL, or hybrids like\n> Mozilla, etc.\n\t-Ned Lilly\n\nI would submit that most businesses don't know the difference. Perhaps\nthey need some education.\n\nI would also submit that that the manner in which a survey was conducted\ncould greatly influence it's own results.\n\nQ: \"Do you prefer a GPL or BSD style license?\"\nA: \"What's the difference?\"\nQ: \"A BSD style license gives you more flexibility in how you administer\nchanges you might make to the software.\"\nA: \"Well, then BSD of course.\"\n\nNevermind that this same business might be running half of its back end\nservices using GPL'd software.\n\nHow about \"Would you like to know that you can take advantage of\ncontributions made to this software by anyone working on it worldwide?\",\nor \"Would you like a *guarantee* that the core development team won't\nduck tail and make you start paying for certain improvements?\"\n\nI'm not accusing anyone of malicious intentions. I'm just saying that\nthe only *guarantee* of good intentions is the license associated with\nthe software.\n\nThrow my response into the survey: my business would prefer GPL'd\nsoftware.\n\n-Ron-\n", "msg_date": "Wed, 05 Jul 2000 14:47:20 -0400", "msg_from": "Ron Peterson <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "\njust to curtail this while thread to a certain point ... switching the\nlicense to GPL is *not* on the table, nor has it every been, nor will it\never be ... \n\nOn Wed, 5 Jul 2000, Ron Peterson wrote:\n\n> Ned,\n> \n> Thanks for inviting the community to participate in this discussion. I\n> wonder, though, if you might like to invite the participation of a wider\n> audience. While I'm sure the subscribers to this list are fervent about\n> all matters related to PostgreSQL, perhaps the subject matter deserves\n> the scrutiny of a larger and more diverse community. I might suggest\n> that beloved cesspool of civil discord - Slashdot.\n> \n> As for the particulars of your proposal, I'd like to suggest, and I see\n> others agree, that it would still be premature to table the discussion\n> of GPL vs. BSD style licensing. If for no other reason than if not now,\n> when?\n> \n> There seem to be two primary objectives here: (1) protect contributers\n> from liability. (2) maintain the code as open source.\n> \n> I don't really understand liability issues or how they relate to the GPL\n> (or any other license for that matter). I'm certainly 100% in favor of\n> protecting PostgreSQL developers from court claims, of course. So I'm\n> not going to chime in about liability issues.\n> \n> One objection to the use of the GPL has been that it has never been\n> tested in court. That may soon change. See\n> http://www.linuxplanet.com/linuxplanet/reports/2000/1/.\n> \n> Then of course there's the discussion about which license is really more\n> \"free\". True, a BSD style license places no restrictions on how someone\n> may use the code. So you are \"free to innovate\", as it were. Isn't\n> anyone worried that PostgreSQL might become it's own competition?\n> \n> > ...we're big fans of the current Berkeley license; we find it\n> > more \"open\" than other open source licenses, in the sense that the\n> > user/hacker has almost total freedom as to what he wants to do with\n> > the code. \n> \t-Ned Lilly\n> \n> To me, it's the difference between the freedom of anarchy, and the\n> freedom afforded by good government. Licenses are inherently\n> restrictive. That's the whole point of having them. This is true even\n> for BSD style licenses. So the question is not \"do you ask the users of\n> your software to make any concessions?\". Of course you do. The\n> question is just what concessions do you require before granting use of\n> your product. Any statement to the effect that BSD is \"really free\" is\n> just navel gazing mumbo jumbo.\n> \n> I keep seeing mention of the \"fact\" that the \"business community\"\n> prefers a BSD style license to the GPL. Might I ask for details on how\n> this conclusion was reached?\n> \n> > We've also found, through some rather extensive market\n> > research, that the business community (to which we'll be selling\n> > products and services) vastly prefers it over GPL, or hybrids like\n> > Mozilla, etc.\n> \t-Ned Lilly\n> \n> I would submit that most businesses don't know the difference. Perhaps\n> they need some education.\n> \n> I would also submit that that the manner in which a survey was conducted\n> could greatly influence it's own results.\n> \n> Q: \"Do you prefer a GPL or BSD style license?\"\n> A: \"What's the difference?\"\n> Q: \"A BSD style license gives you more flexibility in how you administer\n> changes you might make to the software.\"\n> A: \"Well, then BSD of course.\"\n> \n> Nevermind that this same business might be running half of its back end\n> services using GPL'd software.\n> \n> How about \"Would you like to know that you can take advantage of\n> contributions made to this software by anyone working on it worldwide?\",\n> or \"Would you like a *guarantee* that the core development team won't\n> duck tail and make you start paying for certain improvements?\"\n> \n> I'm not accusing anyone of malicious intentions. I'm just saying that\n> the only *guarantee* of good intentions is the license associated with\n> the software.\n> \n> Throw my response into the survey: my business would prefer GPL'd\n> software.\n> \n> -Ron-\n> \n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n", "msg_date": "Wed, 5 Jul 2000 15:57:57 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "* The Hermit Hacker <[email protected]> [000705 12:03] wrote:\n> \n> just to curtail this while thread to a certain point ... switching the\n> license to GPL is *not* on the table, nor has it every been, nor will it\n> ever be ... \n\nGood to hear, I was getting worried for a bit that the code might\nbecome uselessly encumbered by a non-opensource friendly license.\n\nIgnore the hype, keep the code _free_.\n\nthanks,\n-Alfred\n", "msg_date": "Wed, 5 Jul 2000 12:14:03 -0700", "msg_from": "Alfred Perlstein <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "Ron Peterson wrote:\n\n> your software to make any concessions?\". Of course you do. The\n> question is just what concessions do you require before granting use of\n> your product. Any statement to the effect that BSD is \"really free\" is\n> just navel gazing mumbo jumbo.\n\n Good points - alot of. Just that the above has a little flaw.\n\n We're not discussing \"what concessions\" the END-USER has to\n make before USING it. We ask \"what restrictions\" does a\n license apply to a DEVELOPER?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Wed, 5 Jul 2000 22:54:43 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "Ron Peterson wrote:\n> While I'm sure the subscribers to this list are fervent about\n> all matters related to PostgreSQL, perhaps the subject matter deserves\n> the scrutiny of a larger and more diverse community. I might suggest\n> that beloved cesspool of civil discord - Slashdot.\n\nUmm. Let's _not_ drag slashdot into this. I think any of us could easily\npredict what the slashdot consensus would be (hint: less free equals\nbad). And it would be bad publicity for pgsql.\n\n> Then of course there's the discussion about which license is really more\n> \"free\". True, a BSD style license places no restrictions on how someone\n> may use the code. So you are \"free to innovate\", as it were. Isn't\n> anyone worried that PostgreSQL might become it's own competition?\n\nI can't see any cause for worry. If you want to use a closed source\ndatabase there are plenty of very good products out there. How would yet\nanother closed source dbms compete with open-source postgres?\n\nThe X consortium recently tried to make X less free. The community took\nthe most recent free X and continued development. Basicly the X\nconsortium were beaten into submission.\n\n> I keep seeing mention of the \"fact\" that the \"business community\"\n> prefers a BSD style license to the GPL. Might I ask for details on how\n> this conclusion was reached?\n\nSome businesses get concerned about the licencing implications and how\nit might effect their plans. For example I was working for a company\nthat wanted to build TV set-top boxes for internet access. Maybe they\ndon't want the hassle of conforming to the GPL. Of course mostly these\nbusiness concerns are pretty much ill-founded.\n\n> I would submit that most businesses don't know the difference. Perhaps\n> they need some education.\n\nAre you volunteering?\n", "msg_date": "Thu, 06 Jul 2000 10:30:40 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "Chris Bitmead wrote:\n> \n> > I would submit that most businesses don't know the difference. Perhaps\n> > they need some education.\n> \n> Are you volunteering?\n\nOf course. As the systems administrator for my company, I feel it's my\nresponsibility to inform the principals who run this company about\nsoftware licensing issues that may affect them. I would hope that\nanyone with similar duties at any other company would feel the same way.\n\nI would like to future proof any technology investments we make to the\ngreatest degree possible. If I say \"let's build our project management\ndatabase using xyz\", I would like to assure anyone buying into this\nproposal that I have some confidence in the future direction xyz will\ntake.\n\nThe current licensing debate surrounding PostgreSQL has greatly\ndiminished my confidence in it's long term viability as an open source\nproject. I may be very wrong here, but I really don't feel like taking\nunnecessary chances.\n\nThe best argument so far as to how a BSD style license will maintain\nPostgreSQL's viability as an open source project is that if someone take\nthe code proprietary, someone else can fork the code and continue\ndevelopment as an open source project.\n\nThis has happened in PostgreSQL's own history. How long did it take for\nthe project to get picked up again? How long did it take for the people\nwho picked it up to familiarize themselves with the code? How long did\nit take before the community at large developed any confidence in the\nproject's viability? How much talent was lost? How many ideas were\nlost?\n\nI believe developers assurances about their desire to maintain\nPostgreSQL as an open source project are sincere. But I am not going to\ncontinue investing my time in PostgreSQL unless those assurances are\nbacked by contractually binding verbage.\n \n________________________\nRon Peterson\[email protected]\n", "msg_date": "Thu, 06 Jul 2000 10:19:41 -0400", "msg_from": "Ron Peterson <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "At 10:19 6/07/00 -0400, Ron Peterson wrote:\n>\n>This has happened in PostgreSQL's own history. How long did it take for\n>the project to get picked up again? How long did it take for the people\n>who picked it up to familiarize themselves with the code? How long did\n>it take before the community at large developed any confidence in the\n>project's viability? How much talent was lost? How many ideas were\n>lost?\n>\n>I believe developers assurances about their desire to maintain\n>PostgreSQL as an open source project are sincere. But I am not going to\n>continue investing my time in PostgreSQL unless those assurances are\n>backed by contractually binding verbage.\n\nShort of employing the core developers directly, and giving them good\nenough conditions to ensure that they stay with you, I don't think you can\nacheieve your ends. No license change will help. All a licence does is\ntells you what *you* can do with the software.\n\nGPL or BSD, if someone buys up the core developers, their replacements will\nhave a steep learning curve. Your only choice here is to invest local\ntalent from your company in the development project so that if the core\ndevelopers do leave, then there is a higher chance of a quick uptake.\n\nOr have I missed something? Did you have some kind of \"contractually\nbinding verbage\" in mind?\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Fri, 07 Jul 2000 01:59:29 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "On Thu, 6 Jul 2000, Ron Peterson wrote:\n\n> This has happened in PostgreSQL's own history. How long did it take\n> for the project to get picked up again? How long did it take for the\n> people who picked it up to familiarize themselves with the code? How\n> long did it take before the community at large developed any\n> confidence in the project's viability? How much talent was lost? \n> How many ideas were lost?\n\nHuh? When? I, and several others, picked up Postgres95 from the\nUniversity of California at Berkeley, following the graduate studies of\nAndrew and Jolly, in 1996 ... there has been no copyright change since\nthen (wish someone had mentioned soon that alther we didn't do any\ncopyright changes, Berkeley did *sigh*) ...\n\nPrior to that, Postgres95 was a graduate project, that, if I recall\ncorrectly, Andrea and Jolly took on to move Postgres from a PostQuel based\nto an SQL based server, but that part of the history I may be wrong on ...\n\nSince '96, there has only been one development effort on PostgreSQL\n... that I'm aware of ...\n\n\n", "msg_date": "Thu, 6 Jul 2000 13:24:29 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "Philip Warner wrote:\n> \n> At 10:19 6/07/00 -0400, Ron Peterson wrote:\n> >\n> >This has happened in PostgreSQL's own history. How long did it take for\n> >the project to get picked up again? How long did it take for the people\n> >who picked it up to familiarize themselves with the code? How long did\n> >it take before the community at large developed any confidence in the\n> >project's viability? How much talent was lost? How many ideas were\n> >lost?\n> >\n> >I believe developers assurances about their desire to maintain\n> >PostgreSQL as an open source project are sincere. But I am not going to\n> >continue investing my time in PostgreSQL unless those assurances are\n> >backed by contractually binding verbage.\n> \n> Short of employing the core developers directly, and giving them good\n> enough conditions to ensure that they stay with you, I don't think you can\n> acheieve your ends. No license change will help. All a licence does is\n> tells you what *you* can do with the software.\n> \n> GPL or BSD, if someone buys up the core developers, their replacements will\n> have a steep learning curve. Your only choice here is to invest local\n> talent from your company in the development project so that if the core\n> developers do leave, then there is a higher chance of a quick uptake.\n> \n> Or have I missed something? Did you have some kind of \"contractually\n> binding verbage\" in mind?\n\nTechnically, you're right, of course. There are no absolutes here, just\nprobabilities. Developers will come and go in either case. But in\nthese days of multi-billion dollar mergers and acquisisitons, I would\nprefer team as a whole be working under a license that would discourage\nthem from taking the code private. Losing any of the developers would\nbe a loss. But losing a significant fraction of them could be tragic.\n\nBy the way, I hope I'm not coming across as saying \"thanks for all your\nhard work, now go eat peanuts.\" I have great appreciation for what\nPostgreSQL has become, and for the developers who have made it so. That\nis why I keep at this tired argument. I would like to know that this\nproject will continue in a manner that benefits the little people, not\njust big corporate commercial interests.\n\nPerhaps some company _would_ like to pay the developers mucho dinero to\ncontinue PostgreSQL development. In fact I really hope so. I would\njust like to know that if that happened, there would be no obstacle to\nreturning privately developed code to the public domain.\n\nGreat Bridge has announced their intention to sell both products and\nservices. What are the products?\n\nI'm going to bow out of this discussion. (applause!) Sorry if it\nsounds old and tired to the veterans among us. I'm a little green in\nthe gills and I appreciate your sufferance.\n\n________________________\nRon Peterson\[email protected]\n", "msg_date": "Thu, 06 Jul 2000 12:42:38 -0400", "msg_from": "Ron Peterson <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "On Thu, 6 Jul 2000, Ron Peterson wrote:\n\n> By the way, I hope I'm not coming across as saying \"thanks for all\n> your hard work, now go eat peanuts.\" I have great appreciation for\n> what PostgreSQL has become, and for the developers who have made it\n> so. That is why I keep at this tired argument. I would like to know\n> that this project will continue in a manner that benefits the little\n> people, not just big corporate commercial interests.\n\nThere are enough of us here who are commit'd to Open Source in general,\nand PostgreSQL specifically, that that won't happen. \n\n> Great Bridge has announced their intention to sell both products and\n> services. What are the products?\n> \n> I'm going to bow out of this discussion. (applause!) Sorry if it\n> sounds old and tired to the veterans among us. I'm a little green in\n> the gills and I appreciate your sufferance.\n> \n> ________________________\n> Ron Peterson\n> [email protected]\n> \n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n", "msg_date": "Thu, 6 Jul 2000 15:03:28 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "Ron Peterson wrote:\n> > > I would submit that most businesses don't know the difference. Perhaps\n> > > they need some education.\n> >\n> > Are you volunteering?\n> \n> Of course. As the systems administrator for my company,\n\nOne company down. 99,999,999 to go.\n\n> The best argument so far as to how a BSD style license will maintain\n> PostgreSQL's viability as an open source project is that if someone take\n> the code proprietary, someone else can fork the code and continue\n> development as an open source project.\n> \n> This has happened in PostgreSQL's own history. How long did it take for\n> the project to get picked up again? How long did it take for the people\n> who picked it up to familiarize themselves with the code? How long did\n> it take before the community at large developed any confidence in the\n> project's viability? How much talent was lost? How many ideas were\n> lost?\n\nAre you saying that Illustra destroyed the original postgres project? I\ndon't buy that. Postgres back then wasn't an open-source project, it was\na university project. The internet wasn't widely used, there was no wide\ninterest in improving the code.\n", "msg_date": "Fri, 07 Jul 2000 11:01:32 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "\nChris ... kill this thread :) Ron and I talked in the background and I\nguess he got confused by Bruce's HISTORY in his book ... he got the\nimpression that things got \"bounced around\" a few times, and didn't\nrealize that it was all fluid from the days of Jolly/Andrew into us, with\na slow down before I kickstarted things in '96 ...\n\n\n\nOn Fri, 7 Jul 2000, Chris Bitmead wrote:\n\n> Ron Peterson wrote:\n> > > > I would submit that most businesses don't know the difference. Perhaps\n> > > > they need some education.\n> > >\n> > > Are you volunteering?\n> > \n> > Of course. As the systems administrator for my company,\n> \n> One company down. 99,999,999 to go.\n> \n> > The best argument so far as to how a BSD style license will maintain\n> > PostgreSQL's viability as an open source project is that if someone take\n> > the code proprietary, someone else can fork the code and continue\n> > development as an open source project.\n> > \n> > This has happened in PostgreSQL's own history. How long did it take for\n> > the project to get picked up again? How long did it take for the people\n> > who picked it up to familiarize themselves with the code? How long did\n> > it take before the community at large developed any confidence in the\n> > project's viability? How much talent was lost? How many ideas were\n> > lost?\n> \n> Are you saying that Illustra destroyed the original postgres project? I\n> don't buy that. Postgres back then wasn't an open-source project, it was\n> a university project. The internet wasn't widely used, there was no wide\n> interest in improving the code.\n> \n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n", "msg_date": "Fri, 7 Jul 2000 09:42:16 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "Greetings!\n\nRon Peterson <[email protected]> writes:\n\n> Throw my response into the survey: my business would prefer GPL'd\n> software.\n> \n> -Ron-\n> \n> \n\nAs would mine. The major cost associated with any software project is\nthe migration/learning curve time. Effective guarantees that this\ntime investment will have a long payoff period (i.e. there will be no\nforced upgrades, mandatory incompatibilities, binary/linked library\nlegacy issues, proprietary usurpation, and disappearance of the\nproduct altogether) are critical to the decision to begin putting\nsoftware into production. \n\nAs I see it, the only realistic objection to the GPL comes from those\nwho want to sell copies of the software, or software products based\nupon it. Most people in the \"business community\" have no such\ninterest, but just want to use postgresql to get their work done. I'm\nnot saying that a commercial interest in selling software is\nillegitimate. I'm just saying that we ought to call a spade a spade,\nand recognize that turning away from the GPL is in the interest of\n(some) developers, not the customers/end users.\n\nThank you all for your work on postgreSQL!\n\n-- \nCamm Maguire\t\t\t \t\t\[email protected]\n==========================================================================\n\"The earth is but one country, and mankind its citizens.\" -- Baha'u'llah\n", "msg_date": "07 Jul 2000 10:47:46 -0400", "msg_from": "Camm Maguire <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" }, { "msg_contents": "\n> As I see it, the only realistic objection to the GPL comes from those\n> who want to sell copies of the software, or software products based\n> upon it. Most people in the \"business community\" have no such\n> interest, but just want to use postgresql to get their work done. I'm\n> not saying that a commercial interest in selling software is\n> illegitimate. I'm just saying that we ought to call a spade a spade,\n> and recognize that turning away from the GPL is in the interest of\n> (some) developers, not the customers/end users.\n\nIMHO, adding restrictions for (some) developers is not in the interest of\nany one, as it reduces opportunity and choice for both developer and end\nuser.\n\n\n- Andrew\n\n\n", "msg_date": "Sat, 8 Jul 2000 00:58:34 +1000", "msg_from": "\"Andrew Snow\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: proposed improvements to PostgreSQL license" }, { "msg_contents": "Camm Maguire writes:\n\n> As would mine. The major cost associated with any software project is\n> the migration/learning curve time. Effective guarantees that this\n> time investment will have a long payoff period (i.e. there will be no\n> forced upgrades, mandatory incompatibilities, binary/linked library\n> legacy issues, proprietary usurpation, and disappearance of the\n> product altogether) are critical to the decision to begin putting\n> software into production. \n\nI don't see your point. The current BSD license does, all in all, allow\nyou to do whatever you want with the product. So there will be no:\n\n* forced upgrades (No one can force you to upgrade.)\n\n* mandatory incompatibilities (No one can mandate anything, you can fix\n everything yourself.)\n\n* disappearance of the product (If you have the source, you have it and\n can continue to develop it.)\n\nYou seem to be believing that with the BSD license someone can take over\nthe product and retroactively re-license it. This is false on both counts:\nNo one can take it over unless he buys out the copyright of each\ncontributor (but no open source license prevents that), and no one can\nrevoke the currently granted license. (Well, there is a faction of legal\nexperts that believe that gratis licenses are revokable, but then every\nsuch license would be affected.)\n\n\n> As I see it, the only realistic objection to the GPL comes from those\n> who want to sell copies of the software, or software products based\n> upon it.\n\nThe major pragmatic objection to the GPL license has been that it creates\nlegal hassles which we currently don't have. If we allow anyone to use the\nproduct at will then we don't have anything to enforce. If we cover our\nown work with a bunch of conditions then we always have to worry about\nwhat we can do with the code and what we can't.\n\n> Most people in the \"business community\" have no such interest,\n\nBut some do and there's no point in stopping them.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Fri, 7 Jul 2000 21:27:11 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: proposed improvements to PostgreSQL license" } ]
[ { "msg_contents": "Hi.\n\nI suspect that maybe this can be solved by changing some tunables, but the\nproblem I am seeing is that after a \"VACUUM ANALYZE\" on our production\ndatabase, we get the following NOTICE's:\n\nNOTICE: RegisterSharedInvalid: SI buffer overflow\nNOTICE: InvalidateSharedInvalid: cache state reset\n\nThen each subsequent query on the clients which had open connections during\nthis time shows:\nNOTICE: InvalidateSharedInvalid: cache state reset\n\nMaybe I need to change some configuration parameter to prevent this? Is it\nsafe to ignore these?\n\nI'm runnign PostgreSQL 7.0.2\n\nI am running postgres like this:\n\npg_ctl -D /var/lib/pgsql/data -o -B 512 -i -N 128 -p /usr/bin/postmaster start\n\nHost system is Linux 2.2.x, dual P-III's, 768 MB ram. \n\nThanks for any tips anyone might have :)\n\nMike\n\n", "msg_date": "Mon, 3 Jul 2000 11:02:28 -0500 (CDT)", "msg_from": "Michael J Schout <[email protected]>", "msg_from_op": true, "msg_subject": "Strange \"NOTICE\".. Should I be concerned?" }, { "msg_contents": "Michael J Schout <[email protected]> writes:\n> we get the following NOTICE's:\n\n> NOTICE: RegisterSharedInvalid: SI buffer overflow\n> NOTICE: InvalidateSharedInvalid: cache state reset\n\nYou can ignore these --- they're really only still there for debugging\nreasons. See further discussion just a day or two ago (on pgsql-general\nIIRC).\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 15:31:45 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Strange \"NOTICE\".. Should I be concerned? " } ]
[ { "msg_contents": "If you get something like\n\nFATAL 2: Open(\"/home/peter/pg-install/data/global/pg_control\") failed: 2\n\nthen you need to\n\na) initdb\n\nb) cd ....data\n mkdir global\n mv pg_control pg_database pg_group* pg_log pg_shadow pg_variable global/\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Mon, 3 Jul 2000 22:56:15 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Advisory: When your pg_control file goes amiss..." } ]
[ { "msg_contents": " Date: Monday, July 3, 2000 @ 19:09:54\nAuthor: wieck\n\nUpdate of /home/projects/pgsql/cvsroot/pgsql/src/backend/utils/adt\n from hub.org:/tmp/cvs-serv55060/backend/utils/adt\n\nModified Files:\n\tarrayfuncs.c cash.c char.c float.c formatting.c int.c int8.c \n\tlztext.c mac.c nabstime.c network.c oid.c oracle_compat.c \n\tpg_lzcompress.c regproc.c ruleutils.c timestamp.c varbit.c \n\tvarchar.c varlena.c version.c \n\n----------------------------- Log Message -----------------------------\n\nTOAST\n\n WARNING: This is actually broken - we have self-deadlocks\n\t due to concurrent changes in buffer management.\n\t\t\t Vadim and me are working on it.\n\nJan\n\n\n", "msg_date": "Mon, 3 Jul 2000 19:09:56 -0400 (EDT)", "msg_from": "Jan Wieck <wieck>", "msg_from_op": true, "msg_subject": "pgsql/src/backend/utils/adt (arrayfuncs.c cash.c char.c float.c\n\tformatting.c int.c int8.c lztext.c mac.c nabstime.c network.c\n\toid.c oracle_compat.c pg_lzcompress.c regproc.c ruleutils.c\n\ttimestamp.c varbit.c varchar.c varlena.c version.c)" }, { "msg_contents": "Jan Wieck <[email protected]> writes:\n> TOAST\n> WARNING: This is actually broken - we have self-deadlocks\n> \t due to concurrent changes in buffer management.\n> \t\t\t Vadim and me are working on it.\n\nUh, exactly how broken would that be? Do you mean that the CVS tip\nis now nonfunctional and no one else can expect to get any useful\nwork done? Or do the problems only arise in special cases?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 19:23:51 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [COMMITTERS] TOAST" }, { "msg_contents": "Tom Lane wrote:\n> Jan Wieck <[email protected]> writes:\n> > TOAST\n> > WARNING: This is actually broken - we have self-deadlocks\n> > due to concurrent changes in buffer management.\n> > Vadim and me are working on it.\n>\n> Uh, exactly how broken would that be? Do you mean that the CVS tip\n> is now nonfunctional and no one else can expect to get any useful\n> work done? Or do the problems only arise in special cases?\n\n My usual sequence is\n\n - construct a patch\n - run \"cvs update\"\n - apply my patch to a local copy of the CVS tree\n - run the regression test\n - apply my patch to the checked out tree\n - cvs commit\n\n That avoids alot of trouble. The problems only arise if\n someone uses a new command to activate tuple move off to the\n secondary relation.\n\n I had some time where I wasn't able to compile the current\n CVS at all (my old, years grown, box got screwed up a little,\n and my new notebook wasn't installed to a useful level). The\n problem now is, that if the toaster really moved off values\n into the secondary relation, the backend locks up when trying\n to append a main tuple.\n\n Vadim changed something recently in heap_insert() and I\n already gave him a call. He'll look at the test case I posted\n to CORE soon.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Tue, 4 Jul 2000 01:57:07 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: Re: [COMMITTERS] TOAST" }, { "msg_contents": "With current CVS tip, I get an Assert() failure during initdb.\n\n#7 0x19268c in ExceptionalCondition (\n conditionName=0x76578 \"!(TransactionIdIsCurrentTransactionId(tuple->t_xmax))\n\", exceptionP=0x4000a650, detail=0x0, fileName=0x401933ed \"\", lineNumber=202)\n at assert.c:73\n#8 0x1a1b44 in HeapTupleSatisfiesNow (tuple=0x400b4680) at tqual.c:233\n#9 0x89ad8 in heapgettup (relation=0x401804f0, tuple=0x40166a24, dir=1,\n buffer=0x40166a58, snapshot=0x0, nkeys=1, key=0x40167ab0) at heapam.c:439\n#10 0x8ac48 in heap_getnext (scandesc=0x0, backw=0) at heapam.c:1028\n#11 0x18df5c in RelationBuildRuleLock (relation=0x4017dbe8) at relcache.c:801\n#12 0x18e1cc in RelationBuildDesc (buildinfo={infotype = 1, i = {\n info_id = 17280,\n info_name = 0x4380 \"�\\211�\\236g5?\\215�D�\\205\\021\\004�\\a\\230^��`\\2209�[\n� �E�T�\\\"LG��\\005*<�\\225�V\\211�\\016\\032\\020X��^�\\e35\\216y�\\205�\\013��<\\036��\"}}\n, oldrelation=0x4017dbe8) at relcache.c:993\n#13 0x18eb20 in RelationClearRelation (relation=0x4017dbe8, rebuildIt=0 '\\000')\n at relcache.c:1485\n#14 0x18eca0 in RelationFlushRelation (relationPtr=0x0, skipLocalRelations=2)\n at relcache.c:1584\n#15 0x18eebc in RelationIdInvalidateRelationCacheByRelationId (\n relationId=17280) at relcache.c:1653\n#16 0x18c574 in CacheIdInvalidate (cacheId=1663383, hashIndex=2,\n pointer=0x40195ffc) at inval.c:544\n#17 0x18c6f0 in InvalidationMessageCacheInvalidate (message=0x4018ef4c)\n at inval.c:658\n#18 0x18c0b8 in LocalInvalidInvalidate (invalid=0x4018ef4c \"r\",\n function=0x4000da7a <DINFINITY+9970>, freemember=1 '\\001') at inval.c:173\n#19 0x18c8a4 in ImmediateLocalInvalidation (send=0 '\\000') at inval.c:808\n#20 0xa6eb8 in AtCommit_LocalCache () at xact.c:714\n#21 0xa6cc0 in CommandCounterIncrement () at xact.c:535\n#22 0x145e0c in pg_exec_query_dest (\n query_string=0x40174320 \"CREATE VIEW pg_user AS SELECT\", ' ' <repeats 13 times>, \"usename,\", ' ' <repeats 13 times>, \"usesysid,\", ' ' <repeats 13 times>, \"usecreatedb,\", ' ' <repeats 13 times>, \"usetrace,\", ' ' <repeats 13 times>, \"usesuper,\", ' ' <repeats 13 times>, \"usecatupd,\", ' ' <repeats 13 times>, \"'********'::tex\"..., dest=Debug, parse_context=0x40195ffc) at postgres.c:670\n#23 0x146f44 in PostgresMain (argc=1073753344, argv=0x67800, real_argc=7,\n real_argv=0x7b03a6d8) at postgres.c:1584\n#24 0xed708 in main (argc=7, argv=0x7b03a6d8) at main.c:103\n\nAny ideas?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 20:11:22 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: [COMMITTERS] TOAST " }, { "msg_contents": "Tom Lane wrote:\n> With current CVS tip, I get an Assert() failure during initdb.\n>\n> #7 0x19268c in ExceptionalCondition (\n> conditionName=0x76578 \"!(TransactionIdIsCurrentTransactionId(tuple->t_xmax))\n> \", exceptionP=0x4000a650, detail=0x0, fileName=0x401933ed \"\", lineNumber=202)\n> at assert.c:73\n>\n> [...]\n>\n> Any ideas?\n\n Added --enable-cassert to my default configure switches.\n\n Hmmm - with my latest changes I get the same, without them I\n don't. Will backout the changes if I cannot find it quickly.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n\n", "msg_date": "Tue, 4 Jul 2000 02:52:38 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: Re: [COMMITTERS] TOAST" }, { "msg_contents": "[email protected] (Jan Wieck) writes:\n> Hmmm - with my latest changes I get the same, without them I\n> don't. Will backout the changes if I cannot find it quickly.\n\nThe problem appears to be that DataFill() is invoked on the tuple\nafter toasting, and it forces the infomask to zero, losing any\ntransaction status bits that may have been there. I think perhaps\nline 93 of heaptuple.c should be\n\n\t*infomask &= HEAP_XACT_MASK;\n\nnot\n\n\t*infomask = 0;\n\nComments anyone?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 22:19:52 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: [COMMITTERS] TOAST " } ]
[ { "msg_contents": "> The problem appears to be that DataFill() is invoked on the tuple\n> after toasting, and it forces the infomask to zero, losing any\n> transaction status bits that may have been there. I think perhaps\n> line 93 of heaptuple.c should be\n> \n> \t*infomask &= HEAP_XACT_MASK;\n> \n> not\n> \n> \t*infomask = 0;\n> \n> Comments anyone?\n\nOne should check all places from where DataFill called...\n\nBtw, I've found that I unlocked buffer in heap_insert *after* WriteBuffer -\nfixed.\n\n2Jan - please note NO ELOG(ERROR)... comment in heap_update: can you move\ntoast\ncode below xlog one?\n\nVadim\n", "msg_date": "Mon, 3 Jul 2000 19:45:55 -0700 ", "msg_from": "\"Mikheev, Vadim\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Re: [COMMITTERS] TOAST " }, { "msg_contents": "\"Mikheev, Vadim\" <[email protected]> writes:\n> One should check all places from where DataFill called...\n\nI did. All the other ones start with an already-zeroed infomask.\n\nI think this change is correct as far as it goes, though we still\nhave bugs elsewhere...\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 03 Jul 2000 23:01:16 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: [COMMITTERS] TOAST " }, { "msg_contents": "Mikheev, Vadim wrote:\n> 2Jan - please note NO ELOG(ERROR)... comment in heap_update: can you move\n> toast\n> code below xlog one?\n\n Will take a look if that matters.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Tue, 4 Jul 2000 13:37:49 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: Re: [COMMITTERS] TOAST" } ]
[ { "msg_contents": " Date: Tuesday, July 4, 2000 @ 01:17:03\nAuthor: momjian\n\nUpdate of /home/projects/pgsql/cvsroot/pgsql/doc/TODO.detail\n from hub.org:/home/projects/pgsql/tmp/cvs-serv13812/pgsql/doc/TODO.detail\n\nRemoved Files:\n\talpha default distinct flock fsync function limit null \n\tpg_shadow primary \n\n----------------------------- Log Message -----------------------------\n\nRemove unused TODO.detail files.\n\n", "msg_date": "Tue, 4 Jul 2000 01:17:04 -0400 (EDT)", "msg_from": "Bruce Momjian - CVS <momjian>", "msg_from_op": true, "msg_subject": "pgsql/doc/TODO.detail (alpha default distinct flock fsync function\n\tlimit null pg_shadow primary)" }, { "msg_contents": "On Tue, 4 Jul 2000, Bruce Momjian - CVS wrote:\n\n> Removed Files:\n> \talpha default distinct flock fsync function limit null \n> \tpg_shadow primary \n\nWhat are we going to do with the flock? Remove it? I asked about this\nyesterday, I guess it's not useful anymore.\n\n-- \nPeter Eisentraut Sernanders vaeg 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Tue, 4 Jul 2000 07:39:47 -0400 (EDT)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: pgsql/doc/TODO.detail (alpha default distinct flock\n\tfsync function limit null pg_shadow primary)" }, { "msg_contents": "> On Tue, 4 Jul 2000, Bruce Momjian - CVS wrote:\n> \n> > Removed Files:\n> > \talpha default distinct flock fsync function limit null \n> > \tpg_shadow primary \n> \n> What are we going to do with the flock? Remove it? I asked about this\n> yesterday, I guess it's not useful anymore.\n\nWell, we have the postmaster pid thing working now. Is there an issue\nwith flock() I have forgotten?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 4 Jul 2000 10:04:53 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: pgsql/doc/TODO.detail (alpha default distinct flock\n\tfsync function limit null pg_shadow primary)" }, { "msg_contents": "[email protected] writes:\n> On Tue, 4 Jul 2000, Bruce Momjian - CVS wrote:\n>> Removed Files:\n>> alpha default distinct flock fsync function limit null \n>> pg_shadow primary \n\n> What are we going to do with the flock? Remove it? I asked about this\n> yesterday, I guess it's not useful anymore.\n\nIt still is, because the $PGDATA pid file only guards against starting\ntwo postmasters in the same data directory. You still need some lock\nagainst starting two postmasters on the same port number. The TCP\nsocket address can take care of itself, but if one or both postmasters\nis started without -i then there's got to be some interlock on the Unix\nsocket file.\n\nI don't much like depending on flock for that, since it isn't available\neverywhere. The only portable answer is to build a pid-containing\ninterlock file for each socket file, as discussed in the TODO item.\n\nIt was premature of Bruce to delete the flock TODO file, since there is\nstill a lot of undone stuff in there concerning both this issue and\nmoving the Unix-socket file to a safer place than /tmp.\n\n\t\t\tregards, tom lane\n\nPS: Hey Bruce, mail directly to your personal address\[email protected] has been bouncing since Sunday.\n", "msg_date": "Tue, 04 Jul 2000 15:39:23 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: pgsql/doc/TODO.detail (alpha default distinct flock fsync\n\tfunction limit null pg_shadow primary)" }, { "msg_contents": "Bruce Momjian writes:\n\n> > On Tue, 4 Jul 2000, Bruce Momjian - CVS wrote:\n> > \n> > > Removed Files:\n> > > \talpha default distinct flock fsync function limit null \n> > > \tpg_shadow primary \n> > \n> > What are we going to do with the flock? Remove it? I asked about this\n> > yesterday, I guess it's not useful anymore.\n> \n> Well, we have the postmaster pid thing working now. Is there an issue\n> with flock() I have forgotten?\n\nWe still have the locking code in there, together with the broken\nconfigure test. See my message from the other day (\"fcntl(F_SETLK)\") --\nshould we just remove all that stuff?\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Tue, 4 Jul 2000 21:58:06 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [COMMITTERS] pgsql/doc/TODO.detail (alpha default distinct flock\n\tfsync function limit null pg_shadow primary)" }, { "msg_contents": "Tom Lane writes:\n\n> but if one or both postmasters is started without -i then there's got\n> to be some interlock on the Unix socket file.\n> \n> I don't much like depending on flock for that, since it isn't available\n> everywhere. The only portable answer is to build a pid-containing\n> interlock file for each socket file, as discussed in the TODO item.\n\nBut the flock code isn't used because the configure test for it is broken,\nand has been broken ever since it was introduced AFAICT. It seems that we\nhave been relying on the mere existence of the socket file.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Wed, 5 Jul 2000 02:12:15 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [COMMITTERS] pgsql/doc/TODO.detail (alpha default distinct flock\n\tfsync function limit null pg_shadow primary)" }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n> Tom Lane writes:\n>> but if one or both postmasters is started without -i then there's got\n>> to be some interlock on the Unix socket file.\n>> \n>> I don't much like depending on flock for that, since it isn't available\n>> everywhere. The only portable answer is to build a pid-containing\n>> interlock file for each socket file, as discussed in the TODO item.\n\n> But the flock code isn't used because the configure test for it is broken,\n> and has been broken ever since it was introduced AFAICT. It seems that we\n> have been relying on the mere existence of the socket file.\n\nOooh, no kidding? That explains why we're still hearing complaints\nabout the postmaster failing to start up when there's a socket file\nleft over from a previous run: the code that's supposed to delete an\nold socket file is part of the #ifdef HAVE_FCNTL_SETLK path.\n\n(Tries it out ... sure enough, it's broken ...)\n\nThe flock is not really needed to protect the port number; it's there\nto prevent a second postmaster from deleting the socket file that\nbelongs to a still-active old postmaster. But if you have no delete\nlogic at all, you can't cope with a leftover socket file.\n\nThe flock code *did* work at one time; I recall testing it. Evidently\nsomeone broke the configure test for it later on.\n\nI think the shortest path to a solution is to fix the configure test,\nunless you have the ambition to tackle setting up a set of lock files\nfor port numbers --- which'd require resolving such thorny questions\nas where to keep the lock files. (/tmp is right out, IMHO.)\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 20:59:05 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [COMMITTERS] pgsql/doc/TODO.detail (alpha default distinct flock\n\tfsync function limit null pg_shadow primary)" }, { "msg_contents": "I will add the flock TODO.detail file for people to review.\n\n[ Charset ISO-8859-1 unsupported, converting... ]\n> Bruce Momjian writes:\n> \n> > > On Tue, 4 Jul 2000, Bruce Momjian - CVS wrote:\n> > > \n> > > > Removed Files:\n> > > > \talpha default distinct flock fsync function limit null \n> > > > \tpg_shadow primary \n> > > \n> > > What are we going to do with the flock? Remove it? I asked about this\n> > > yesterday, I guess it's not useful anymore.\n> > \n> > Well, we have the postmaster pid thing working now. Is there an issue\n> > with flock() I have forgotten?\n> \n> We still have the locking code in there, together with the broken\n> configure test. See my message from the other day (\"fcntl(F_SETLK)\") --\n> should we just remove all that stuff?\n> \n> \n> -- \n> Peter Eisentraut Sernanders v?g 10:115\n> [email protected] 75262 Uppsala\n> http://yi.org/peter-e/ Sweden\n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 4 Jul 2000 22:54:36 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [COMMITTERS] pgsql/doc/TODO.detail (alpha default distinct flock\n\tfsync function limit null pg_shadow primary)" }, { "msg_contents": "> [email protected] writes:\n> > On Tue, 4 Jul 2000, Bruce Momjian - CVS wrote:\n> >> Removed Files:\n> >> alpha default distinct flock fsync function limit null \n> >> pg_shadow primary \n> \n> > What are we going to do with the flock? Remove it? I asked about this\n> > yesterday, I guess it's not useful anymore.\n> \n> It still is, because the $PGDATA pid file only guards against starting\n> two postmasters in the same data directory. You still need some lock\n> against starting two postmasters on the same port number. The TCP\n> socket address can take care of itself, but if one or both postmasters\n> is started without -i then there's got to be some interlock on the Unix\n> socket file.\n> \n> I don't much like depending on flock for that, since it isn't available\n> everywhere. The only portable answer is to build a pid-containing\n> interlock file for each socket file, as discussed in the TODO item.\n> \n> It was premature of Bruce to delete the flock TODO file, since there is\n> still a lot of undone stuff in there concerning both this issue and\n> moving the Unix-socket file to a safer place than /tmp.\n\nOK, I have re-added TODO.detail, and re-added the TODO item.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 5 Jul 2000 00:39:05 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: pgsql/doc/TODO.detail (alpha default distinct flock\n\tfsync function limit null pg_shadow primary)" } ]
[ { "msg_contents": "\nSorry about that; I didn't check the 'to' address before sending the patch...\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 04 Jul 2000 15:24:37 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: AW: AW: AW: Modified pg_dump & new pg_restore\n need testing..." } ]
[ { "msg_contents": "\nI am wondering if anyone can see an immediate problem in creating a function:\n\n heap_create_oid\n\nwhich is just like heap_create, but takes an oid arg. This oid (assuming\nit's free) would be used to create the object.\n\nThis could then bes called from a new \n\n heap_create_with_catalog_oid\n\nwhich in turn will be called from \n\n inv_create_oid\n\nto ultimately allow pg_restore to recreate BLOBs, via lo_create_oid.\n\nThis plan was developed by simply looking at the LO code in total\nisolation, so I am aware it could be totally flawed for the rest of the DB,\nbut I would appreciate some input...\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 04 Jul 2000 16:38:26 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "heap_create with OID?" }, { "msg_contents": "\nWould this allow you to create a tuble with a user-specified oid?\n\nWe definitely need this functionality for all tables, not just large\nobjects.\n\nHowever you won't be able to discover if the oid is free or not. You\ncould end up with duplicate oids. Thus anybody who uses the feature\nshould know what they're doing.\n\nPhilip Warner wrote:\n> \n> I am wondering if anyone can see an immediate problem in creating a function:\n> \n> heap_create_oid\n> \n> which is just like heap_create, but takes an oid arg. This oid (assuming\n> it's free) would be used to create the object.\n> \n> This could then bes called from a new\n> \n> heap_create_with_catalog_oid\n> \n> which in turn will be called from\n> \n> inv_create_oid\n> \n> to ultimately allow pg_restore to recreate BLOBs, via lo_create_oid.\n> \n> This plan was developed by simply looking at the LO code in total\n> isolation, so I am aware it could be totally flawed for the rest of the DB,\n> but I would appreciate some input...\n", "msg_date": "Tue, 04 Jul 2000 16:44:32 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "Philip Warner <[email protected]> writes:\n> I am wondering if anyone can see an immediate problem in creating a function:\n> heap_create_oid\n> which is just like heap_create, but takes an oid arg. This oid (assuming\n> it's free) would be used to create the object.\n\nHow exactly do you propose to determine that the OID is free?\nIf the requested OID is greater than the current OID counter, how\nwill you prevent a future conflict?\n\nThere's no inherent problem in forcing a particular choice of OID;\ninitdb does it on every run. The trick is to avoid a conflict of\nOID assignments.\n\nMy own feeling is that the current LO setup is fundamentally flawed\nby its reliance on specific OID values, and that the right answer is\nto find a way to avoid that. contrib/lo might provide some food for\nthought here (although it's clearly not the whole answer either).\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 02:57:36 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID? " }, { "msg_contents": "At 16:44 4/07/00 +1000, Chris Bitmead wrote:\n>\n>Would this allow you to create a tuble with a user-specified oid?\n\nTo be honest, I have no idea - like I said, I just looked at the LO code\nand heap.c. But I believe it will let me create a lo with the right OID.\n\n>\n>However you won't be able to discover if the oid is free or not. You\n>could end up with duplicate oids. Thus anybody who uses the feature\n>should know what they're doing.\n\nOK; then it needs to be done by pg_restore ONLY when a full restore is\nbeing done. pg_dump already has a crude way of setting the max OID at the\nstart of a restore; all I need to do is use that process when restoring blobs.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 04 Jul 2000 17:33:43 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "At 02:57 4/07/00 -0400, Tom Lane wrote:\n>Philip Warner <[email protected]> writes:\n>> I am wondering if anyone can see an immediate problem in creating a\nfunction:\n>> heap_create_oid\n>> which is just like heap_create, but takes an oid arg. This oid (assuming\n>> it's free) would be used to create the object.\n>\n>How exactly do you propose to determine that the OID is free?\n\nBased on the comments so far, I don't.\n\npg_dump already allows a 'with-oid' option for table dumping & loading, and\nthe way it gets around the problem is to save the max oid at dump time,\nthen set the max oid at restore time. I am only seeking a way of doing a\nvalid backup/restore which includes BLOBS; so even if I say to the user\nthat BLOBs can only be restored as part of a full restore, then I think\nit's a lot better than the current system.\n\n\n>There's no inherent problem in forcing a particular choice of OID;\n>initdb does it on every run. The trick is to avoid a conflict of\n>OID assignments.\n\nOK. This is good news. The way pg_dump sets the max oid is:\n\n CREATE TABLE pgdump_oid (dummy int4);\n COPY pgdump_oid WITH OIDS FROM stdin;\n 1282249 0\n \\.\n\nwhere 1282249 is the max oid you want to set.\n\nIt's pretty crude, but I assume effective. What would be very nice is if I\ncould 'reserve' the OIDs of the to-be-created BLOBS without having to\ncreate rows/tables etc, and without affecting the max oid. But that is\nprobably pointless if a decent BLOB implementation is coming along.\n\n\n>My own feeling is that the current LO setup is fundamentally flawed\n\nYou will get no disagreement from me.\n\n\n>the right answer is\n>to find a way to avoid that.\n\nI'm reluctant to do too much in this area until the capabilities of TOAST\nare sorted out...\n\nISTM that, even if a new BLOB architecture comes along, some kind of\nmigration utility will be needed.\n\nDoes this all sound reasonable?\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 04 Jul 2000 17:43:43 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: heap_create with OID? " }, { "msg_contents": "Tom Lane wrote:\n> My own feeling is that the current LO setup is fundamentally flawed\n> by its reliance on specific OID values, and that the right answer is\n> to find a way to avoid that. contrib/lo might provide some food for\n> thought here (although it's clearly not the whole answer either).\n\n I have some ideas to replace the entire LO handling by\n something completely different. More compatible with the\n Oracle way of handling large objects. That is, that on\n INSERT/UPDATE someone uses a function to place an LO\n reference into the tuple. Then having a new interface in\n libpq, that works like file I/O on the references that appear\n in a result set. To open them for writing, the user must have\n selected them for update, otherwise he can open them for\n reading. The file I/O itself can be based on the fastpath\n interface.\n\n The LO's follow Oracles copy semantic, so if someone does\n INSERT ... SELECT, the LO gets duplicated.\n\n All LO's for one base table can be stored in one big,\n segmented external heap (more or less like toast values). The\n system would automatically know when objects are obsolete.\n\n An INSERT operation might then return a result set as if\n someone did a SELECT FOR UPDATE of exactly what he just\n inserted. So he immediately has access to the LO references\n to place the values.\n\n Don't know yet how to interface it from psql - but let me\n sleep another night or so over it.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Tue, 4 Jul 2000 14:41:23 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "> \n> Would this allow you to create a tuble with a user-specified oid?\n> \n> We definitely need this functionality for all tables, not just large\n> objects.\n> \n> However you won't be able to discover if the oid is free or not. You\n> could end up with duplicate oids. Thus anybody who uses the feature\n> should know what they're doing.\n\nYes, this type of thing is clearly needed. We need to be able to UPDATE\nthe oid column too.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 4 Jul 2000 10:00:08 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "At 10:00 4/07/00 -0400, Bruce Momjian wrote:\n>\n>Yes, this type of thing is clearly needed. We need to be able to UPDATE\n>the oid column too.\n>\n\nUpdating OIDs sounds like a *very* scary thing to do. My motivation for\nheap_create_oid is very specific - I want to be able to do a valid database\nbackup/restore, and need to handle the current (bad) BLOB implementation. I\nwould sincerely hope that (a) heap_create_oid would be marked in some way\nas 'internal use only', and (b) go away in 7.2 or 7.3 when TOASTED BLOBs\ncome along...\n\nFWIW, I now have a working lo_create_oid, that does what is expected.\n\nWhen (and if) the new pg_restore is accepted, I'll probably have a look at\nimplementing BLOB backup, but there might be some very justifiable\nresistance to heap_create_oid...so before I go much further I'd appreciate\na barometer reading on the likely acceptability of heap_create_oid.\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 00:57:16 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "At 14:41 4/07/00 +0200, Jan Wieck wrote:\n>Tom Lane wrote:\n>\n> I have some ideas to replace the entire LO handling by\n> something completely different. More compatible with the\n> Oracle way of handling large objects. That is, that on\n> INSERT/UPDATE someone uses a function to place an LO\n> reference into the tuple. Then having a new interface in\n> libpq, that works like file I/O on the references that appear\n> in a result set. To open them for writing, the user must have\n> selected them for update, otherwise he can open them for\n> reading. \n\nIt might be worth looking at the SQL standard and it's 'locators' for BLOB\ndata. They sound a lot like this, but may have some requirements that\naffect your design.\n\n\n> The file I/O itself can be based on the fastpath\n> interface.\n\nWhat's the 'fastpath interface'?\n\n\n> The LO's follow Oracles copy semantic, so if someone does\n> INSERT ... SELECT, the LO gets duplicated.\n\nGreat. The standard seems (by default) to treat them as just another\n'value', which this is consistent with. It may be relevant to mention that\nthe standard also allows substr, ||, etc to work on BLOBs.\n\n\n> An INSERT operation might then return a result set as if\n> someone did a SELECT FOR UPDATE of exactly what he just\n> inserted. So he immediately has access to the LO references\n> to place the values.\n>\n> Don't know yet how to interface it from psql - but let me\n> sleep another night or so over it.\n\nFWIW, Dec/Rdb uses a kind of cursor for the BLOB data. The row insert must\nbe done in a normal insert cursor, then another cursor is opened to write\nthe BLOB data. It's very cumbersome. But if you use an insert cursor for\nyour table rows, then at least you do have a context available for BLOB\ninsertion.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 01:43:03 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "> \n> > The file I/O itself can be based on the fastpath\n> > interface.\n> \n> What's the 'fastpath interface'?\n> \n\nThe interface between PG backend and libpq that is for example used for \nCOPY command. (Simplification) this intarface directly connect two routines\nbetween backend and frondend without standard \"the path of query\". \n\nLook at src/backend/tcop/fastpath.c\n\n\t\t\t\t\tKarel\n\n", "msg_date": "Tue, 4 Jul 2000 18:06:02 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "> Updating OIDs sounds like a *very* scary thing to do. My motivation for\n> heap_create_oid is very specific - I want to be able to do a valid database\n> backup/restore, and need to handle the current (bad) BLOB implementation. I\n> would sincerely hope that (a) heap_create_oid would be marked in some way\n> as 'internal use only', and (b) go away in 7.2 or 7.3 when TOASTED BLOBs\n> come along...\n\nIf you accidentally delete a row with a specific oid, how do you re-add\nthe row?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 4 Jul 2000 12:14:41 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "Philip Warner <[email protected]> writes:\n> The way pg_dump sets the max oid is:\n\n> CREATE TABLE pgdump_oid (dummy int4);\n> COPY pgdump_oid WITH OIDS FROM stdin;\n> 1282249 0\n> \\.\n\n> where 1282249 is the max oid you want to set.\n\n> It's pretty crude, but I assume effective.\n\nAh, right. IIRC there is a check in COPY to set the OID counter to\nmax(current OID counter, user-supplied OID). So you could use this\ntechnique to install all the BLOBs with the right OIDs, as long as\nyou were careful to load them in OID order: just load another row\ninto the temp table each time you need to advance the OID counter.\nKinda klugy, but no backend changes needed.\n\nBTW I'd recommend making pgdump_oid a TEMP table, so that it'll go away\nby itself.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 12:48:22 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID? " }, { "msg_contents": "> Philip Warner <[email protected]> writes:\n> > The way pg_dump sets the max oid is:\n> \n> > CREATE TABLE pgdump_oid (dummy int4);\n> > COPY pgdump_oid WITH OIDS FROM stdin;\n> > 1282249 0\n> > \\.\n> \n> > where 1282249 is the max oid you want to set.\n> \n> > It's pretty crude, but I assume effective.\n> \n> Ah, right. IIRC there is a check in COPY to set the OID counter to\n> max(current OID counter, user-supplied OID). So you could use this\n> technique to install all the BLOBs with the right OIDs, as long as\n> you were careful to load them in OID order: just load another row\n> into the temp table each time you need to advance the OID counter.\n> Kinda klugy, but no backend changes needed.\n> \n> BTW I'd recommend making pgdump_oid a TEMP table, so that it'll go away\n> by itself.\n\nDone. Renamed pgdump_oid to pg_dump_oid too. We didn't have TEMPORARY\ntables at the time I did this. The temporary allows multiple concurrent\nloads because there is no table name conflict.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 4 Jul 2000 12:58:40 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "Philip Warner wrote:\n\n> Updating OIDs sounds like a *very* scary thing to do.\n\nIt's scary, but useful in some specific situations. Restoring lost data\nis an obvious one. There would be applications in writing certain forms\nof database replication schemes. There is also a technique in object\ndatabases where the client side code has to know what the oid for an\nobject will be before it is actually added to the database.\n", "msg_date": "Wed, 05 Jul 2000 09:42:38 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "At 09:42 5/07/00 +1000, Chris Bitmead wrote:\n>Philip Warner wrote:\n>\n>> Updating OIDs sounds like a *very* scary thing to do.\n>\n>It's scary, but useful in some specific situations. Restoring lost data\n>is an obvious one. There would be applications in writing certain forms\n>of database replication schemes. There is also a technique in object\n>databases where the client side code has to know what the oid for an\n>object will be before it is actually added to the database.\n\nNone of these strictly involve updating OIDs; they seem to be choosing OIDs\nat creation time, which is a very different thing.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 15:12:56 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "At 12:48 4/07/00 -0400, Tom Lane wrote:\n>Philip Warner <[email protected]> writes:\n>\n>Ah, right. IIRC there is a check in COPY to set the OID counter to\n>max(current OID counter, user-supplied OID). So you could use this\n>technique to install all the BLOBs with the right OIDs, as long as\n>you were careful to load them in OID order: just load another row\n>into the temp table each time you need to advance the OID counter.\n>Kinda klugy, but no backend changes needed.\n\nThis might work...the idea would presumably be to load the BLOBs before the\nschema, since the OIDs of the BLOBS might be low, and hence conflict could\noccur if many tables were defined after the first blob was loaded in the\noriginal DB.\n\nSo:\n\n1. createdb\n2. load all blobs\n3. Do remaining restore steps\n\nThe only real problem (and this is in pg_dump as well), is if a table is\ndumped with OIDs and the OIDs conflict with automatically generated ones\nfor the schema. Although this may not be an issue: how unique are OIDs,\nanyway? Will corruption of any kind occur in the above case (even with the\nold pg_dump)?\n\nI guess I'll throw out lo_create_oi, heap_create_oid & related code. I\nguess it taught me a little more about PG internals.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 15:24:34 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: heap_create with OID? " }, { "msg_contents": "Philip Warner <[email protected]> writes:\n> This might work...the idea would presumably be to load the BLOBs before the\n> schema, since the OIDs of the BLOBS might be low, and hence conflict could\n> occur if many tables were defined after the first blob was loaded in the\n> original DB.\n\n> 1. createdb\n> 2. load all blobs\n> 3. Do remaining restore steps\n\nIf you're reloading everything in OID order, then blobs need no special\nordering treatment ;-)\n\n> The only real problem (and this is in pg_dump as well), is if a table is\n> dumped with OIDs and the OIDs conflict with automatically generated ones\n> for the schema. Although this may not be an issue: how unique are OIDs,\n> anyway? Will corruption of any kind occur in the above case (even with the\n> old pg_dump)?\n\nIn theory OIDs are unique (within one database anyway). There are a\ncouple of ways that theory can fail:\n\n1. User-supplied OIDs (via COPY WITH OIDs) might conflict.\n\n2. Run the system long enough, the OID counter will wrap around and\n start generating already-used numbers.\n\nThe next question is what happens if we do have duplicate OIDs. AFAIK,\nif the duplicate OIDs are for different kinds of entities (eg, rows in\ndifferent tables) the answer is \"no problem\". Duplicate OIDs for, say,\ntwo tables would be disastrous --- but the unique index on pg_class.oid\nshould prevent you from creating duplicates. I *think* that 7.0 has\nunique indexes on every system table where it's important to prevent\nduplicate OIDs. (If not, it's an easily corrected omission; anyone want\nto run through the tables and double-check?)\n\nOIDs aren't magic, they're just another number. The system depends on\nthe assumption that OIDs are unique keys for certain system tables, and\nit enforces (or should enforce) this assumption. Applications might\ndepend on the assumption that OIDs are unique keys for their own tables;\nif so they can and should enforce that assumption with unique indexes.\nBut there's no hidden mechanism that will make the system go belly-up\njust because there are identical OIDs floating around.\n\n\t\t\tregards, tom lane\n\nPS: I believe the \"comment\" support will misbehave if there are\nduplicate OIDs in different system tables, since it assumes that an OID\nis sufficient to identify any database object regardless of type.\nBut the comment stuff is not exactly mission-critical...\n", "msg_date": "Wed, 05 Jul 2000 02:14:04 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID? " }, { "msg_contents": "At 02:14 5/07/00 -0400, Tom Lane wrote:\n>\n>> 1. createdb\n>> 2. load all blobs\n>> 3. Do remaining restore steps\n>\n>If you're reloading everything in OID order, then blobs need no special\n>ordering treatment ;-)\n\nIt's actually a bit more complex than that, in that for efficiency I put\nindexes, triggers, and ACLs at the end, just after date, which once more\nintroduces the possibility of a conflict. \n\nAlso, the ordering is fine, but with data loading (without OID), it still\nmay fail (unless I put the data after the BLOBs), so a strict OID order\nwon't work without OID based data load. And again, that would also need to\nbe done in OID order, which is a big ask.\n\nPersonally, I dislike the the concept of being able to set OID on some\nthings but not on others. I'd prefer a lower level backup interface to\nallow creation of dumped entities with specified OIDs (ie. tables, indexes\netc). That way we get a database that is indistinguishable from the\noriginal. This might wait to see what backup techniques are possible after\nWAL & slot reuse.\n\nThis is not an argument for heap_create_oid, unless it is extended to *all*\ndatabase entities, which some other people might favour. I'd be interested\nin other people's thoughts on this, since I am somewhat out of my depth!\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 18:55:22 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: heap_create with OID? " }, { "msg_contents": "> > The only real problem (and this is in pg_dump as well), is if a table is\n> > dumped with OIDs and the OIDs conflict with automatically generated ones\n> > for the schema. Although this may not be an issue: how unique are OIDs,\n> > anyway? Will corruption of any kind occur in the above case (even with the\n> > old pg_dump)?\n> \n> In theory OIDs are unique (within one database anyway). There are a\n> couple of ways that theory can fail:\n> \n> 1. User-supplied OIDs (via COPY WITH OIDs) might conflict.\n> \n> 2. Run the system long enough, the OID counter will wrap around and\n> start generating already-used numbers.\n\nSeems we should throw an error on rollover, or at least jump from 0 to\n17k to skip the system oid's. That zero oid is going to cause a\nproblem, perhaps. Of course, it is has never happened, but maybe we\nshould do it for people's sanity.\n\n> \n> The next question is what happens if we do have duplicate OIDs. AFAIK,\n> if the duplicate OIDs are for different kinds of entities (eg, rows in\n> different tables) the answer is \"no problem\". Duplicate OIDs for, say,\n> two tables would be disastrous --- but the unique index on pg_class.oid\n> should prevent you from creating duplicates. I *think* that 7.0 has\n> unique indexes on every system table where it's important to prevent\n> duplicate OIDs. (If not, it's an easily corrected omission; anyone want\n> to run through the tables and double-check?)\n\nIt sure does, or at least ones that are used by the cache.\n\n> OIDs aren't magic, they're just another number. The system depends on\n> the assumption that OIDs are unique keys for certain system tables, and\n> it enforces (or should enforce) this assumption. Applications might\n> depend on the assumption that OIDs are unique keys for their own tables;\n> if so they can and should enforce that assumption with unique indexes.\n> But there's no hidden mechanism that will make the system go belly-up\n> just because there are identical OIDs floating around.\n> \n> \t\t\tregards, tom lane\n> \n> PS: I believe the \"comment\" support will misbehave if there are\n> duplicate OIDs in different system tables, since it assumes that an OID\n> is sufficient to identify any database object regardless of type.\n> But the comment stuff is not exactly mission-critical...\n\nAgreed.\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 6 Jul 2000 10:16:28 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n>> 2. Run the system long enough, the OID counter will wrap around and\n>> start generating already-used numbers.\n\n> Seems we should throw an error on rollover, or at least jump from 0 to\n> 17k to skip the system oid's. That zero oid is going to cause a\n> problem, perhaps.\n\nThrowing an error will definitely not do --- that'd mean your whole\ninstallation comes to a screeching halt as soon as you hit 4G OIDs.\nThat cure is way worse than the disease...\n\nWe do need to avoid generating a zero OID. I vaguely recall having\nseen some code somewhere that did that, but it was pretty crude/ugly,\nand I'm not even sure it was in the OID generator rather than in one\nspecific routine that used OIDs :-(. (What I seem to recall having\nseen was some code that called newoid() a second time if it got a zero.\nThis is bletcherous of course.)\n\nAs long as we've gotta avoid zero anyway, skipping over the whole\nsystem-reserved range doesn't sound like a bad idea, even though it's\nnot going to guarantee anything. But it's got to be done at the bottom\nlevel of the counter-incrementing logic, not higher up.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 06 Jul 2000 10:55:41 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID? " }, { "msg_contents": "Bruce Momjian wrote:\n\n> Seems we should throw an error on rollover, or at least jump from 0 to\n> 17k to skip the system oid's. That zero oid is going to cause a\n> problem, perhaps. Of course, it is has never happened, but maybe we\n> should do it for people's sanity.\n\nSkipping the system oids is no guarantee things will keep working on\nrollover. In fact I'm not sure it would help at all. The only way would\nbe to compress the existing oids via a utility or go to 64 bit. Has\nanyone actually run out?\n\n>\n", "msg_date": "Fri, 07 Jul 2000 10:46:35 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "Tom Lane wrote:\n> Throwing an error will definitely not do --- that'd mean your whole\n> installation comes to a screeching halt as soon as you hit 4G OIDs.\n> That cure is way worse than the disease...\n\nIs there really an alternative? There is no guarantee things will keep\nworking after a wrap around. Some applications might, some might not\ndepending on how they work. Isn't throwing an error better than\ncontinuing on, possibly doing the wrong thing?\n", "msg_date": "Fri, 07 Jul 2000 11:05:15 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID?" }, { "msg_contents": "Chris Bitmead <[email protected]> writes:\n> Tom Lane wrote:\n>> Throwing an error will definitely not do --- that'd mean your whole\n>> installation comes to a screeching halt as soon as you hit 4G OIDs.\n>> That cure is way worse than the disease...\n\n> Is there really an alternative? There is no guarantee things will keep\n> working after a wrap around. Some applications might, some might not\n> depending on how they work. Isn't throwing an error better than\n> continuing on, possibly doing the wrong thing?\n\nNo. Throwing an error (refusing to wrap the counter around) means that\nyou will refuse to create any more rows until after the dbadmin dumps,\ninitdbs, reloads. (If you think that's bad, consider also that in its\ncurrent form pg_dumpall would fail under those conditions, meaning you\ncouldn't even dump.) Stopping the whole DB dead in the water cannot\nbe a better answer than *anything*.\n\nThe main point of my prior post is that a well-designed DB will have\nunique indexes to guarantee uniqueness of anything it needs to assume\nis unique ... such as OIDs in particular tables. The presence of\nthose indexes might mean that individual transactions fail (when they\ntry to use a duplicate OID value), and that's surely nasty. But it's\nnot as nasty as forcibly shutting down ALL inserts, which is what\nthrowing an error implies.\n\nObviously this isn't a perfect, no-need-for-improvement approach.\nWe need 64-bit OIDs or the ability not to assign OIDs to tables that\ndon't need 'em before we can really say \"problem solved\". But throwing\nan error on wraparound is *worse* than the laissez-faire approach we\ntake now. It's at least possible to design a DB to survive wraparound\nnow; if we throw an error survival is not possible.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 06 Jul 2000 23:20:42 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: heap_create with OID? " } ]
[ { "msg_contents": "At 03:23 4/07/00 -0400, Tom Lane wrote:\n>IMHO we'd be damned fools to\n>ignore his advice completely. Sticking your head in the sand is not\n>a good defense mechanism.\n\nFWIW, I think the disclaimer could be strengthened to protect people who\nsell the PostgreSQL CD, and people who offer it on servers, and people who\napply patches from other people (who may not themselves be contributors or\ndevelopers). It's just a box of worms - no source can be too open, and no\nindemnity can be too strong.\n\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 04 Jul 2000 17:59:14 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [GENERAL] Re: [HACKERS] proposed improvements to\n\tPostgreSQL license" } ]
[ { "msg_contents": "At 02:57 4/07/00 -0400, Tom Lane wrote:\n>Philip Warner <[email protected]> writes:\n>> I am wondering if anyone can see an immediate problem in creating a\nfunction:\n>> heap_create_oid\n>> which is just like heap_create, but takes an oid arg. This oid (assuming\n>> it's free) would be used to create the object.\n\n>There's no inherent problem in forcing a particular choice of OID;\n>initdb does it on every run. The trick is to avoid a conflict of\n>OID assignments.\n\nI've modified the relevant routines up to and including defining\nlo_create_oid, and was wondering what the best way to procede in testing\nthe new routine might be...is there a document that decribes modifying the\nstandard set of functions and adding lo_create_oid?\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Tue, 04 Jul 2000 19:19:07 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": true, "msg_subject": "Re: heap_create with OID? " } ]
[ { "msg_contents": "Can someone comment on this? Why doesn't OID work in foreign keys. \nSeems this is a bug.\n\n> First off I want to say thank your for creating a sorely needed reference and\n> sharing it with the online world via the PDF document.\n> \n> I do have one minor complaint though. I'm learning Postgres from your book\n> although I'm familiar with the basics of SQL. I was at first excited then\n> discouraged by the OID section of your book. I think a much needed\n> addition to the OID limitations section is that the OID field can not be\n> referenced in a foreign key. IE\n> \n> CREATE TABLE t1( i INTEGER);\n> CREATE TABLE t2( id OID REFERENCES t1(oid));//dose not work\n> \n> You use the OID is a foreign key of sorts in the example, but you never qualify\n> that using the foreign key constraint system in Postgres will not allow you to\n> automate this. I realize the OID section comes long before the constraints\n> chapter, but a quick qualification would still make things clearer. Without\n> the ability to use the OID field as a reference for foreign keys, it's\n> usefulness is very limited.\n> \n> Keep up the good work, I can't wait to buy the finished product!\n> \n> _________________________\n> Richard Rowell\n> [email protected]\n> \n> Perl is the unix world's equivalent of VB - and just as much of an abomination.\n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 4 Jul 2000 09:56:51 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: PDF book July 2, 2000 version" }, { "msg_contents": "Bruce Momjian wrote:\n> Can someone comment on this? Why doesn't OID work in foreign keys.\n> Seems this is a bug.\n\n The RI triggers check for attribute existance by looking up\n in the relations descriptor via SPI_fnumber(). Since OID is a\n system attribute, it returns SPI_ERROR_NOATTRIBUTE and that\n causes the error.\n\n Later on, I'm not sure that it is possible to use SPI\n functions to obtain the OID of a given trigger tuple. So it'd\n be a little more to fix it.\n\n\nJan\n\n>\n> > First off I want to say thank your for creating a sorely needed reference and\n> > sharing it with the online world via the PDF document.\n> >\n> > I do have one minor complaint though. I'm learning Postgres from your book\n> > although I'm familiar with the basics of SQL. I was at first excited then\n> > discouraged by the OID section of your book. I think a much needed\n> > addition to the OID limitations section is that the OID field can not be\n> > referenced in a foreign key. IE\n> >\n> > CREATE TABLE t1( i INTEGER);\n> > CREATE TABLE t2( id OID REFERENCES t1(oid));//dose not work\n> >\n> > You use the OID is a foreign key of sorts in the example, but you never qualify\n> > that using the foreign key constraint system in Postgres will not allow you to\n> > automate this. I realize the OID section comes long before the constraints\n> > chapter, but a quick qualification would still make things clearer. Without\n> > the ability to use the OID field as a reference for foreign keys, it's\n> > usefulness is very limited.\n> >\n> > Keep up the good work, I can't wait to buy the finished product!\n> >\n> > _________________________\n> > Richard Rowell\n> > [email protected]\n> >\n> > Perl is the unix world's equivalent of VB - and just as much of an abomination.\n> >\n>\n>\n> --\n> Bruce Momjian | http://www.op.net/~candle\n> [email protected] | (610) 853-3000\n> + If your life is a hard drive, | 830 Blythe Avenue\n> + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n>\n\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Tue, 4 Jul 2000 17:08:35 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: Re: PDF book July 2, 2000 version" }, { "msg_contents": "Added to TODO:\n\n\t* Allow oid to act as a foreign key\n\n\n> Bruce Momjian wrote:\n> > Can someone comment on this? Why doesn't OID work in foreign keys.\n> > Seems this is a bug.\n> \n> The RI triggers check for attribute existance by looking up\n> in the relations descriptor via SPI_fnumber(). Since OID is a\n> system attribute, it returns SPI_ERROR_NOATTRIBUTE and that\n> causes the error.\n> \n> Later on, I'm not sure that it is possible to use SPI\n> functions to obtain the OID of a given trigger tuple. So it'd\n> be a little more to fix it.\n> \n> \n> Jan\n> \n> >\n> > > First off I want to say thank your for creating a sorely needed reference and\n> > > sharing it with the online world via the PDF document.\n> > >\n> > > I do have one minor complaint though. I'm learning Postgres from your book\n> > > although I'm familiar with the basics of SQL. I was at first excited then\n> > > discouraged by the OID section of your book. I think a much needed\n> > > addition to the OID limitations section is that the OID field can not be\n> > > referenced in a foreign key. IE\n> > >\n> > > CREATE TABLE t1( i INTEGER);\n> > > CREATE TABLE t2( id OID REFERENCES t1(oid));//dose not work\n> > >\n> > > You use the OID is a foreign key of sorts in the example, but you never qualify\n> > > that using the foreign key constraint system in Postgres will not allow you to\n> > > automate this. I realize the OID section comes long before the constraints\n> > > chapter, but a quick qualification would still make things clearer. Without\n> > > the ability to use the OID field as a reference for foreign keys, it's\n> > > usefulness is very limited.\n> > >\n> > > Keep up the good work, I can't wait to buy the finished product!\n> > >\n> > > _________________________\n> > > Richard Rowell\n> > > [email protected]\n> > >\n> > > Perl is the unix world's equivalent of VB - and just as much of an abomination.\n> > >\n> >\n> >\n> > --\n> > Bruce Momjian | http://www.op.net/~candle\n> > [email protected] | (610) 853-3000\n> > + If your life is a hard drive, | 830 Blythe Avenue\n> > + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n> >\n> \n> \n> --\n> \n> #======================================================================#\n> # It's easier to get forgiveness for being wrong than for being right. #\n> # Let's break this rule - forgive me. #\n> #================================================== [email protected] #\n> \n> \n> \n\n\n-- \n Bruce Momjian | http://candle.pha.pa.us\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 6 Jul 2000 23:23:43 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Re: PDF book July 2, 2000 version" } ]
[ { "msg_contents": "Hi,\n\n while moving my development environment onto my notebook I\n found a little module that needed to go into contrib. It\n provides aggregates for variance and standard deviation for\n float8. Can be found now under contrib/statmath.\n\n While doing so I noticed that at least the fti Makefile has a\n bug. The MODULE_PATH in the .sql.in is replaced with .sql at\n the end instead of .so as it should be. Might have to do with\n the contrib cleanup done lately.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Tue, 4 Jul 2000 16:40:32 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": true, "msg_subject": "Statistical aggregates" }, { "msg_contents": "\nOn Tue, 4 Jul 2000, Jan Wieck wrote:\n\n> Hi,\n> \n> while moving my development environment onto my notebook I\n> found a little module that needed to go into contrib. It\n> provides aggregates for variance and standard deviation for\n> float8. Can be found now under contrib/statmath.\n> \n> While doing so I noticed that at least the fti Makefile has a\n> bug. The MODULE_PATH in the .sql.in is replaced with .sql at\n> the end instead of .so as it should be. Might have to do with\n> the contrib cleanup done lately.\n> \n\n Sorry, my faul. I fog '$@' instead '$<'. Nice on this bug is that \nI copy this to more contrib Makefiles..... grrrr :-((\n\n Well, I directly now correct it. The patch will after one hour...\n\n\t\t\t\tKarel\n\n", "msg_date": "Tue, 4 Jul 2000 16:56:26 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Statistical aggregates" }, { "msg_contents": "\nOn Tue, 4 Jul 2000, Karel Zak wrote:\n\n> \n> On Tue, 4 Jul 2000, Jan Wieck wrote:\n> \n> > Hi,\n> > \n> > while moving my development environment onto my notebook I\n> > found a little module that needed to go into contrib. It\n> > provides aggregates for variance and standard deviation for\n> > float8. Can be found now under contrib/statmath.\n> > \n> > While doing so I noticed that at least the fti Makefile has a\n> > bug. The MODULE_PATH in the .sql.in is replaced with .sql at\n> > the end instead of .so as it should be. Might have to do with\n> > the contrib cleanup done lately.\n> > \n> \n> Sorry, my faul. I fog '$@' instead '$<'. Nice on this bug is that \n\t\t\t\t ^^^^^^\n\t\t\tidiotism - it must be directly .so\n\n> I copy this to more contrib Makefiles..... grrrr :-((\n> \n> Well, I directly now correct it. The patch will after one hour...\n\n", "msg_date": "Tue, 4 Jul 2000 16:59:02 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Statistical aggregates" }, { "msg_contents": "On Tue, 4 Jul 2000, Jan Wieck wrote:\n\n> Hi,\n> \n> while moving my development environment onto my notebook I\n> found a little module that needed to go into contrib. It\n> provides aggregates for variance and standard deviation for\n> float8. Can be found now under contrib/statmath.\n> \n> While doing so I noticed that at least the fti Makefile has a\n> bug. The MODULE_PATH in the .sql.in is replaced with .sql at\n> the end instead of .so as it should be. Might have to do with\n> the contrib cleanup done lately.\n\n Already fixed in the attache. Can anyone apply it to CVS?\n\n\t\t\t\tKarel", "msg_date": "Tue, 4 Jul 2000 17:18:48 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Statistical aggregates" }, { "msg_contents": "Applied.\n\n> \n> On Tue, 4 Jul 2000, Jan Wieck wrote:\n> \n> > Hi,\n> > \n> > while moving my development environment onto my notebook I\n> > found a little module that needed to go into contrib. It\n> > provides aggregates for variance and standard deviation for\n> > float8. Can be found now under contrib/statmath.\n> > \n> > While doing so I noticed that at least the fti Makefile has a\n> > bug. The MODULE_PATH in the .sql.in is replaced with .sql at\n> > the end instead of .so as it should be. Might have to do with\n> > the contrib cleanup done lately.\n> \n> Already fixed in the attache. Can anyone apply it to CVS?\n> \n> \t\t\t\tKarel\n> \n> \t\t\t\t\nContent-Description: \n\n[ application/x-gzip is not supported, skipping... ]\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 4 Jul 2000 12:16:29 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PATCHES] Re: Statistical aggregates" }, { "msg_contents": "Karel Zak wrote:\n> \n> On Tue, 4 Jul 2000, Karel Zak wrote:\n> \n> > \n> > On Tue, 4 Jul 2000, Jan Wieck wrote:\n> > \n> > > Hi,\n> > > \n> > > while moving my development environment onto my notebook I\n> > > found a little module that needed to go into contrib. It\n> > > provides aggregates for variance and standard deviation for\n> > > float8. Can be found now under contrib/statmath.\n> > > \n> > > While doing so I noticed that at least the fti Makefile has a\n> > > bug. The MODULE_PATH in the .sql.in is replaced with .sql at\n> > > the end instead of .so as it should be. Might have to do with\n> > > the contrib cleanup done lately.\n> > > \n> > \n> > Sorry, my faul. I fog '$@' instead '$<'. Nice on this bug is that \n> \t\t\t\t ^^^^^^\n> \t\t\tidiotism - it must be directly .so\n\n I used $(NAME)$(DLSUFFIX) in my Makefile. Not all platforms\n use .so for dynamic loadable objects.\n\n> \n> > I copy this to more contrib Makefiles..... grrrr :-((\n> > \n> > Well, I directly now correct it. The patch will after one hour...\n> \n\n\nJan\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n", "msg_date": "Tue, 4 Jul 2000 18:46:09 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": true, "msg_subject": "Re: Statistical aggregates" }, { "msg_contents": "Shouldn't the new statistics routines go into the main tree?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 4 Jul 2000 12:59:17 -0400 (EDT)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Statistical aggregates" }, { "msg_contents": "\nOn Tue, 4 Jul 2000, Jan Wieck wrote:\n\n> > > Sorry, my faul. I fog '$@' instead '$<'. Nice on this bug is that \n> > \t\t\t\t ^^^^^^\n> > \t\t\tidiotism - it must be directly .so\n> \n> I used $(NAME)$(DLSUFFIX) in my Makefile. Not all platforms\n> use .so for dynamic loadable objects.\n> \n\n Yes. In the contrib it is already fixed. I used:\n\n$(SED) \"s|MODULE_PATHNAME|$(CONTRIB_MODDIR)/$(@:.sql=$(DLSUFFIX))|\" < $< >$@\n\n I hope that it will right.\n\n\t\t\t\t\tKarel\n\n", "msg_date": "Tue, 4 Jul 2000 19:26:43 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Statistical aggregates" } ]
[ { "msg_contents": "\nOkay, from seeing the responses so far on the list, I'm not the only one\nthat has issues with the whole \"juristiction of virginia\" issue *or* the\n\"slam this copyright in ppls faces\" ... I do like the part in BOLD about\n\"ANY DEVELOPER\" instead of just the \"UNIVERSITY OF CALIFORNIA\" ... but I\nconsider that an appendum/extension of what is already stated ...\n\nIs the following more palatable to those of us that aren't US citizens? \n\nThe only part that I believe at least one person had an issue with was:\n\n\"Any person who contributes or submits any modification or other change to\nthe PostgreSQL software or documentation grants irrevocable,\nnon-exclusive, worldwide permission, without charge, to use, copy, further\nmodify and distribute the same under the terms of this license.\"\n\nQuite frankly, all I'm reading into this paragraph is that once committed,\nJan (as a recent example) couldn't come along and pull out all his TOAST\nchanges ... could you imagine the hell that would wreak were he (or anyone\nelse) were to pull crucial changes after others have built upon it?\n\nThe only change in this is the \"Juristiction\" para is removed ... I've\nread this over several times now, and personal feel that all its doing is\nextending the existing copyright to cover *ALL* developers, and not just\nthe \"UNIVERITY OF CALIFORNIA\" ones ...\n\nIf someone wants to point out something I have missed, I'm more then open\nto re-reading it again ;)\n\nI consider it an appendum to the existing copyright ... I don't know, does\nthat make it any less BSD/open?\n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n====================================================================\n\nPostgreSQL Data Base Management System (formerly known as Postgres95)\n\nThis directory contains the _______ release of PostgreSQL, as well as\nvarious post-release patches in the patches directory. See INSTALL for\nthe installation notes and HISTORY for the changes.\n\nWe also have a WWW home page located at: http://www.postgreSQL.org\n\n-------------------------\n\nPostgreSQL is not public domain software. It is copyrighted by the\nUniversity of California but may be used according to the following\nlicensing terms:\n\nPOSTGRES95 Data Base Management System (formerly known as Postgres, then\nas Postgres95).\n\nCopyright (c) 1994-8 Regents of the University of California\n\nPermission to use, copy, modify, and distribute this software and its\ndocumentation for any purpose, without fee, and without a written\nagreement is hereby granted, provided that the above copyright notice and\nthis paragraph and the following two paragraphs appear in all copies.\n\nIN NO EVENT SHALL THE UNIVERSITY OF CALIFORNIA BE LIABLE TO ANY PARTY FOR\nDIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING\nLOST PROFITS, ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS\nDOCUMENTATION, EVEN IF THE UNIVERSITY OF CALIFORNIA HAS BEEN ADVISED OF\nTHE POSSIBILITY OF SUCH DAMAGE.\n\nTHE UNIVERSITY OF CALIFORNIA SPECIFICALLY DISCLAIMS ANY WARRANTIES,\nINCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY\nAND FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS\nON AN \"AS IS\" BASIS, AND THE UNIVERSITY OF CALIFORNIA HAS NO OBLIGATIONS\nTO PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.\n\n-------------------------\n\nCopyright ( 1996, 1997, 1998, 1999, 2000 by various contributors (as\nidentified in HISTORY) (collectively \"Developers\") which may be used\naccording to the following licensing terms:\n\nWorldwide permission to use, copy, modify, and distribute this software\nand its documentation for any purpose, without fee, and without a written\nagreement is hereby granted, on a non-exclusive basis, provided that the\nabove copyright notice, this paragraph and the following paragraphs appear\nin all copies:\n\nAny person who contributes or submits any modification or other change to\nthe PostgreSQL software or documentation grants irrevocable,\nnon-exclusive, worldwide permission, without charge, to use, copy, further\nmodify and distribute the same under the terms of this license.\n\nIN NO EVENT SHALL ANY DEVELOPER BE LIABLE TO ANY PARTY FOR DIRECT,\nINDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING,\nWITHOUT LIMITATION, LOST PROFITS, ARISING OUT OF THE USE OF THIS SOFTWARE\nAND ITS DOCUMENTATION, EVEN IF THE DEVELOPER HAS BEEN ADVISED OF THE\nPOSSIBILITY OF SUCH DAMAGE.\n\nTHE DEVELOPERS SPECIFICALLY DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED\nINCLUDING BUT NOT LIMITED TO IMPLIED WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE, NEED, OR QUALITY, AND ANY IMPLIED\nWARRANTY FROM COURSE OF DEALING OR USAGE OF TRADE. IN ADDITION, THERE IS\nNO IMPLIED WARRANTY AGAINST INTERFERENCE WITH ENJOYMENT OR AGAINST\nINFRINGEMENT. THE SOFTWARE AND DOCUMENTATION PROVIDED HEREUNDER IS ON AN\n\"AS IS\" BASIS. NO DEVELOPER HAS ANY OBLIGATION TO PROVIDE MAINTENANCE,\nSUPPORT, UPDATES, ENHANCEMENTS OR MODIFICATIONS TO OR FOR THE SOFTWARE OR\nDOCUMENTATION.\n\nBY USING THIS SOFTWARE YOU AGREE TO THESE TERMS AND CONDITIONS. IF YOU DO\nNOT AGREE TO THESE TERMS AND CONDITIONS, YOU SHOULD NOT USE THIS SOFTWARE.\n\n", "msg_date": "Tue, 4 Jul 2000 11:42:00 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Revised Copyright: is this more palatable?" }, { "msg_contents": "At 11:42 4/07/00 -0300, The Hermit Hacker wrote:\n>\n>The only part that I believe at least one person had an issue with was:\n>\n>\"Any person who contributes or submits any modification or other change to\n>the PostgreSQL software or documentation grants irrevocable,\n>non-exclusive, worldwide permission, without charge, to use, copy, further\n>modify and distribute the same under the terms of this license.\"\n>\n>Quite frankly, all I'm reading into this paragraph is that once committed,\n>Jan (as a recent example) couldn't come along and pull out all his TOAST\n>changes ... could you imagine the hell that would wreak were he (or anyone\n>else) were to pull crucial changes after others have built upon it?\n\nI am (still) waiting to hear from my IP lawyer, but it is my understanding\nthat if Jan puts TOAST into CVS, then he has given an implied license for\nuse to use it in the open source project. As a result I doubt he could\nactually force it's removal. What he could do is stop a third party from\nusing it in another product. This does not seem bad to me.\n\nUnfortunately, with your revised clause, he no longer has that right.\n\nWhy not just leave the clause out? The more you diverge from BSD, the more\nyou make me want GPL. \n\n\n>The only change in this is the \"Juristiction\" para is removed ... I've\n>read this over several times now, and personal feel that all its doing is\n>extending the existing copyright to cover *ALL* developers, and not just\n>the \"UNIVERITY OF CALIFORNIA\" ones ...\n\nIt's reducing the rights of developers.\n\n\n>I consider it an appendum to the existing copyright ... I don't know, does\n>that make it any less BSD/open?\n\nI think it does; but I'd be open to any reason why I as a developer should\nfeel stronger as a result of your suggested clause.\n\nCertainly if I were a private company who wanted to use PG, I would feel\nmore comfortable with this clause, but that is not how you are marketing it.\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 01:13:05 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Revised Copyright: is this more palatable?" }, { "msg_contents": "The Hermit Hacker wrote:\n>\n> Okay, from seeing the responses so far on the list, I'm not the only one\n> that has issues with the whole \"juristiction of virginia\" issue *or* the\n> \"slam this copyright in ppls faces\" ... I do like the part in BOLD about\n> \"ANY DEVELOPER\" instead of just the \"UNIVERSITY OF CALIFORNIA\" ... but I\n> consider that an appendum/extension of what is already stated ...\n>\n> Is the following more palatable to those of us that aren't US citizens?\n>\n> The only part that I believe at least one person had an issue with was:\n>\n> \"Any person who contributes or submits any modification or other change to\n> the PostgreSQL software or documentation grants irrevocable,\n> non-exclusive, worldwide permission, without charge, to use, copy, further\n> modify and distribute the same under the terms of this license.\"\n>\n> Quite frankly, all I'm reading into this paragraph is that once committed,\n> Jan (as a recent example) couldn't come along and pull out all his TOAST\n> changes ... could you imagine the hell that would wreak were he (or anyone\n> else) were to pull crucial changes after others have built upon it?\n\n The new license should clearly make it impossible to later\n pull out things again. To stay with me as example, what would\n happen if I take out PL/pgSQL, FOREIGN KEY (not all mine I\n know), the fixes to the rewriter and so on. They all where\n contributed under the old license, so I still hold the\n copyright on 'em - don't I. Can a new license change the\n legal state of previous contributions? I don't think so. What\n do we have to do to reversely apply this \"irrevocable\" term\n to all so far done contributions?\n\n And some words to all the people who think GPL is better.\n IMHO it is a kind of Open Source Fashism. Forcing everything\n that uses a little snippet of open code to be open too\n doesn't have anything to do with free software. There are a\n couple of things Open Source can never offer. For example a\n native DB-link interface between a Postgres DB and a\n commercial one might require NDA to get internals. Surely a\n useful thing that must be a closed source product, so what\n would it be good for to make it's development impossible?\n\n If someone needs a feature and is willing to pay alot money\n to get it right now, why shouldn't a company or some\n individual grab it and implement the feature. At some point,\n those will learn that it is a good idea to contribute these\n things to the free source too, because they'll get rid of\n most maintainence efford and gain that future development on\n our side doesn't collide with what they're responsible for.\n It's so obvious to me that I don't need a license that\n enforces it from the very first second.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Tue, 4 Jul 2000 17:51:14 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "On Wed, 5 Jul 2000, Philip Warner wrote:\n\n> At 11:42 4/07/00 -0300, The Hermit Hacker wrote:\n> >\n> >The only part that I believe at least one person had an issue with was:\n> >\n> >\"Any person who contributes or submits any modification or other change to\n> >the PostgreSQL software or documentation grants irrevocable,\n> >non-exclusive, worldwide permission, without charge, to use, copy, further\n> >modify and distribute the same under the terms of this license.\"\n> >\n> >Quite frankly, all I'm reading into this paragraph is that once committed,\n> >Jan (as a recent example) couldn't come along and pull out all his TOAST\n> >changes ... could you imagine the hell that would wreak were he (or anyone\n> >else) were to pull crucial changes after others have built upon it?\n> \n> I am (still) waiting to hear from my IP lawyer, but it is my understanding\n> that if Jan puts TOAST into CVS, then he has given an implied license for\n> use to use it in the open source project. As a result I doubt he could\n> actually force it's removal. What he could do is stop a third party from\n> using it in another product. This does not seem bad to me.\n> \n> Unfortunately, with your revised clause, he no longer has that right.\n> \n> Why not just leave the clause out? The more you diverge from BSD, the more\n> you make me want GPL. \n> \n> \n> >The only change in this is the \"Juristiction\" para is removed ... I've\n> >read this over several times now, and personal feel that all its doing is\n> >extending the existing copyright to cover *ALL* developers, and not just\n> >the \"UNIVERITY OF CALIFORNIA\" ones ...\n> \n> It's reducing the rights of developers.\n> \n> \n> >I consider it an appendum to the existing copyright ... I don't know, does\n> >that make it any less BSD/open?\n> \n> I think it does; but I'd be open to any reason why I as a developer should\n> feel stronger as a result of your suggested clause.\n> \n> Certainly if I were a private company who wanted to use PG, I would feel\n> more comfortable with this clause, but that is not how you are marketing it.\n\nWait ... you had me on the first section, but this second one does confuse\nme ... \"reducing the rights of developers\" applies to the \"Any person who\ncontributes...\" clause, or the BOLD liability clauses?\n\nI'm definitely not sold on the \"Any person who contributes or submits any\nmodification...\" clause, and *if* your IP lawyer comes back that your\nunderstanding is accurate, I'm even less sold on it ... look forward to\nhearing back on that ...\n\nFor me, the only thing that I really like is the three extra BOLD paras\nthat extend the protection from liability to encompass ALL DEVELOPERS\ninstead of just \"UNIVERSITY OF CALIFORNIA\", which I don't believe any of\nus falls under? :)\n\n\n\n", "msg_date": "Tue, 4 Jul 2000 13:04:55 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Revised Copyright: is this more palatable?" }, { "msg_contents": "On Tue, 4 Jul 2000, Jan Wieck wrote:\n\n> The Hermit Hacker wrote:\n> >\n> > Okay, from seeing the responses so far on the list, I'm not the only one\n> > that has issues with the whole \"juristiction of virginia\" issue *or* the\n> > \"slam this copyright in ppls faces\" ... I do like the part in BOLD about\n> > \"ANY DEVELOPER\" instead of just the \"UNIVERSITY OF CALIFORNIA\" ... but I\n> > consider that an appendum/extension of what is already stated ...\n> >\n> > Is the following more palatable to those of us that aren't US citizens?\n> >\n> > The only part that I believe at least one person had an issue with was:\n> >\n> > \"Any person who contributes or submits any modification or other change to\n> > the PostgreSQL software or documentation grants irrevocable,\n> > non-exclusive, worldwide permission, without charge, to use, copy, further\n> > modify and distribute the same under the terms of this license.\"\n> >\n> > Quite frankly, all I'm reading into this paragraph is that once committed,\n> > Jan (as a recent example) couldn't come along and pull out all his TOAST\n> > changes ... could you imagine the hell that would wreak were he (or anyone\n> > else) were to pull crucial changes after others have built upon it?\n> \n> The new license should clearly make it impossible to later\n> pull out things again. To stay with me as example, what would\n> happen if I take out PL/pgSQL, FOREIGN KEY (not all mine I\n> know), the fixes to the rewriter and so on. They all where\n> contributed under the old license, so I still hold the\n> copyright on 'em - don't I. Can a new license change the\n> legal state of previous contributions? I don't think so. What\n> do we have to do to reversely apply this \"irrevocable\" term\n> to all so far done contributions?\n> \n> And some words to all the people who think GPL is better.\n> IMHO it is a kind of Open Source Fashism. Forcing everything\n> that uses a little snippet of open code to be open too\n> doesn't have anything to do with free software. There are a\n> couple of things Open Source can never offer. For example a\n> native DB-link interface between a Postgres DB and a\n> commercial one might require NDA to get internals. Surely a\n> useful thing that must be a closed source product, so what\n> would it be good for to make it's development impossible?\n> \n> If someone needs a feature and is willing to pay alot money\n> to get it right now, why shouldn't a company or some\n> individual grab it and implement the feature. At some point,\n> those will learn that it is a good idea to contribute these\n> things to the free source too, because they'll get rid of\n> most maintainence efford and gain that future development on\n> our side doesn't collide with what they're responsible for.\n> It's so obvious to me that I don't need a license that\n> enforces it from the very first second.\n\nSo you are in the \"make no changes to existing license\" camp? Or just\nagainst that one para above?\n\nMarc G. Fournier ICQ#7615664 IRC Nick: Scrappy\nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n", "msg_date": "Tue, 4 Jul 2000 13:54:15 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "On Tue, Jul 04, 2000 at 05:51:14PM +0200, Jan Wieck wrote:\n\n> The new license should clearly make it impossible to later\n> pull out things again. \n\nI'm confused about this. I'm not a coder, so I beg forgiveness for my\nintrusion, but how would it be possible to revoke the license on code once\ncontributed?\n\nIf I distribute something under terms t(1), and then later distribute the\nsame thing under terms t(2), even if terms t(2) revoke terms t(1), I can't\ngo back and get the original distribution back. \n\nNow, in the case of something easily distributed (like code), if terms t(1)\nallow free distribution, then all one needs is to argue that one is copying\nthat original distribution. Am I missing something?\n\nBecause of the above, it seems to me that once some copyrighted work has\nbeen opened, it can't be closed again. Future developments by the original\ncopyright owner can, of course, include the original copyrightedwork under\ndifferent terms. But even the original copyright owner can't go back and\nchange the license forsomething in the past, no?\n\n-- \nAndrew Sullivan Computer Services\n<[email protected]> Burlington Public Library\n+1 905 639 3611 x158 2331 New Street\n Burlington, Ontario, Canada L7R 1J4\n", "msg_date": "Tue, 4 Jul 2000 14:13:59 -0400", "msg_from": "Andrew Sullivan <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "On Tue, 4 Jul 2000, Andrew Sullivan wrote:\n\n> On Tue, Jul 04, 2000 at 05:51:14PM +0200, Jan Wieck wrote:\n> \n> > The new license should clearly make it impossible to later\n> > pull out things again. \n> \n> I'm confused about this. I'm not a coder, so I beg forgiveness for my\n> intrusion, but how would it be possible to revoke the license on code once\n> contributed?\n> \n> If I distribute something under terms t(1), and then later distribute the\n> same thing under terms t(2), even if terms t(2) revoke terms t(1), I can't\n> go back and get the original distribution back. \n> \n> Now, in the case of something easily distributed (like code), if terms t(1)\n> allow free distribution, then all one needs is to argue that one is copying\n> that original distribution. Am I missing something?\n> \n> Because of the above, it seems to me that once some copyrighted work has\n> been opened, it can't be closed again. Future developments by the original\n> copyright owner can, of course, include the original copyrightedwork under\n> different terms. But even the original copyright owner can't go back and\n> change the license forsomething in the past, no?\n\nTo the best of *my* knowledge, a copyright cannot be retro-actively\nimposed on software ... but, I'm not a copyright lawyer, so may be wrong\non this ...\n\nI *believe* what Jan was getting at was that the copyright should be made\nsuch that, as our example has gone so far, if his TOAST contribution falls\nunder said copyright, he can't, at some later date, decide to pull *his*\ncode out of the tree ... but, it only works \"from that day forward\", not\nretro-actively on any previous code he's submitted ...\n\n\n", "msg_date": "Tue, 4 Jul 2000 15:24:21 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "The Hermit Hacker wrote:\n>\n> So you are in the \"make no changes to existing license\" camp? Or just\n> against that one para above?\n>\n\n My home is my castle - since I where in the army I don't like\n camps any more :-)\n\n All I want is to be safe against beeing sued for something I\n did for free, available for nothing on an AS IS base.\n\n The aspect of \"this license permits to make it proprietary\"\n is IMHO weakened by itself. More and more companies see the\n benefits of going open source (getting high experienced\n programming and maintainence for free) as the Mozilla,\n Interbase and other examples demonstrate.\n\n I started with individual software development on the\n customer side. The next step was standard software, to\n reduce development costs. Now we are at the beginning of the\n next wave, open source. Again reducing development costs. The\n closed shop companies will be the loosers in the long run, I\n believe. So my focus isn't really on keeping open what once\n became open - that'll take care for itself because someone\n will be there taking the last open version and continue from\n that.\n\n My focus is on beeing protected. And I'm not sure if beeing\n bound to any US States law is a good choice for that.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Wed, 5 Jul 2000 01:09:15 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "The Hermit Hacker writes:\n\n> Is the following more palatable to those of us that aren't US citizens? \n\nIs it really necessary that the developers' DISCLAIMER is different in\nwording than the UCB DISCLAIMER? Is there any deeper thought behind it? If\nnot, I humbly suggest using the same text.\n\n> The only part that I believe at least one person had an issue with was:\n> \n> \"Any person who contributes or submits any modification or other change to\n> the PostgreSQL software or documentation grants irrevocable,\n> non-exclusive, worldwide permission, without charge, to use, copy, further\n> modify and distribute the same under the terms of this license.\"\n\nThere is certainly a point behind this. But this point essentially applies\nto all open source projects, even GPL'ed ones. Is there an implicit grant\nof license when you submit patches? If we feel that this issue needs to be\naddressed then we should involve the open source community at large, not\nmove ahead by ourselves. Ask the Open Source Initiative, Software in the\nPublic Interest, the FSF, other large scale projects (FreeBSD), other\ncommercial entities, such as RedHat. You just can't sneak in this sort of\nnew language without anyone noticing.\n\nThere are also particular issues I have with the wording. What's a\n\"contribution\"? One could interpret that any modification to PostgreSQL\nthat you make available to the public at large is a contribution. That\nwould make re-publishing PostgreSQL under a different license impossible.\n\nIt would also be quite tricky to enforce this change. Consider this: All\nof the current code is under the current license, you can't change that.\nTherefore, anyone who contributes changes to the *current* code is not\nbound by this *new* license; only truly *new* code is bound by the new\nlicense. But if it's truly new code then it does not constitute a\n\"modification or other change\". So this license could never take effect\nunless the all the authors of the exiting code explicitly agree that all\nof their old code is now under the new license. That's not going to\nhappen.\n\nYes, I'm language lawyering. But the fact that such interpretations are\npossible makes me uneasy.\n\n> I consider it an appendum to the existing copyright ... I don't know, does\n> that make it any less BSD/open?\n\ns/copyright/license/\n\nThe _copyright_ is held by the entity who wrote the code, the _license_\nare the conditions under which he makes them available to the rest of the\nworld. This also makes the current wording \"Portions Copyright ...\nPostgreSQL, Inc.\" kind of questionable.\n\nAnd yes, IMHO any change to the BSD license makes it less BSD.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Wed, 5 Jul 2000 02:11:53 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "At 01:09 5/07/00 +0200, Jan Wieck wrote:\n>\n> So my focus isn't really on keeping open what once\n> became open - that'll take care for itself because someone\n> will be there taking the last open version and continue from\n> that.\n\nThis is called 'faith', and is commendable, but I'm and old testament sort\nof person - a few thunderbolts never go amiss!\n\nI truly hope you are right, and that the result of Software Manufacturers\nVs. The Hackers is not the destruction of open source software through\nextensive legal battles and intense self-interest. \n\nAm I correct in saying that you agree that the GPL is where we should be,\nbut you want people to go there of their own free will?\n\n\n> My focus is on beeing protected. And I'm not sure if beeing\n> bound to any US States law is a good choice for that.\n\nI think it's a bad idea.\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 14:12:55 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more\n palatable?" }, { "msg_contents": "Philip Warner wrote:\n> \n> At 01:09 5/07/00 +0200, Jan Wieck wrote:\n> >\n> > So my focus isn't really on keeping open what once\n> > became open - that'll take care for itself because someone\n> > will be there taking the last open version and continue from\n> > that.\n> \n> This is called 'faith', and is commendable, but I'm and old testament sort\n> of person - a few thunderbolts never go amiss!\n> \n> I truly hope you are right, and that the result of Software Manufacturers\n> Vs. The Hackers is not the destruction of open source software through\n> extensive legal battles and intense self-interest.\n> \n> Am I correct in saying that you agree that the GPL is where we should be,\n> but you want people to go there of their own free will?\n\nWhy do you continue to insist that GPL is superior to BSD? GPL is\nBSD *with restrictions*. If someone comes along and sweeps up the\nmajor developers:\n\nA) Good for the major developers - they deserve to have large\nsums of cash thrown their way, particularly for many of them who\nhave been working on this *for years*\n\nB) The moment it happens, the project forks and another \"Marc\"\nout-there offers to host development on his machine and the\nprocess begins again. PostgreSQL exists despite Illustra's\nexistence.\n\nThis is not something new. SunOS, AIX, HPUX, etc. all have (at\none time or another) considerable BSD roots. And yet FreeBSD\nstill exists... All GPL does is 'poison' the pot by prohibiting\ncommercial spawns which may leverage the code. If someone makes\nsome money selling CommercialGres by integrating replication,\ndistributive, and parallel query, good for them.\n\nMike Mascari\n", "msg_date": "Wed, 05 Jul 2000 00:24:31 -0400", "msg_from": "Mike Mascari <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "At 13:04 4/07/00 -0300, The Hermit Hacker wrote:\n>> \n>> Certainly if I were a private company who wanted to use PG, I would feel\n>> more comfortable with this clause, but that is not how you are marketing\nit.\n>\n>Wait ... you had me on the first section, but this second one does confuse\n>me ... \"reducing the rights of developers\" applies to the \"Any person who\n>contributes...\" clause, or the BOLD liability clauses?\n\nIt was meant to apply to \"Any person who...\", but based on legal advice, I\nwas a little wrong. Assuming I could be reasonably expected to be aware of\nthe BSD nature of the project, and the terms of the BSD license, then by\nmaking a submission, my submission comes under the BSD. \n\nI do still own copyright, but I can't prevent people using it in a\nPG-derived project. But I *can* prevent them using it in software to run a\nmeat-grinder, assuming that software is not recognizably a PG derived\ndatabase (as perceived by a reasonable person). This could be relevant to\nJan who wrote compression code, although I suspect he's philosophically\nopposed to preventing his actual source code being used in, eg, \"The New\nFangled Microsoft Disk Compressor, using advanced Compression Technology\n(tm)\".\n\nThe new clause removes this right.\n\n\n>I'm definitely not sold on the \"Any person who contributes or submits any\n>modification...\" clause, and *if* your IP lawyer comes back that your\n>understanding is accurate, I'm even less sold on it ... look forward to\n>hearing back on that ...\n\nI still don't like it for the above reason.\n\n\n>For me, the only thing that I really like is the three extra BOLD paras\n>that extend the protection from liability to encompass ALL DEVELOPERS\n>instead of just \"UNIVERSITY OF CALIFORNIA\", which I don't believe any of\n>us falls under? :)\n\nYes, but legal advice so far suggests it is useless in Australia (&\nprobably the UK), where it is not lawful to remove warranties. All you can\ndo here is reduce liability. More on this in a future email.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 14:40:43 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Revised Copyright: is this more palatable?" }, { "msg_contents": "At 00:24 5/07/00 -0400, Mike Mascari wrote:\n>> \n>> Am I correct in saying that you agree that the GPL is where we should be,\n>> but you want people to go there of their own free will?\n>\n>Why do you continue to insist that GPL is superior to BSD? GPL is BSD \n>*with restrictions*\n\nI don't. The above was a question to Jan.\n\nI have stated in the past that I would prefer PG to be GPL, but that is\nbased on my perception of PG as a 'strategic resource' for my company. The\nGPL Vs. BSD discussion is a religious war that will only be resolved in\ntime. I do, honestly, hope Jan is right about the convergence of open\nsource and industry.\n\nThe context in which I said I would prefer PG to be GPL was when someone\nsuggested leaving the BSD licence for a more restrictive (to developers\nrights over their submissions) version.\n\n\n>If someone comes along and sweeps up the\n>major developers:\n>\n>A) Good for the major developers - they deserve to have large\n>sums of cash thrown their way, particularly for many of them who\n>have been working on this *for years*\n\nI totally agree. This can happen under GPL. If I were a company wanting to\ndevelop PG, the source would be less of an issue than access to the core\ndevelopers who are the real resource. As Jan has said elsewhere, keeping\nsource secret is a waste of effort.\n\n\n>B) The moment it happens, the project forks and another \"Marc\"\n>out-there offers to host development on his machine and the\n>process begins again. PostgreSQL exists despite Illustra's\n>existence.\n\nNo problem here but wasted effort.\n\n\n>This is not something new. SunOS, AIX, HPUX, etc. all have (at\n>one time or another) considerable BSD roots. And yet FreeBSD\n>still exists... All GPL does is 'poison' the pot by prohibiting\n>commercial spawns which may leverage the code. If someone makes\n>some money selling CommercialGres by integrating replication,\n>distributive, and parallel query, good for them.\n\nThis is the place where the religious war starts, so I'll confine my\ncomments to the issue at hand:\n\nIn summary of my position:\n\n1. I am happy to continue with vanilla BSD + extra warranty & liability\ndisclaimers.\n\n2. If the license goes anywhere else, I *believe* it probably should go GPL.\n\n3. If people really want to write yeat another license (as opposed to\nwarranty), then they should do it properly - look at the objectives, look\nat all existing public licenses, find the once closest to what the\ndeveloper & user base wants, modify it according to the specific needs.\nPerhaps a modified LGPL might be better. Someone (Jan?) noted that adding a\ndriver is a problem under GPL - perhaps that should be addressed, rather\nthan just taking away more rights of developers as is proposed by Great\nBridge.\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 15:00:08 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this\n morepalatable?" }, { "msg_contents": "At 17:51 4/07/00 +0200, Jan Wieck wrote:\n>The Hermit Hacker wrote:\n>>\n> The new license should clearly make it impossible to later\n> pull out things again. \n\nMy legal advice is that, assuming they knew it was a BSD project, they\ncan't take it out of PostgreSQL. But you could, for example, stop Microsoft\nusing your compression code in one of their products. The new license\nremoves this right from you.\n\n\n> I still hold the\n> copyright on 'em - don't I. \n\nYou will have the copyright, even under the new license. All you are doing\nis waiving your rights to restrict it's use in any way whatsoever for any\npurpose.\n\n\n> Can a new license change the\n> legal state of previous contributions?\n\nNo.\n\n\n> I don't think so. What\n> do we have to do to reversely apply this \"irrevocable\" term\n> to all so far done contributions?\n\nYes. Sticking with BSD looks good to me.\n\n\n> And some words to all the people who think GPL is better.\n> IMHO it is a kind of Open Source Fashism. Forcing everything\n> that uses a little snippet of open code to be open too\n> doesn't have anything to do with free software. There are a\n> couple of things Open Source can never offer. For example a\n> native DB-link interface between a Postgres DB and a\n> commercial one might require NDA to get internals. Surely a\n> useful thing that must be a closed source product, so what\n> would it be good for to make it's development impossible?\n\nI agree this is a problem with GPL; perhaps not with LGPL if the new code\ncould be written to require minimum changes to existing PG code, and if the\nPG changes were not related to the NDA. I don't really want to start a GPL\nvs. BSD argument here, I'm just asking if you had thought of that possibility.\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 15:10:49 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more\n palatable?" }, { "msg_contents": "Philip Warner wrote:\n\n> My legal advice is that, assuming they knew it was a BSD project, they\n> can't take it out of PostgreSQL. But you could, for example, stop Microsoft\n> using your compression code in one of their products. The new license\n> removes this right from you.\n\nWhy wouldn't MS be able to take the code and use it while abiding by its\nterms and conditions?\n", "msg_date": "Wed, 05 Jul 2000 15:15:12 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "At 15:15 5/07/00 +1000, Chris Bitmead wrote:\n>\n>Why wouldn't MS be able to take the code and use it while abiding by its\n>terms and conditions?\n>\n\nI am told that the most likely interpretation of this is that it is for use\nin PostgreSQL or one of its descendants. The new clause changes that to\n'any use whatsoever'.\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Wed, 05 Jul 2000 15:34:41 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this\n morepalatable?" }, { "msg_contents": "Philip Warner wrote:\n>\n> Am I correct in saying that you agree that the GPL is where we should be,\n> but you want people to go there of their own free will?\n\n Right. Someone who doesn't want to make his code \"FREE\" in\n the entire meaning of this word but want to make it open for\n any non-commercial use should choose it. IMHO the GPL\n includes \"this is the one and only truth and must propagate\n up into everything started on something that once went under\n this license\". Who am I to restrict my code in that way? Can\n I see the future?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Wed, 5 Jul 2000 11:10:27 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more palatable??" }, { "msg_contents": "Mike Mascari wrote:\n> This is not something new. SunOS, AIX, HPUX, etc. all have (at\n> one time or another) considerable BSD roots. And yet FreeBSD\n> still exists... All GPL does is 'poison' the pot by prohibiting\n> commercial spawns which may leverage the code. If someone makes\n> some money selling CommercialGres by integrating replication,\n> distributive, and parallel query, good for them.\n\n Let them! It's good for the customer too, because he mustn't\n wait until we lazy dogs implement all that.\n\n If they are smart, they will contribute it to the open source\n tree sometimes after having their ROI. Otherwise they run the\n risk of getting stuck someday when their changes don't apply\n any more to our tree but they are still responsible for it's\n functionality.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Wed, 5 Jul 2000 11:31:26 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "Philip Warner wrote:\n> I do still own copyright, but I can't prevent people using it in a\n> PG-derived project. But I *can* prevent them using it in software to run a\n> meat-grinder, assuming that software is not recognizably a PG derived\n> database (as perceived by a reasonable person). This could be relevant to\n> Jan who wrote compression code, although I suspect he's philosophically\n> opposed to preventing his actual source code being used in, eg, \"The New\n> Fangled Microsoft Disk Compressor, using advanced Compression Technology\n> (tm)\".\n\n Especially the compression code was only coding. The\n underlying algorithm (slightly modified) was taken from a 10+\n years old usenet article and is the idea of Adisak\n Pochanayon.\n\n If anyone on earth can make money out of it, I don't have a\n problem with that. This includes M$ and any other company\n with bad reputation. But if there's a bug in my code causing\n severe damage, I don't want to get sued for it. That's all.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Wed, 5 Jul 2000 11:42:39 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Revised Copyright: is this more palatable?" }, { "msg_contents": "Thus spake Jan Wieck\n> Right. Someone who doesn't want to make his code \"FREE\" in\n> the entire meaning of this word but want to make it open for\n> any non-commercial use should choose it. IMHO the GPL\n\nWhile I am a proponent of keeping the BSD style license, there is nothing\nin the GPL about using code for commercial use one way or the other.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n", "msg_date": "Wed, 5 Jul 2000 08:02:09 -0400 (EDT)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more palatable??" }, { "msg_contents": "On Wed, 5 Jul 2000, Philip Warner wrote:\n\n> At 00:24 5/07/00 -0400, Mike Mascari wrote:\n> >> \n> >> Am I correct in saying that you agree that the GPL is where we should be,\n> >> but you want people to go there of their own free will?\n> >\n> >Why do you continue to insist that GPL is superior to BSD? GPL is BSD \n> >*with restrictions*\n> \n> I don't. The above was a question to Jan.\n> \n> I have stated in the past that I would prefer PG to be GPL, but that is\n> based on my perception of PG as a 'strategic resource' for my company. The\n> GPL Vs. BSD discussion is a religious war that will only be resolved in\n> time. I do, honestly, hope Jan is right about the convergence of open\n> source and industry.\n\nPhilip ... I abhor GPL myself, which is why PostgreSQL will never fall\nunder it ... I think it is just this side of 'MicroSloth evil' in that it\ncreates way more restrictions on code that are necessary. Its been around\nso long that ppl have been effectively brainwashed into thinking that\n\"this is the only open source license\" ...\n\nYou cannot close source open source ... unless ppl don't care. If\nsomeone were to come along and try, someone else comes along and forks the\ncode off at the point *just before* the license changed and continues\nalong their own thread. Quite frankly, that person forking it off would\nbe me, since PostgreSQL was never intended to be closed source ...\n\n... it doesn't matter if the code is under BSD or GPL, that fork can (and\nwill) happen ... with GPL, its near impossible to do ... with BSD, its\neasier, but it buys little for the commercial enterprise doing so ...\n\nI was going to say that what BSD buys someone over GPL is the ability to\ncreate modules taht are binary only, but even GPL allows for that\n... *shrug*\n\n\n> >A) Good for the major developers - they deserve to have large\n> >sums of cash thrown their way, particularly for many of them who\n> >have been working on this *for years*\n> \n> I totally agree. This can happen under GPL. If I were a company wanting to\n> develop PG, the source would be less of an issue than access to the core\n> developers who are the real resource. As Jan has said elsewhere, keeping\n> source secret is a waste of effort.\n\nOkay, so BSD vs GPL matters not here ...\n\n> >B) The moment it happens, the project forks and another \"Marc\"\n> >out-there offers to host development on his machine and the\n> >process begins again. PostgreSQL exists despite Illustra's\n> >existence.\n> \n> No problem here but wasted effort.\n\nAnd BSD vs GPL matters not here ...\n\n> In summary of my position:\n> \n> 1. I am happy to continue with vanilla BSD + extra warranty & liability\n> disclaimers.\n\nThis is my feeling too ... I won't agree to changing the license over to a\n\"under juristiction of ...\", nor will I agreee with the \"slam this in\nfront of ppls faces and force them to read it ...\". \n\nPersonally, from all the 'legal' issues that FreeBSD has gone through over\nthe years, especially recently with the BSDi/FreeBSD merger and the whole\ncryptology merger, I would think they would have been the first to\nadopt/change their BSD license to something else, and I've never even seen\ndiscussions on it ...\n\nPutting the license up as a README on the ftp site, and maybe including it\nas part of the download page ... no probs, not obnoxious ... hell, how\nmany ppl even read the license on sites that require a 'I agree'? \n\n", "msg_date": "Wed, 5 Jul 2000 09:11:43 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "On Wed, 5 Jul 2000, The Hermit Hacker wrote:\n\n> Personally, from all the 'legal' issues that FreeBSD has gone through over\n> the years, especially recently with the BSDi/FreeBSD merger and the whole\n> cryptology merger, I would think they would have been the first to\n> adopt/change their BSD license to something else, and I've never even seen\n> discussions on it ...\n\nThe only thing that comes close are the periodic discussions (ignored by\nFBSD developers) about the removal of the 'offensive' mascot. ;-)\n\nTop Ten Gratuitous Discussions:\n\n1. FreeBSD v. Linux\n2. GPL v. BSD licence\n3. Removal of 'Chuckie' (Berkeley Daemon) from BSD \n in favour of something less 'Satanic' or 'Demonic'\n4. (Official) renaming of 'Chuckie' so as not offend McKusick\n over the popular (mis)perception that the Berkeley Daemon \n is (nick)named 'Chuckie'\n5. MySQL v. Postgres\n6. RedHat v. Debian v. SuSe\n7. Why was Slackware not involved in the above discussion\n8. Perl v. Python\n9. Coke v. Pepsi\n\nlastly:\n\n10. Altering the PG licence...I agree with *my perception* of Scrappy's\n position: If FBSD get nervous and changes their licence, Pg should\n follow suit. Otherwise it seems counterproductive.\n\nAnyway, thanks for the great code. I promise not to abuse it.\nI promise not to sue anyone if I am too damn stupid to use it properly.\n\n--------------------------------------------------------------------\n SVCMC - Center for Behavioral Health \n--------------------------------------------------------------------\nThomas Good tomg@ { admin | q8 } .nrnet.org\nIS Coordinator / DBA Phone: 718-354-5528 \n Fax: 718-354-5056 \n--------------------------------------------------------------------\nPowered by: PostgreSQL s l a c k w a r e FreeBSD:\n RDBMS |---------- linux The Power To Serve\n--------------------------------------------------------------------\n\n", "msg_date": "Wed, 5 Jul 2000 09:00:55 -0400 (EDT)", "msg_from": "Thomas Good <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "Mike Mascari <[email protected]> writes:\n\n> This is not something new. SunOS, AIX, HPUX, etc. all have (at\n> one time or another) considerable BSD roots. And yet FreeBSD\n> still exists... All GPL does is 'poison' the pot by prohibiting\n> commercial spawns which may leverage the code.\n\nGPL doesn't prohibit commercial spawns - it just requires you to send\nthe source along.\n\n-- \nTrond Eivind Glomsr�d\nRed Hat, Inc.\n", "msg_date": "05 Jul 2000 10:30:42 -0400", "msg_from": "[email protected] (Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?=)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?= wrote:\n> Mike Mascari <[email protected]> writes:\n>\n> > This is not something new. SunOS, AIX, HPUX, etc. all have (at\n> > one time or another) considerable BSD roots. And yet FreeBSD\n> > still exists... All GPL does is 'poison' the pot by prohibiting\n> > commercial spawns which may leverage the code.\n>\n> GPL doesn't prohibit commercial spawns - it just requires you to send\n> the source along.\n\n So if someone offers $$$ for implementation of Postgres\n feature XYZ I don't have to make that code open source? Only\n need to ship the code to the one paying (under NDA so he\n cannot publish it) and grab the money?\n\n That's currently possible with our license, and if someone\n would pay $30,000 for something I need a month of unpayed\n vacation to implement, why not? Maybe I can sell it to\n others, that have the same urgent need for this feature, too.\n Making another $xK out of it.\n\n No question, the result will finally get contributed so\n everyone benefits from it. The reason why someone is even\n willing to pay that amount is just to get me out of my\n dayjob, focussing on his problem NOW, so it's done in a\n month.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Wed, 5 Jul 2000 19:22:45 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "[email protected] (Jan Wieck) writes:\n\n> Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?= wrote:\n> > Mike Mascari <[email protected]> writes:\n> >\n> > > This is not something new. SunOS, AIX, HPUX, etc. all have (at\n> > > one time or another) considerable BSD roots. And yet FreeBSD\n> > > still exists... All GPL does is 'poison' the pot by prohibiting\n> > > commercial spawns which may leverage the code.\n> >\n> > GPL doesn't prohibit commercial spawns - it just requires you to send\n> > the source along.\n> \n> So if someone offers $$$ for implementation of Postgres\n> feature XYZ I don't have to make that code open source?\n\nYou don't have to tell the world they can have it for free - you can\nsell it, and develop it by demand.\n\n> Only need to ship the code to the one paying\n\nYes.\n\n> (under NDA so he cannot publish it) and grab the money?\n\nAn NDA can't be done, but you certainly can grab the money. If he\nwanted you to add a feature, he could pay you for it.\n\n> That's currently possible with our license,\n\nWith GPL as well - you just can't deny the customer to distribute the\nsource if he wants to.\n> \n> No question, the result will finally get contributed so\n> everyone benefits from it. The reason why someone is even\n> willing to pay that amount is just to get me out of my\n> dayjob, focussing on his problem NOW, so it's done in a\n> month.\n\nThat would certainly be possible. \n-- \nTrond Eivind Glomsr�d\nRed Hat, Inc.\n", "msg_date": "05 Jul 2000 14:33:08 -0400", "msg_from": "[email protected] (Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?=)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?= wrote:\n> [email protected] (Jan Wieck) writes:\n>\n> > Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?= wrote:\n> > > Mike Mascari <[email protected]> writes:\n> > >\n> > > > This is not something new. SunOS, AIX, HPUX, etc. all have (at\n> > > > one time or another) considerable BSD roots. And yet FreeBSD\n> > > > still exists... All GPL does is 'poison' the pot by prohibiting\n> > > > commercial spawns which may leverage the code.\n> > >\n> > > GPL doesn't prohibit commercial spawns - it just requires you to send\n> > > the source along.\n> >\n> > So if someone offers $$$ for implementation of Postgres\n> > feature XYZ I don't have to make that code open source?\n>\n> You don't have to tell the world they can have it for free - you can\n> sell it, and develop it by demand.\n>\n> > Only need to ship the code to the one paying\n>\n> Yes.\n\n Now I don't want to ship the source code. My customer would\n be happy with a patched 8.2.3 binary as long as I'm\n responsible to patch future versions until I release the\n sources. Is that OK?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Wed, 5 Jul 2000 23:04:45 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "Mike Mascari wrote:\n> \n> Why do you continue to insist that GPL is superior to BSD? GPL is\n> BSD *with restrictions*. If someone comes along and sweeps up the\n> major developers:\n> \n> A) Good for the major developers - they deserve to have large\n> sums of cash thrown their way, particularly for many of them who\n> have been working on this *for years*\n> \n> B) The moment it happens, the project forks and another \"Marc\"\n> out-there offers to host development on his machine and the\n> process begins again. PostgreSQL exists despite Illustra's\n> existence.\n> \n> This is not something new. SunOS, AIX, HPUX, etc. all have (at\n> one time or another) considerable BSD roots. And yet FreeBSD\n> still exists... All GPL does is 'poison' the pot by prohibiting\n> commercial spawns which may leverage the code. If someone makes\n> some money selling CommercialGres by integrating replication,\n> distributive, and parallel query, good for them.\n\nIs perhaps GPL more restrictive for *developers*? And BSD more\nrestrictive for *consumers*?\n\nAs a consumer I prefer the GPL. But Mike's point is well taken. I\nagree that the GPL is rather idealistic. It makes it very difficult,\nalmost impossible, for someone to make money doing software development.\n\nIs there a middle ground? Somewhere where perhaps I can be assured that\n*someday* in the not-so-distant future I, as a consumer, will have\naccess to source code? Is there any such thing as a license with\nbuilt-in time limits? Reasonably short time limits, as opposed to those\nprovided by the U.S. patent office?\n\nOr is there a way to write an open-source license that allows developers\nto make money? I know, I know, there are too many licenses already. \nBut if talented hard working people can't make a living, there's a\nproblem. This will probably sound very stupid, but would it be possible\nto write a license that said something to the effect of \"if you are a\nbig corporate commercial interest worth more than $X, you must donate $Y\nto postgresql.org.\"?\n\nI'm not trying to rankle the developers who have benefited me so much by\npromoting the GPL. I'm just trying to protect myself as a consumer from\nbeing left in the cold when the product I've spent so much time learning\nand implementing suddenly goes proprietary.\n\nSorry to be cynical, but as a consumer, I can't help seeing BSD licenses\nas good old bait and switch. And this discussion doesn't reassure me\notherwise.\n\nSure, the code can fork. SunOS, AIX, HPUX are good examples. Examples\nof the kind of code forking and corporatism I thought, I hoped, the\nworld was moving away from.\n\n________________________\nRon Peterson\[email protected]\n", "msg_date": "Wed, 05 Jul 2000 17:32:30 -0400", "msg_from": "Ron Peterson <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "Ron, probably the best example to reassure you here is Illustra/Informix,\nwhich is based on the old Berkeley Postgres code. A group of people at\nBerkeley \"forked\" the Postgres code into the closed Illustra system, but it\nsurvived as Postgres95, then later PostgreSQL when Marc and Bruce got\nstarted.\n\nAs a number of people have said, if someone (like Great Bridge or anyone\nelse) ever took the then-current PostgreSQL code proprietary, it would still\nremain as an open source project - and believe me, there are plenty of\npeople who would rather work on it as an open source project than a\nproprietary death-spiral.\n\nWe think the proprietary software development model for large scale projects\n(operating systems, databases, wide-ranging applications) is stupid and\ndead. We don't think open source is going away - in fact, we think it's the\nway most software is going to be developed in the future. There will\ncertainly be companies that try and fork off open source projects and make a\nquick buck; they will fail.\n\nAs I understand your concern, you don't want to make a learning investment\nin something you think is open source, only to have it go closed? I think I\ncan safely say that PostgreSQL as an open source project will never go away\n- the momentum is too strong, the product is too good, the developers are\ntoo committed, for that to happen.\n\nBest,\nNed\n\n\n\nRon Peterson wrote:\n\n> I'm not trying to rankle the developers who have benefited me so much by\n> promoting the GPL. I'm just trying to protect myself as a consumer from\n> being left in the cold when the product I've spent so much time learning\n> and implementing suddenly goes proprietary.\n>\n> Sorry to be cynical, but as a consumer, I can't help seeing BSD licenses\n> as good old bait and switch. And this discussion doesn't reassure me\n> otherwise.\n>\n> Sure, the code can fork. SunOS, AIX, HPUX are good examples. Examples\n> of the kind of code forking and corporatism I thought, I hoped, the\n> world was moving away from.\n>\n> ________________________\n> Ron Peterson\n> [email protected]\n\n", "msg_date": "Wed, 05 Jul 2000 18:06:08 -0400", "msg_from": "Ned Lilly <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "[email protected] (Jan Wieck) writes:\n\n> Trond Eivind Glomsr�d wrote:\n> > [email protected] (Jan Wieck) writes:\n> >\n> > > Trond Eivind Glomsr�d wrote:\n> > > > Mike Mascari <[email protected]> writes:\n> > > >\n> > > > > This is not something new. SunOS, AIX, HPUX, etc. all have (at\n> > > > > one time or another) considerable BSD roots. And yet FreeBSD\n> > > > > still exists... All GPL does is 'poison' the pot by prohibiting\n> > > > > commercial spawns which may leverage the code.\n> > > >\n> > > > GPL doesn't prohibit commercial spawns - it just requires you to send\n> > > > the source along.\n> > >\n> > > So if someone offers $$$ for implementation of Postgres\n> > > feature XYZ I don't have to make that code open source?\n> >\n> > You don't have to tell the world they can have it for free - you can\n> > sell it, and develop it by demand.\n> >\n> > > Only need to ship the code to the one paying\n> >\n> > Yes.\n> \n> Now I don't want to ship the source code. My customer would\n> be happy with a patched 8.2.3 binary as long as I'm\n> responsible to patch future versions until I release the\n> sources. Is that OK?\n\nYou don't have to give the customer the source, as long as you\ngurantee that he gets it (for cost of distribution) if he wants it. \n\n\n-- \nTrond Eivind Glomsr�d\nRed Hat, Inc.\n", "msg_date": "05 Jul 2000 18:49:37 -0400", "msg_from": "[email protected] (Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?=)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?= wrote:\n> [email protected] (Jan Wieck) writes:\n>\n> > Trond Eivind Glomsr�d wrote:\n> > > [email protected] (Jan Wieck) writes:\n> > >\n> > > > Trond Eivind Glomsr�d wrote:\n> > > > > Mike Mascari <[email protected]> writes:\n> > > > >\n> > > > > > This is not something new. SunOS, AIX, HPUX, etc. all have (at\n> > > > > > one time or another) considerable BSD roots. And yet FreeBSD\n> > > > > > still exists... All GPL does is 'poison' the pot by prohibiting\n> > > > > > commercial spawns which may leverage the code.\n> > > > >\n> > > > > GPL doesn't prohibit commercial spawns - it just requires you to send\n> > > > > the source along.\n> > > >\n> > > > So if someone offers $$$ for implementation of Postgres\n> > > > feature XYZ I don't have to make that code open source?\n> > >\n> > > You don't have to tell the world they can have it for free - you can\n> > > sell it, and develop it by demand.\n> > >\n> > > > Only need to ship the code to the one paying\n> > >\n> > > Yes.\n> >\n> > Now I don't want to ship the source code. My customer would\n> > be happy with a patched 8.2.3 binary as long as I'm\n> > responsible to patch future versions until I release the\n> > sources. Is that OK?\n>\n> You don't have to give the customer the source, as long as you\n> gurantee that he gets it (for cost of distribution) if he wants it.\n\n Wordy, but how can I prevent him to ask for?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Thu, 6 Jul 2000 00:50:38 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "[email protected] (Jan Wieck) writes:\n\n> Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?= wrote:\n> > [email protected] (Jan Wieck) writes:\n> >\n> > > Trond Eivind Glomsr�d wrote:\n> > > > [email protected] (Jan Wieck) writes:\n> > > >\n> > > > > Trond Eivind Glomsr�d wrote:\n> > > > > > Mike Mascari <[email protected]> writes:\n> > > > > >\n> > > > > > > This is not something new. SunOS, AIX, HPUX, etc. all have (at\n> > > > > > > one time or another) considerable BSD roots. And yet FreeBSD\n> > > > > > > still exists... All GPL does is 'poison' the pot by prohibiting\n> > > > > > > commercial spawns which may leverage the code.\n> > > > > >\n> > > > > > GPL doesn't prohibit commercial spawns - it just requires you to send\n> > > > > > the source along.\n> > > > >\n> > > > > So if someone offers $$$ for implementation of Postgres\n> > > > > feature XYZ I don't have to make that code open source?\n> > > >\n> > > > You don't have to tell the world they can have it for free - you can\n> > > > sell it, and develop it by demand.\n> > > >\n> > > > > Only need to ship the code to the one paying\n> > > >\n> > > > Yes.\n> > >\n> > > Now I don't want to ship the source code. My customer would\n> > > be happy with a patched 8.2.3 binary as long as I'm\n> > > responsible to patch future versions until I release the\n> > > sources. Is that OK?\n> >\n> > You don't have to give the customer the source, as long as you\n> > gurantee that he gets it (for cost of distribution) if he wants it.\n> \n> Wordy, but how can I prevent him to ask for?\n\nBy doing everything he wants (and perfect) so he doesn't have a need\nfor it? \n\nBasically, GPL is intended to protect the end user and guaranteeing\nhim the source if he wants it - and that he can do what he wants to\nwith it, as long as he doesn't prevent others from doing so.\n\n\n-- \nTrond Eivind Glomsr�d\nRed Hat, Inc.\n", "msg_date": "05 Jul 2000 19:28:04 -0400", "msg_from": "[email protected] (Trond Eivind=?iso-8859-1?q?_Glomsr=F8d?=)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this morepalatable?" }, { "msg_contents": "On 5 Jul 2000, Trond Eivind[iso-8859-1] Glomsr�d wrote:\n> Basically, GPL is intended to protect the end user and guaranteeing\n> him the source if he wants it - and that he can do what he wants to\n> with it, as long as he doesn't prevent others from doing so.\n\nWhich has the potential to really screw over developers who want to devote\nthemselves to for pay, (in this case) PostgreSQL hacking.\n\nThe threat of major pieces of code diverging is enough to assure that most\n\"closed source\" enhancements will eventually be donated once the R&D\neffort has been paid off.\n\nWe see these sort of contributions all the time in the FreeBSD project.\n\nI don't think that the GPL or a GPL like license has anything to offer the\nPostgreSQL project. Extending the liability clause seems like a smart\nmove, as FreeBSD has done (most new stuff is commited under a '2 clause'\nBSD derived license.)\n\n-- \n| Matthew N. Dodd | '78 Datsun 280Z | '75 Volvo 164E | FreeBSD/NetBSD |\n| [email protected] | 2 x '84 Volvo 245DL | ix86,sparc,pmax |\n| http://www.jurai.net/~winter | This Space For Rent | ISO8802.5 4ever |\n\n", "msg_date": "Thu, 6 Jul 2000 21:19:41 -0400 (EDT)", "msg_from": "\"Matthew N. Dodd\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: Revised Copyright: is this morepalatable?" } ]
[ { "msg_contents": "Hi:\n\nSeveral people have complained about forking from the BSD license. If the \nBSD license is so flawed, why not open the discussion to FreeBSD and other \nBSD license users. If the license truely is flawed, it can be \"fixed\" for \nall. Then no one can claim: 1) a PostgreSQL fork, 2) kow tow to corporate \ninterests.\n\nPeople joining this discussion have varying levels of legal knowledge. It \nseems that some clarification by a legal expert on many of these issues is \nneeded. And knowing the variability of \"expertise\" in the legal profession, \nand the importance of the issue, I'd recommend a second or third opinion \n(opening the discusion as above could help with this).\n\nJohn\n\n________________________________________________________________________\nGet Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com\n\n", "msg_date": "Tue, 04 Jul 2000 16:23:05 EDT", "msg_from": "\"John Daniels\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\nHash: SHA1\n\nOn Tue, 4 Jul 2000, John Daniels wrote:\n\n>Several people have complained about forking from the BSD license. If the \n>BSD license is so flawed, why not open the discussion to FreeBSD and other \n>BSD license users. If the license truely is flawed, it can be \"fixed\" for \n>all. Then no one can claim: 1) a PostgreSQL fork, 2) kow tow to corporate \n>interests.\n>\n>People joining this discussion have varying levels of legal knowledge. It \n>seems that some clarification by a legal expert on many of these issues is \n>needed. And knowing the variability of \"expertise\" in the legal profession, \n>and the importance of the issue, I'd recommend a second or third opinion \n>(opening the discusion as above could help with this).\n\nOne question has been asked several times in this thread, and not,\nAFAICT, answered:\n\nWhat is wrong with the current license?\n\nIt's that simple. What's wrong with the current license?\n\nI'd like to point out a couple things that are _not_ wrong with the\ncurrent license:\n\n1.) With the current license, contributors to the code are not opened\n to legal liability for the code they contribute. The BSD license\n very clearly disclaims all warranty on the part of not only UCB but\n also all contributors\n\n2.) The current license does not interfere with commercial products\n based on PostgreSQL. To pick a solid example of this, NCOS is an\n almost direct port of NetBSD 1.3 to various `thin client' hardware.\n Each year, IBM, Oracle, and NCI sell thousands of copies of this\n software. In addition to the great ease with which NetBSD can be\n customized to a specific purpose or ported to new hardware, a key\n reason that NetBSD was chosen over Linux was that if they spend a\n lot of money improving it, they can profit by their work if they\n see fit to.\n\n Remember, in the end that's what its all about, isn't it? We _want_\n people to use PostgreSQL...\n\n3.) The current license does not prevent these companies from\n contributing back. IBM, for example, is preparing to donate back\n a lot of the work they did to make NetBSD run on their (PowerPC\n based) thin client systems.\n\n4.) The current license does not interfere with PostgreSQL being used\n with products under other license. Look at all the claims that KDE\n is violating the GPL. Why? Because its authors put a hell of a lot\n of work into releasing a huge piece of software under GPL, but God\n forbid, some of the other code they used was not GPL'ed...\n\nIn short, there is only one thing that people are accusing the BSD\nlicense of not being which it in fact is not: it is not the GPL.\n\nIt may in fact be that the goals and ideology of the PostgreSQL project\nhave changed so drastically that a move from a BSD license to a GPL is\nin tune with the project's desires. If so, fine, but let's not claim\nthat this is `fixing' the license, or `furthering the purposes\noriginally set out by the PostgreSQL project'. This would be a change\nin those goals, and not one which should take place without consensus\namong those who have worked so hard on it.\n\nThat's my 2 cents. I'm a user of Postgres, not a developer, so I'll\nshut up now :-) \n\n- --\n\t\t\t\tJim Wise\n\t\t\t\[email protected]\n\n-----BEGIN PGP SIGNATURE-----\nVersion: PGPfreeware 5.0i for non-commercial use\nCharset: noconv\n\niQA/AwUBOWPc0y2NgFbJL33VEQIsiQCfWAZuaYbXZu6X3xvYo8e2D/vtcCwAnAhN\nBsRLhw1ninosT/ytRYlBYVDP\n=3NDF\n-----END PGP SIGNATURE-----\n\n", "msg_date": "Wed, 5 Jul 2000 21:11:42 -0400 (EDT)", "msg_from": "Jim Wise <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "On Wed, 5 Jul 2000, Jim Wise wrote:\n\n> -----BEGIN PGP SIGNED MESSAGE-----\n> Hash: SHA1\n> \n> On Tue, 4 Jul 2000, John Daniels wrote:\n> \n> >Several people have complained about forking from the BSD license. If the \n> >BSD license is so flawed, why not open the discussion to FreeBSD and other \n> >BSD license users. If the license truely is flawed, it can be \"fixed\" for \n> >all. Then no one can claim: 1) a PostgreSQL fork, 2) kow tow to corporate \n> >interests.\n> >\n> >People joining this discussion have varying levels of legal knowledge. It \n> >seems that some clarification by a legal expert on many of these issues is \n> >needed. And knowing the variability of \"expertise\" in the legal profession, \n> >and the importance of the issue, I'd recommend a second or third opinion \n> >(opening the discusion as above could help with this).\n> \n> One question has been asked several times in this thread, and not,\n> AFAICT, answered:\n> \n> What is wrong with the current license?\n> \n> It's that simple. What's wrong with the current license?\n> \n> I'd like to point out a couple things that are _not_ wrong with the\n> current license:\n> \n> 1.) With the current license, contributors to the code are not opened\n> to legal liability for the code they contribute. The BSD license\n> very clearly disclaims all warranty on the part of not only UCB but\n> also all contributors\n\nActually, this is the only thing that I do feel the current license is\nmissing ... unless I'm reading something wrong, it all focuses on\ndisclaming \"UNIVERSITY OF CALIFORNIA\"s liability ... that one is very\nspecific ...\n\n\n", "msg_date": "Wed, 5 Jul 2000 23:13:45 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "On Wed, Jul 05, 2000 at 11:13:45PM -0300, The Hermit Hacker wrote:\n> On Wed, 5 Jul 2000, Jim Wise wrote:\n> \n> > I'd like to point out a couple things that are _not_ wrong with the\n> > current license:\n> > \n> > 1.) With the current license, contributors to the code are not opened\n> > to legal liability for the code they contribute. The BSD license\n> > very clearly disclaims all warranty on the part of not only UCB but\n> > also all contributors\n> \n> Actually, this is the only thing that I do feel the current license is\n> missing ... unless I'm reading something wrong, it all focuses on\n> disclaming \"UNIVERSITY OF CALIFORNIA\"s liability ... that one is very\n> specific ...\n\nSince no-one else has mentioned this yet, I will: the Postgres license,\ni.e., the file COPYRIGHT at the top level of the distribution, isn't\nexactly identical to what's commonly known as \"the BSD license\". The\nPostgres copyright, the BSD 4.4 copyright \n(http://www.freebsd.org/copyright/license.html), and the FreeBSD copyright\n(http://www.freebsd.org/copyright/freebsd-license.html), are all\ndifferently worded in parts, although clearly the same in intent. The\nlatter is almost identical to the BSD license template at\nhttp://www.opensource.org/licenses/bsd-license.html . All of them\nexcept ours say something like \"REGENTS AND CONTRIBUTORS\" when they're\ndisclaiming warranties; we just have the University of California doing\nso.\n\nThe simplest way to change our license if we want to make sure that\nit explicitly disclaims warranties on behalf of all contributors seems\nto be to add to the existing California paragraphs a dead standard\nBSD license with our contributors referred to collectively, which is\nwhat Marc has proposed. There may be people who for one reason or\nanother (usually US law, as far as I can see) would like to see more\nchanges, but I can't see what's objectionable about this one.\n\nRichard\n", "msg_date": "Thu, 6 Jul 2000 16:55:16 +0100", "msg_from": "Richard Poole <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "On Thu, 6 Jul 2000, Richard Poole wrote:\n\n> On Wed, Jul 05, 2000 at 11:13:45PM -0300, The Hermit Hacker wrote:\n> > On Wed, 5 Jul 2000, Jim Wise wrote:\n> > \n> > > I'd like to point out a couple things that are _not_ wrong with the\n> > > current license:\n> > > \n> > > 1.) With the current license, contributors to the code are not opened\n> > > to legal liability for the code they contribute. The BSD license\n> > > very clearly disclaims all warranty on the part of not only UCB but\n> > > also all contributors\n> > \n> > Actually, this is the only thing that I do feel the current license is\n> > missing ... unless I'm reading something wrong, it all focuses on\n> > disclaming \"UNIVERSITY OF CALIFORNIA\"s liability ... that one is very\n> > specific ...\n> \n> Since no-one else has mentioned this yet, I will: the Postgres license,\n> i.e., the file COPYRIGHT at the top level of the distribution, isn't\n> exactly identical to what's commonly known as \"the BSD license\". \n\n\tYa, I just clued into that ... throughout all the discussions, I\nnever once thought to do a 'cat /usr/src/COPYRIGHT' on my machine :(\n\n> The\n> Postgres copyright, the BSD 4.4 copyright \n> (http://www.freebsd.org/copyright/license.html), and the FreeBSD copyright\n> (http://www.freebsd.org/copyright/freebsd-license.html), are all\n> differently worded in parts, although clearly the same in intent. The\n> latter is almost identical to the BSD license template at\n> http://www.opensource.org/licenses/bsd-license.html . All of them\n\nDamn, why didn't anyone ever actually look at this stuff before? And ya,\nI'm just as guilty as the rest ...\n\n> The simplest way to change our license if we want to make sure that it\n> explicitly disclaims warranties on behalf of all contributors seems to\n> be to add to the existing California paragraphs a dead standard BSD\n> license with our contributors referred to collectively, which is what\n> Marc has proposed. \n\nQuite frankly, I like the one that OpenSource.Org provides as standard for\nBSD License ... it encompasses everything as one Para instead of repeating\nthings ...\n\nWith wu-ftpd, each source file has this included, as well as a line\nconsisting of \"Copyright (c) <YEAR>, <OWNER>\" for each developer that did\nwork in that file ... \n\nMy personal opinion is to replace the BSD License of 1996 with the BSD\nLicense of today (and keep up with changes to it), as it has been adopt'd\nby other Open Source Projects ... as is provided on\n\n http://www.opensource.org/licenses/bsd-license.html\n\nSomething nice, simple and industry standard:\n\n======================[ README file ]===============================\n\nPostgreSQL Data Base Management System (formerly known as Postgres95)\n\nThis directory contains the _______ release of PostgreSQL, as well as\nvarious post-release patches in the patches directory. See INSTALL for\nthe installation notes and HISTORY for the changes.\n\nWe also have a WWW home page located at: http://www.postgreSQL.org\n\n\n=====================[ COPYRIGHT file ]===============================\n\nCopyright (c) 1994-1996, Regents of the University of California\nCopyright (c) 1996-2000, various contributors (as identified in HISTORY)\n (collectively \"Contributors\")\nAll rights reserved.\n\nRedistribution and use in source and binary forms, with or without\nmodification, are permitted provided that the following conditions are\nmet:\n\n\tRedistributions of source code must retain the above copyright\n\tnotice, this list of conditions and the following disclaimer.\n\n\tRedistributions in binary form must reproduce the above copyright\n\tnotice, this list of conditions and the following disclaimer in\n\tthe documentation and/or other materials provided with the\n\tdistribution.\n\n\tNeither name of the University nor the names of its contributors\n\tmay be used to endorse or promote products derived from this\n\tsoftware without specific prior written permission.\n\nTHIS SOFTWARE IS PROVIDED BY THE UNIVERSITY AND CONTRIBUTORS ``AS IS'' AND\nANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\nIMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\nARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE\nFOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL\nDAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR\nSERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER\nCAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT\nLIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY\nOUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF\nSUCH DAMAGE.\n\n==========================================\n\n", "msg_date": "Thu, 6 Jul 2000 13:19:09 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more palatable?" }, { "msg_contents": "> My personal opinion is to replace the BSD License of 1996 with the BSD\n> License of today (and keep up with changes to it), as it has been adopt'd\n> by other Open Source Projects ... as is provided on\n\nWe can't do this *exactly*: afaict we will need to ship the original UCB\nlicense along with the new license, since that original license required\nthat it continue to travel with the code. We can't unilaterally\nsubstitute another, similar, license.\n\n - Thomas\n", "msg_date": "Fri, 07 Jul 2000 05:28:37 +0000", "msg_from": "Thomas Lockhart <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Revised Copyright: is this more palatable?" } ]
[ { "msg_contents": "Hi,\n\n\n./configure --prefix=~/PostgreSQL --enable-locale --enable-multibyte=LATIN2\nmake:\n\nmake[4]: Entering directory `/home/pavel/pgsql/src/backend/utils/mb'\nmake[4]: `SUBSYS.o' is up to date.\nmake[4]: Leaving directory `/home/pavel/pgsql/src/backend/utils/mb'\nmake[3]: Leaving directory `/home/pavel/pgsql/src/backend/utils'\ngcc -I../include -O2 -Wall -Wmissing-prototypes -Wmissing-declarations -o postgres access/SUBSYS.o bootstrap/SUBSYS.o catalog/SUBSYS.o commands/SUBSYS.o executor/SUBSYS.o lib/SUBSYS.o libpq/SUBSYS.o main/SUBSYS.o parser/SUBSYS.o nodes/SUBSYS.o optimizer/SUBSYS.o port/SUBSYS.o postmaster/SUBSYS.o regex/SUBSYS.o rewrite/SUBSYS.o storage/SUBSYS.o tcop/SUBSYS.o utils/SUBSYS.o -lcrypt -lnsl -ldl -lm -lbsd -lreadline -ltermcap -lncurses -export-dynamic\nutils/SUBSYS.o: In function `lztextlen':\nutils/SUBSYS.o(.text+0x11c68): undefined reference to `PGLZ_RAW_SIZE'\nmake[2]: *** [postgres] Error 1\nmake[2]: Leaving directory `/home/pavel/pgsql/src/backend'\nmake[1]: *** [all] Error 2\nmake[1]: Leaving directory `/home/pavel/pgsql/src'\nmake: *** [all] Error 2\n\nRed Hat Linux 5.2 (I think).\n\nPlease CC: to me, I'm not on the list.\n-- \nPavel Janďż˝k ml.\[email protected]\n", "msg_date": "Tue, 4 Jul 2000 18:32:58 +0200", "msg_from": "[email protected] (Pavel =?iso-8859-2?q?Jan=EDk?= ml.)", "msg_from_op": true, "msg_subject": "current CVS: undefined reference to `PGLZ_RAW_SIZE'" }, { "msg_contents": "> utils/SUBSYS.o: In function `lztextlen':\n> utils/SUBSYS.o(.text+0x11c68): undefined reference to `PGLZ_RAW_SIZE'\n\nDid you do this from a completely fresh tree? If not, did you do a \"make\nclean\" first? I built this morning without trouble...\n\n - Thomas\n", "msg_date": "Wed, 05 Jul 2000 04:28:11 +0000", "msg_from": "Thomas Lockhart <[email protected]>", "msg_from_op": false, "msg_subject": "Re: current CVS: undefined reference to `PGLZ_RAW_SIZE'" }, { "msg_contents": "Thomas Lockhart <[email protected]> writes:\n>> utils/SUBSYS.o: In function `lztextlen':\n>> utils/SUBSYS.o(.text+0x11c68): undefined reference to `PGLZ_RAW_SIZE'\n\n> Did you do this from a completely fresh tree? If not, did you do a \"make\n> clean\" first? I built this morning without trouble...\n\nNo, he's right: the lztext stuff has an undefined reference that's\nburied inside #ifdef MULTIBYTE. Not sure if the answer is to add\nthe missing #include to lztext.c, or if the code is wrong. Jan?\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 05 Jul 2000 02:32:06 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: current CVS: undefined reference to `PGLZ_RAW_SIZE' " }, { "msg_contents": "> -----Original Message-----\n> From: [email protected] [mailto:[email protected]]On\n> Behalf Of Tom Lane\n>\n> Thomas Lockhart <[email protected]> writes:\n> >> utils/SUBSYS.o: In function `lztextlen':\n> >> utils/SUBSYS.o(.text+0x11c68): undefined reference to `PGLZ_RAW_SIZE'\n>\n> > Did you do this from a completely fresh tree? If not, did you do a \"make\n> > clean\" first? I built this morning without trouble...\n>\n> No, he's right: the lztext stuff has an undefined reference that's\n> buried inside #ifdef MULTIBYTE. Not sure if the answer is to add\n> the missing #include to lztext.c, or if the code is wrong. Jan?\n>\n\nAdding a header file pg_lzcompress.h isn't sufficient.\nType lztext was changed to mean 'varattrib' not 'PGLZ_Header'.\nIn addtion lztext.c was pretty changed yesterday and other\nPGLZ_RAW_SIZE() calls were removed completely.\nI don't know how to fix it for now and the future changes.\n\nRegards.\n\nHiroshi Inoue\[email protected]\n\n", "msg_date": "Wed, 5 Jul 2000 16:11:07 +0900", "msg_from": "\"Hiroshi Inoue\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: current CVS: undefined reference to `PGLZ_RAW_SIZE' " }, { "msg_contents": "Tom Lane wrote:\n> Thomas Lockhart <[email protected]> writes:\n> >> utils/SUBSYS.o: In function `lztextlen':\n> >> utils/SUBSYS.o(.text+0x11c68): undefined reference to `PGLZ_RAW_SIZE'\n>\n> > Did you do this from a completely fresh tree? If not, did you do a \"make\n> > clean\" first? I built this morning without trouble...\n>\n> No, he's right: the lztext stuff has an undefined reference that's\n> buried inside #ifdef MULTIBYTE. Not sure if the answer is to add\n> the missing #include to lztext.c, or if the code is wrong. Jan?\n\n To me it looks more that the code is wrong, left over from\n the old self compressing version of lztext. I think it should\n be strlen(s1) instead.\n\n I changed it to that. Could you please check if it works?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Wed, 5 Jul 2000 12:06:49 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: current CVS: undefined reference to `PGLZ_RAW_SIZE'" }, { "msg_contents": " From: [email protected] (Jan Wieck)\n Date: Wed, 5 Jul 2000 12:06:49 +0200 (MEST)\n\nHi,\n\n > I changed it to that. Could you please check if it works?\n\nyes, it can be build now. Thank you.\n-- \nPavel Janďż˝k ml.\[email protected]\n", "msg_date": "Thu, 6 Jul 2000 23:06:09 +0200", "msg_from": "[email protected] (Pavel =?iso-8859-2?q?Jan=EDk?= ml.)", "msg_from_op": true, "msg_subject": "Re: current CVS: undefined reference to `PGLZ_RAW_SIZE'" } ]
[ { "msg_contents": "Just a short note that I'm leaving the office for vacation. ECPG stuff will\nhave to wait until I'm back, or until someone else takes care. :-)\n\nMichael\n-- \nMichael Meskes\[email protected]\nGo SF 49ers! Go Rhein Fire!\nUse Debian GNU/Linux! Use PostgreSQL!\n", "msg_date": "Tue, 4 Jul 2000 19:56:26 +0200", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "vacation" } ]
[ { "msg_contents": "Hi,\n\nWhen I run a sequence test, I get the following error message:\n\n=============== installing languages... =================\ninstalling PL/pgSQL .. createlang: missing required argument PGLIB\ndirectory\n\n\nCould anyone tell me why and how to resolv it?\n\n\nThanks.\n\n-Steven\n [email protected]\n", "msg_date": "Wed, 05 Jul 2000 02:39:40 +0800", "msg_from": "steven wu <[email protected]>", "msg_from_op": true, "msg_subject": "sequential test error" }, { "msg_contents": "\nOn Wed, 5 Jul 2000, steven wu wrote:\n\n> Hi,\n> \n> When I run a sequence test, I get the following error message:\n> \n> =============== installing languages... =================\n> installing PL/pgSQL .. createlang: missing required argument PGLIB\n> directory\n> \n> \n> Could anyone tell me why and how to resolv it?\n\nfor example:\n\nexport PGLIB=/usr/lib/postgresql/lib\nmake runtest\n\n\n\t\t\t\t\tKarel\n\n", "msg_date": "Tue, 4 Jul 2000 20:54:44 +0200 (CEST)", "msg_from": "Karel Zak <[email protected]>", "msg_from_op": false, "msg_subject": "Re: sequential test error" }, { "msg_contents": "Karel Zak <[email protected]> writes:\n> On Wed, 5 Jul 2000, steven wu wrote:\n>> When I run a sequence test, I get the following error message:\n>> \n>> =============== installing languages... =================\n>> installing PL/pgSQL .. createlang: missing required argument PGLIB\n>> directory\n\n> export PGLIB=/usr/lib/postgresql/lib\n> make runtest\n\nHmm, I had not realized that createlang fails without a PGLIB setting\n(probably because I do all my Postgres work with PGLIB and PGDATA\ndefined... :-()\n\nWouldn't it be a good idea to install the right path into it as a\ndefault value at configure time? Likewise for the other scripts that\nneed this info?\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 22:01:58 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "createlang and PGLIB (was Re: [GENERAL] sequential test error)" }, { "msg_contents": "On Tue, 4 Jul 2000, Tom Lane wrote:\n\n> Wouldn't it be a good idea to install the right path into it as a\n> default value at configure time? Likewise for the other scripts that\n> need this info?\n\nYes, that is precisely the plan.\n\n-- \nPeter Eisentraut Sernanders vaeg 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Wed, 5 Jul 2000 07:48:39 -0400 (EDT)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: createlang and PGLIB (was Re: [GENERAL] sequential\n test error)" } ]
[ { "msg_contents": "For discussion:\n\n First what the current implementation and the yet to be done\n proposals do.\n\n All varlena data types (text, char, varchar, arrays) will\n finally be toastable. Every table that uses such types\n will have a secondary relation to move off attributes.\n\n The toaster allways tries to keep a main tuple small\n enough so that at minimum 4 tuples fit into a block. One\n had complained about, and I explain later why I think\n it's a good decision anyway.\n\n This strategy already covers most possible index problems. If\n the main tuple fits into 2K after toasting, any combination\n of attributes out of it will too. The only thing not covered\n are functional indices.\n\n In real world scenarios, indices are usually set up on small\n key values. These are very likely to be kept plain in the\n main tuple by the toaster, becuase it looks at the biggest\n values first. So an index (built out of the values in the\n main tuple after toasting) will also contain the plain\n values. Thus, index scans will not require toast fetches in\n turn. Except the indexed attribute had at some point a huge\n value.\n\n The current TOAST implementation hooks into the heap access\n methods only. Automagically covering the index issues due to\n the 2K approach. Fact is, that if more toast entries can get\n produced during index inserts, we need to take care for them\n during vacuum (the only place where index items get removed).\n Alot of work just to support huge functional indices - IMHO\n not worth the efford right now. Let's better get some\n experience with the entire thing before going too far.\n\n Why is it good to keep the main tuple below 2K? First because\n of the above side effects for indices. Second, because in the\n most likely case of small indexed attributes, more main\n tuples (that must be fetched for the visibility checks\n anyway) will fit into one block. That'll cause more blocks of\n the relation to fit into the given shared memory buffer cache\n and avoids I/O during index scans.\n\n My latest tests load a 1.1M tree full of .html files into a\n database. The result is a 140K heap plus 300K toast\n relation. Without that 2K approach, the result is a 640K heap\n plus 90K toastrel only. Since all compression is done on\n single entries, it scales linear, so that a 1.1G tree will\n result in a 140M heap plus 300M toastrel vs. a 640M heap plus\n 90M toastrel. No need to bechmark it - I know which strategy\n wins.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Tue, 4 Jul 2000 20:42:48 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": true, "msg_subject": "TOAST on indices" }, { "msg_contents": "At 20:42 4/07/00 +0200, Jan Wieck wrote:\n>\n> So an index (built out of the values in the\n> main tuple after toasting) will also contain the plain\n> values. Thus, index scans will not require toast fetches in\n> turn. Except the indexed attribute had at some point a huge\n> value.\n\nSo that, for toasted attrs, the indexes will be no use for sorting. I agree\nthat in the majority of cases this is not a problem, but if the entire\ntuple gets toasted because it is too large it becomes a problem.\n\nI agree that this is only a problem if indexes are used in sorting, and may\nnot be a problem if one builds an index on 'substr(toastable-field,1,20)',\nbut I think you are suggesting not supporting functional indexes,\nbelow...but maybe I've missed the point.\n\n\n\n> The current TOAST implementation hooks into the heap access\n> methods only. Automagically covering the index issues due to\n> the 2K approach. Fact is, that if more toast entries can get\n> produced during index inserts, we need to take care for them\n> during vacuum (the only place where index items get removed).\n> Alot of work just to support huge functional indices - IMHO\n> not worth the efford right now. Let's better get some\n> experience with the entire thing before going too far.\n>\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Thu, 06 Jul 2000 13:10:18 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: TOAST on indices" }, { "msg_contents": "Philip Warner wrote:\n> At 20:42 4/07/00 +0200, Jan Wieck wrote:\n> >\n> > So an index (built out of the values in the\n> > main tuple after toasting) will also contain the plain\n> > values. Thus, index scans will not require toast fetches in\n> > turn. Except the indexed attribute had at some point a huge\n> > value.\n>\n> So that, for toasted attrs, the indexes will be no use for sorting. I agree\n> that in the majority of cases this is not a problem, but if the entire\n> tuple gets toasted because it is too large it becomes a problem.\n\n That ain't true, entirely. First of all, only single\n attributes get toasted. Never a complete tuple but maybe all\n of it's attributes (if it is a table with many, many\n attributes or all of them are big).\n\n If so, well, the sort might become damned slow. Assuming all\n the rows selected have the attribute to sort on toasted, each\n comparision will require two index scans (plus possibly\n decompression) during the sort.\n\n But tell me, do you know of real world DB installations where\n indices on fields likely to be >1K exist? What is such an\n index good for? Fast reverse lookup of 65536-bit RSA keys?\n\n The system won't complain, nor will it bail out in such a\n situation. That it won't behave as good as it could is a\n con. Maybe we should tell on our web pages that someone who\n wants to create indices on multi-K attributes should better\n look for another DB, because Postgres is slow in that case?\n\n> I agree that this is only a problem if indexes are used in sorting, and may\n> not be a problem if one builds an index on 'substr(toastable-field,1,20)',\n> but I think you are suggesting not supporting functional indexes,\n> below...but maybe I've missed the point.\n>\n> > The current TOAST implementation hooks into the heap access\n> > methods only. Automagically covering the index issues due to\n> > the 2K approach. Fact is, that if more toast entries can get\n> > produced during index inserts, we need to take care for them\n> > during vacuum (the only place where index items get removed).\n> > Alot of work just to support huge functional indices - IMHO\n> > not worth the efford right now. Let's better get some\n> > experience with the entire thing before going too far.\n\n Yeah - you missed me here.\n\n In the case of a functional index, the function would seldom\n return one of the original tuples attribute values. Usually\n those functions manipulate one or more attributes to compute\n a completely new value (like your substr() example above).\n In the TOAST world, any such function returns a plain, fully\n expanded, in memory value.\n\n So even if the toaster had worked on the main tuple and\n compressed/moved off some attributes, the value that is\n computed during index tuple creation is of full size. Having\n a char(20000) attribute, the toaster will shrink it down so\n the main tuple will fit. But a functional index like\n \"substr(att, 1, 10000)\" must fail, because during index tuple\n creation the funtion is evalueated and creates a 10000 byte\n value.\n\n In the current implementation, non-functional indices on huge\n fields should be supported (there still are bugs because it\n doesn't work right now). For functional ones, the old\n restriction of \"index-tuple must be smaller than supported\n tuple size of index method\" applies.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#================================================== [email protected] #\n\n\n", "msg_date": "Fri, 7 Jul 2000 00:41:43 +0200 (MEST)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": true, "msg_subject": "Re: TOAST on indices" } ]
[ { "msg_contents": "> Before I do that I want to confirm the major problem I had w/postgres:\n> the 8K tuple limit. When trying to import some tables from MySQL,\n> postgres kept choking because MySQL has no such limit on the size of a\n> row in the database (text fields on MySQL can be multi-megabyte).\n\nJan is working on TOAST for 7.1 - there will be no 8K limit any more...\n\n> Long-term stability:\n> Postgres is undoubtably the long-run winner in stability, \n> whereas MySQL will freak out or die when left running for more than a\nmonth \n> at a time. But if you ever do have a problem with postgres, you generally\n> have to nuke the database and recover from a backup, as there are no \n> known tools to fix index and database corruption. For a long-running\npostgres\n> database, you will occasionally have to drop indexes and re-create them,\n> causing downtime.\n\nI'm implementing WAL for 7.1 - there will be true after crash recovery...\n\nVadim\n", "msg_date": "Tue, 4 Jul 2000 12:47:32 -0700 ", "msg_from": "\"Mikheev, Vadim\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: Article on MySQL vs. Postgres" } ]
[ { "msg_contents": "I think we need a --with-zlib switch for the new pg_dump. Or at least a\n--without-zlib switch, if you think that it's widespread enough.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Tue, 4 Jul 2000 21:57:31 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "zlib for pg_dump" }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n> I think we need a --with-zlib switch for the new pg_dump. Or at least a\n> --without-zlib switch, if you think that it's widespread enough.\n\nSeems reasonable to me to look for libz automatically in the library\nsearch path. A --with switch would only serve as an addition to the\nLIBS directory list, and as I commented in another connection I see no\ngood reason for adding a bunch of variant spellings of --with-libs ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 16:11:15 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: zlib for pg_dump " }, { "msg_contents": "Tom Lane writes:\n\n> > I think we need a --with-zlib switch for the new pg_dump. Or at least a\n> > --without-zlib switch, if you think that it's widespread enough.\n> \n> Seems reasonable to me to look for libz automatically in the library\n> search path. A --with switch would only serve as an addition to the\n> LIBS directory list, and as I commented in another connection I see no\n> good reason for adding a bunch of variant spellings of --with-libs ...\n\nAs I might have commented before, I'm not so excited about adding and\ndropping features without explicit user interaction. zlib is probably\nwidespread enough to assume it by default, but then the build must fail if\nthe library is not installed, and the user must indicate his consent by\nusing the --without-zlib flag. We have to be able to determine the set of\nfeatures that will be build by examining the configure command line, not\nby scanning the configure output, or worse yet, by finding out that the\nprogram doesn't behave as expected.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Wed, 5 Jul 2000 18:36:33 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Re: zlib for pg_dump " }, { "msg_contents": "At 18:36 5/07/00 +0200, Peter Eisentraut wrote:\n>Tom Lane writes:\n>> \n>> Seems reasonable to me to look for libz automatically in the library\n>> search path. A --with switch would only serve as an addition to the\n>> LIBS directory list, and as I commented in another connection I see no\n>> good reason for adding a bunch of variant spellings of --with-libs ...\n...\n>We have to be able to determine the set of\n>features that will be build by examining the configure command line\n\nThe feature (compressed backup output in pg_dump) is not readily deducible\nfrom '--with-zlib'. It might be better to output a message from configure\nif a needed library is not found:\n\n\"zlib compression library not found, pg_dump archives will not support\ncompression\".\n\nif, in the future, there is a gzip datatype, we could also output 'gzip\ndata types will not be supported' etc.\n\nAlternatively, have a very specific flag: --with-compressed-dumps, then die\nif there is no zlib.\n\nBut I'd stick with the check for zlib, since it is pretty common. The\n*only* feature you lose by not having it is the compressed output in dumps:\nif you request compression, it issues a warning and produces uncompressed\noutput. The archives are still readable when zlib is installed (and vice\nverca in the case of -Z0).\n\n\n\n\n\n\n\n\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Thu, 06 Jul 2000 02:49:31 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: zlib for pg_dump " }, { "msg_contents": "On Wed, 5 Jul 2000, Peter Eisentraut wrote:\n\n> Tom Lane writes:\n> \n> > > I think we need a --with-zlib switch for the new pg_dump. Or at least a\n> > > --without-zlib switch, if you think that it's widespread enough.\n> > \n> > Seems reasonable to me to look for libz automatically in the library\n> > search path. A --with switch would only serve as an addition to the\n> > LIBS directory list, and as I commented in another connection I see no\n> > good reason for adding a bunch of variant spellings of --with-libs ...\n> \n> As I might have commented before, I'm not so excited about adding and\n> dropping features without explicit user interaction. zlib is probably\n> widespread enough to assume it by default, but then the build must\n> fail if the library is not installed, and the user must indicate his\n> consent by using the --without-zlib flag. We have to be able to\n> determine the set of features that will be build by examining the\n> configure command line, not by scanning the configure output, or worse\n> yet, by finding out that the program doesn't behave as expected.\n\nI don't agree ... unless we're gonna add this for libreadline\n(optional) and anything else that is \"optional\". IMHO, if the system has\nthe functionality that extends the usefulness of the software, don't make\nthe installer go off looking for how to make use of it ...\n\n\n", "msg_date": "Wed, 5 Jul 2000 14:44:27 -0300 (ADT)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: zlib for pg_dump " }, { "msg_contents": "Philip Warner writes:\n\n> The feature (compressed backup output in pg_dump) is not readily deducible\n> from '--with-zlib'. It might be better to output a message from configure\n> if a needed library is not found:\n> \n> \"zlib compression library not found, pg_dump archives will not support\n> compression\".\n\nThat's exactly what we should not do. You forget that not only humans run\nconfigure scripts, it might be other programs, which can't \"read\".\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Thu, 6 Jul 2000 02:15:33 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Re: zlib for pg_dump " }, { "msg_contents": "The Hermit Hacker writes:\n\n> I don't agree ... unless we're gonna add this for libreadline\n> (optional) and anything else that is \"optional\".\n\nWe already do this for everything else that's optional, except for\nreadline, which I consider wrong, but there's probably no support to be\ngained for that. The fact that it is virtually impossible to determine\nwhether readline and/or history support is supported in psql until you\nhave installed and used it is a major source of user problems.\n\n> IMHO, if the system has the functionality that extends the usefulness\n> of the software, don't make the installer go off looking for how to\n> make use of it ...\n\nThat's why I suggested assuming it by default and turn if *off* manually.\n\nThe reason why I'm so keen about this is this: configure scripts are not\nonly run by humans, but also by programs, and the only thing programs can\nrely on is a non-zero exit status. config.status is one such program.\nOthers are source RPMs and other packaging mechanisms. Rebuilding a source\nRPM is an enlightening experience: You get all kinds of garbage flying by\non your screen, and the end you only hope that something reasonable came\nout. But if not even the build script can determine what it's going to\nbuild, then you really lose.\n\nThe failure modes are plenty. You don't even have to be on a different\nmachine or fiddle with your installed packages. Maybe the sysadmin simply\nreran /sbin/ldconfig after forgetting for a few months. Before you know it\nyou have shipped the package off to thousands of people who wonder why\nfeature X won't work.\n\nThe bottom line is, the packager/builder needs a way to specify what\nexactly is going to be built.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Thu, 6 Jul 2000 18:13:31 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Re: zlib for pg_dump " }, { "msg_contents": "At 18:13 6/07/00 +0200, Peter Eisentraut wrote:\n>\n>The bottom line is, the packager/builder needs a way to specify what\n>exactly is going to be built.\n>\n\nWhat about a compromise: --enable-zlib, --disable-zlib. \n\nconfigure should check for zlib if neither of these is specified, otherwise\nif just uses these. \n\nYou also need to be careful: if --enable-zlib is specified and configure\ncould not find zlib.h, something needs to be done...\n\n\n----------------------------------------------------------------\nPhilip Warner | __---_____\nAlbatross Consulting Pty. Ltd. |----/ - \\\n(A.C.N. 008 659 498) | /(@) ______---_\nTel: (+61) 0500 83 82 81 | _________ \\\nFax: (+61) 0500 83 82 82 | ___________ |\nHttp://www.rhyme.com.au | / \\|\n | --________--\nPGP key available upon request, | /\nand from pgp5.ai.mit.edu:11371 |/\n", "msg_date": "Fri, 07 Jul 2000 02:25:27 +1000", "msg_from": "Philip Warner <[email protected]>", "msg_from_op": false, "msg_subject": "Re: zlib for pg_dump " }, { "msg_contents": "Philip Warner writes:\n\n> What about a compromise: --enable-zlib, --disable-zlib. \n> \n> configure should check for zlib if neither of these is specified, otherwise\n> if just uses these. \n\nI like that.\n\n> You also need to be careful: if --enable-zlib is specified and configure\n> could not find zlib.h, something needs to be done...\n\nTrue, especially on RPM systems that don't install the header files by\ndefault.\n\nActually, there is a CHECK_ZLIB macro in the official Autoconf macro\narchive that does this sort of thing, but it rather follows my original\nsuggestions of yes/no, rather than your idea of yes/no/maybe. But maybe we\ncan reutilize that anyway in some form.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Fri, 7 Jul 2000 18:16:41 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Re: zlib for pg_dump " } ]
[ { "msg_contents": "A while ago I suggested to create a backend internal function that takes\nan internal type name and an atttypmod and formats it to a canonical SQL\ntype, to be used by pg_dump, psql, and anyone who feels like it.\n\nI have it written now, here's an example of how it works. Any comments?\n\n\nregression=# select c.relname, a.attname, format_type(t.typname, a.atttypmod)\nfrom pg_class c, pg_attribute a, pg_type t\nwhere c.oid=a.attrelid and t.oid=a.atttypid\nand c.relname not like 'pg\\_%' and a.attnum > 0\norder by 1;\n\n relname | attname | format_type\n-------------------------+---------------+------------------------\n a | aa | text\n a_star | class | character(1)\n a_star | aa | integer\n a_star | a | text\n abstime_tbl | f1 | abstime\n aggtest | a | smallint\n aggtest | b | real\n arrtest | a | smallint[]\n arrtest | b | integer[]\n arrtest | c | name[]\n arrtest | d | text[]\n arrtest | e | double precision[]\n arrtest | f | character(5)[]\n arrtest | g | character varying(5)[]\n b | aa | text\n...\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n\n", "msg_date": "Tue, 4 Jul 2000 21:57:52 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Dumping SQL type names" }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n> A while ago I suggested to create a backend internal function that takes\n> an internal type name and an atttypmod and formats it to a canonical SQL\n> type, to be used by pg_dump, psql, and anyone who feels like it.\n\n> I have it written now, here's an example of how it works. Any comments?\n\nFor the most part, a function working from type OID would be more\nconvenient, I should think. Possibly offer both ...\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 04 Jul 2000 22:13:08 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Dumping SQL type names " } ]
[ { "msg_contents": "There seems to be a general issue here with usernames.\n\nPG creates only lower case usernames. I am happy that if usEr can't type, it's\nhis problem but I have a definite problem with Identd authentication from an NT\nclient (or ident server whichever you choose ) which insists on pretty-printing\nthe identd response.\n\nI have therefore hacked my code in auth.c. This seems to be of no other\nconsequence but I am not a C programmer nor did I have time to check it out\nelsewhere.\n\nI may be masking another, more general, issue but this does for me ! Can we get\nit or equivalent in future releases please ?\n\nRegards.\n\nBTW The actaul code works - this is a cut&paste.\n\n*---------------------------------------------------------------------------\n Talk to the ident server on the remote host and find out who owns the\n connection described by \"port\". Then look in the usermap file under\n the usermap *auth_arg and see if that user is equivalent to\n Postgres user *user.\n\n Return STATUS_OK if yes.\n---------------------------------------------------------------------------*/\n\tbool\t\tchecks_out;\n\tbool\t\tident_failed;\n\n\t/* We were unable to get ident to give us a username */\n\tchar\t\tident_username[IDENT_USERNAME_MAX + 1];\n!!!!!!!!!\tchar*\t\tc = ident username; \n \n\n\n\t/* The username returned by ident */\n\n\tident(raddr->sin_addr, laddr->sin_addr,\n\t\t raddr->sin_port, laddr->sin_port,\n\t\t &ident_failed, ident_username);\n\n\tif (ident_failed)\n\t\treturn STATUS_ERROR;\n\n|||||||\twhile (*c) { *c++ |= 0x20;}\n\n\tverify_against_usermap(postgres_username, ident_username, auth_arg,\n\t\t\t\t\t\t &checks_out);\n\n\treturn checks_out ? STATUS_OK : STATUS_ERROR;\n}\n", "msg_date": "Tue, 4 Jul 2000 21:37:21 +0100", "msg_from": "Brian Piatkus <[email protected]>", "msg_from_op": true, "msg_subject": "Case sensitivity" }, { "msg_contents": "Brian Piatkus <[email protected]> writes:\n> PG creates only lower case usernames.\n\nNot at all. However, if we hacked the ident code as you suggest,\nit's true that the ident code would support only lower-case names.\n\nUse double quotes if you need to create mixed-case names.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 05 Jul 2000 12:06:07 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Case sensitivity " }, { "msg_contents": "You haven't given the hba.conf entry you are using, so I'm assuming\nyou're using ident with the sameuser usermap.\n\nYou can create mixed case names by double quoting the name\n(as mentioned by Tom Lane)\n\nOr, if you're using both unix and NT machines so the names come\nout differently for the same users, you might be able to get away\nwith just using a usermap file. If you really didn't feel like making\nand keep the usermap up to date, you might want to look at how\nthe sameuser usermap is defined and make an equivalent\nthat did what you wanted (or do case insensitive matches),\nprobably something like [not at all tested or even tried to compile]:\n\nelse if (strcmp(usermap_name, \"caseless_sameuser\")) == 0)\n{\n if (strcasecmp(ident_username, pguser) == 0)\n *checks_out_p = true;\n else\n *check_out_p = false;\n}\n\n\nI think the changes below are a bit drastic and make sure that\nnoone will ever be able to use mixed case names with it and\nalso are assuming that the usernames can't contain any characters\nthat might get translated by the |= that aren't A-Z.\n\n----- Original Message -----\nFrom: \"Brian Piatkus\" <[email protected]>\nTo: <[email protected]>\nSent: Tuesday, July 04, 2000 1:37 PM\nSubject: [HACKERS] Case sensitivity\n\n\n> There seems to be a general issue here with usernames.\n>\n> PG creates only lower case usernames. I am happy that if usEr can't type,\nit's\n> his problem but I have a definite problem with Identd authentication from\nan NT\n> client (or ident server whichever you choose ) which insists on\npretty-printing\n> the identd response.\n>\n> I have therefore hacked my code in auth.c. This seems to be of no other\n> consequence but I am not a C programmer nor did I have time to check it\nout\n> elsewhere.\n>\n> I may be masking another, more general, issue but this does for me ! Can\nwe get\n> it or equivalent in future releases please ?\n>\n> Regards.\n>\n> BTW The actaul code works - this is a cut&paste.\n>\n>\n*---------------------------------------------------------------------------\n> Talk to the ident server on the remote host and find out who owns the\n> connection described by \"port\". Then look in the usermap file under\n> the usermap *auth_arg and see if that user is equivalent to\n> Postgres user *user.\n>\n> Return STATUS_OK if yes.\n> --------------------------------------------------------------------------\n-*/\n> bool checks_out;\n> bool ident_failed;\n>\n> /* We were unable to get ident to give us a username */\n> char ident_username[IDENT_USERNAME_MAX + 1];\n> !!!!!!!!! char* c = ident username;\n>\n>\n>\n> /* The username returned by ident */\n>\n> ident(raddr->sin_addr, laddr->sin_addr,\n> raddr->sin_port, laddr->sin_port,\n> &ident_failed, ident_username);\n>\n> if (ident_failed)\n> return STATUS_ERROR;\n>\n> ||||||| while (*c) { *c++ |= 0x20;}\n>\n> verify_against_usermap(postgres_username, ident_username, auth_arg,\n> &checks_out);\n>\n> return checks_out ? STATUS_OK : STATUS_ERROR;\n> }\n>\n\n", "msg_date": "Wed, 5 Jul 2000 14:09:00 -0700", "msg_from": "\"Stephan Szabo\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Case sensitivity" } ]
[ { "msg_contents": "\n>The only part that I believe at least one person had an issue with was:\n\n>\"Any person who contributes or submits any modification or other change\n>to the PostgreSQL software or documentation grants irrevocable,\n>non-exclusive, worldwide permission, without charge, to use, copy,\n>further modify and distribute the same under the terms of this license.\"\n\n>Quite frankly, all I'm reading into this paragraph is that once\n>committed, Jan (as a recent example) couldn't come along and pull out all\n>his TOAST changes ... could you imagine the hell that would wreak were he\n>(or anyone else) were to pull crucial changes after others have built\n>upon it?\n\nA problem I see here is that \"contributes or submits\" is not defined. I\npresume that what is probably intended is something like \"submits to the\nCVS repository or the patches mailing list of the PostgreSQL project\".\nOtherwise the term has no meaning. However, this would require some\ndefinition of \"PostgreSQL project\", and IMHO any such definition would be\na Bad Thing. I suggest it is better for PostgreSQL, the software, to stand\nalone as the entity distributed. The fact that a \"project\" exists to\nmaintain and develop it should not figure in the licence. Otherwise, to\npick an extreme example, suppose Oracle makes all the core developers an\noffer too good to refuse, and someone else is left to pick up the pieces\nand fork the project to keep it open. What rights does this forked project\nhave in the scheme of things? What terms apply to someone who submits\npatches to this new project?\n\nFor that matter, suppose someone forks the project for other reasons; the\ncurrent licence gives anyone the right to do this at any time. With this\nnew suggested clause, the right to fork becomes murky. Forking is a Bad\nThing in general, but having the right to fork a project is the only\nguarantee that the software will always be open.\n\nNow one thing that might be sensible is to adopt a policy of only\naccepting patches or cvs checkins that are provided irrevocably under the\nterms of the PostgreSQL licence. But that is a policy for the PostgreSQL\nproject to adopt, not a term of the licence for redistribution of the\nPostgreSQL software.\n\nTim\n\n--\n-----------------------------------------------\nTim Allen [email protected]\nProximity Pty Ltd http://www.proximity.com.au/\n http://www4.tpg.com.au/users/rita_tim/\n\n\n", "msg_date": "Wed, 5 Jul 2000 10:09:22 +1000 (EST)", "msg_from": "Tim Allen <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Revised Copyright: is this more palatable?" }, { "msg_contents": "Tim Allen wrote:\n\n> Now one thing that might be sensible is to adopt a policy of only\n> accepting patches or cvs checkins that are provided irrevocably under the\n> terms of the PostgreSQL licence. But that is a policy for the PostgreSQL\n> project to adopt, not a term of the licence for redistribution of the\n> PostgreSQL software.\n\nThat's exactly right.\n\nFirstly I don't believe there is an issue here. There is an implied\nlicence for patches.\n\nBut if there is any doubt it is the project's responsibility to ensure\nthat code that is accepted is under the right licence. If that means\npatches are submitted with the appropriate licence so be it.\n\nIf find it somewhat doubtful also that putting this clause into the\ncopyright would have the desired effect. The clause effectively says\n\"Whatever patches you make available - well we have a licence whether\nyou like it or not\". I can't see a court enforcing this. Either the\nlicence is implied by merely submitting it to the existing project by\ncommon law, or the writer explicitely grants the licence. Trying to grab\nthe licence from the other party just by name it and claim it, I can't\nsee it working.\n", "msg_date": "Wed, 05 Jul 2000 10:33:14 +1000", "msg_from": "Chris Bitmead <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Re: [HACKERS] Revised Copyright: is this more palatable?" } ]
[ { "msg_contents": "As we know, the inet and cidr types are still broken in several ways,\namongst others input and output functions, operators, ordering. I've\ncollected the bug reports from the last year or so from the archives.\n\nThere's apparently a lack of understanding of what exactly are these types\nare supposed to do. Therefore, instead of addressing each bug\nindividually, let me first state what I reconstructed as the specification\nof these types, and then add what is currently wrong with it.\n\n* CIDR\n\nThe cidr type stores the identity of an IP _network_. A network\nspecification is of the form 'x.x.x.x/y'. The documentation states that if\ny is omitted then it is constructed from the old A, B, C class scheme. So\nbe it. In a real world network, the bits (y+1)...32 have to be zero, but\nthe cidr type does not currently enforce this. This has been the source of\nbugs in the past, and no doubt the source of some confusion as well. I\npropose that cidr _reject_ input of the form '127.0.0.5/16'. If you think\nabout it, this is the same as int4 rejecting 3.5 as input.\n\n* INET\n\nThe inet type stores the identity of an IP _host_. A host specification is\nof the form 'x.x.x.x'. Optionally, the inet type also stores the identity\nof the network the host is in. E.g., '127.0.0.5/16' means the host\n127.0.0.5 in the network 127.0/16.\n\n* Type equivalency\n\nThis has also been a source of problems. I propose that cidr and inet are\nnot made equivalent types at any level. No automatic casting either. A\nnetwork and a host are not the same thing. To construct a cidr value from\nan inet value, you'd have to use some sort of (to be created) network()\nfunction, e.g., network('127.0.0.5/16') => '127.0/16'. IMO, there is no\nreasonable way to construct an inet value from a cidr value.\n\n* Operators\n\nBecause the types are equivalent, the operators have also been bunched\ntogether in confusing ways. I propose that ordering operators (>, +, <)\nbetween inet and cidr be eliminated, they do not make sense. The only\nuseful operation between cidr and inet is the << (\"contains\") operator.\nOrdering withing cidr and inet be defined in terms of their bit\nrepresentation, as is the case now. The << family of operators should also\nbe removed for the inet type -- a host cannot \"contain\" another host. What\nyou probably wanted is `inet1 << network(inet2)'.\n\n\nDoes anyone see this differently? If not, can we agree on this\nspecification?\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Wed, 5 Jul 2000 02:12:35 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Repair plan for inet and cidr types" }, { "msg_contents": "Thus spake Peter Eisentraut\n> There's apparently a lack of understanding of what exactly are these types\n> are supposed to do. Therefore, instead of addressing each bug\n> individually, let me first state what I reconstructed as the specification\n> of these types, and then add what is currently wrong with it.\n\nI have been browsing through the old messages on the topic. There was, in\nfact some very good work defining the type before anyone actually started\nto code. There was a surprising amount of controversy over the actual\ndefinitions but I think in the end we hammered it out at least to the\npoint that everyone could work with it.\n\n> * CIDR\n> \n> The cidr type stores the identity of an IP _network_. A network\n> specification is of the form 'x.x.x.x/y'. The documentation states that if\n> y is omitted then it is constructed from the old A, B, C class scheme. So\n> be it. In a real world network, the bits (y+1)...32 have to be zero, but\n> the cidr type does not currently enforce this. This has been the source of\n> bugs in the past, and no doubt the source of some confusion as well. I\n> propose that cidr _reject_ input of the form '127.0.0.5/16'. If you think\n> about it, this is the same as int4 rejecting 3.5 as input.\n\nThere is also the option of accepting it but masking out the host bits\nbefore storing it. That gives us automatic conversion if we store an\ninet into a cidr if our intent is to store the network part.\n\nWhat sort of bugs do you think it caused btw?\n\n> * INET\n> \n> The inet type stores the identity of an IP _host_. A host specification is\n> of the form 'x.x.x.x'. Optionally, the inet type also stores the identity\n> of the network the host is in. E.g., '127.0.0.5/16' means the host\n> 127.0.0.5 in the network 127.0/16.\n\nThat sounds right. We also allowed for hosts to be stored implicitely by\nsimply making the netmask /32.\n\n> * Type equivalency\n> \n> This has also been a source of problems. I propose that cidr and inet are\n> not made equivalent types at any level. No automatic casting either. A\n> network and a host are not the same thing. To construct a cidr value from\n> an inet value, you'd have to use some sort of (to be created) network()\n> function, e.g., network('127.0.0.5/16') => '127.0/16'. IMO, there is no\n> reasonable way to construct an inet value from a cidr value.\n\nI'm not sure I understand why this is necessary. I can see not allowing\ncidr ==> inet conversions but inet ==> cidr can be done as it is a matter\nof dropping information - the host part.\n\n\n> * Operators\n> \n> Because the types are equivalent, the operators have also been bunched\n> together in confusing ways. I propose that ordering operators (>, +, <)\n> between inet and cidr be eliminated, they do not make sense. The only\n> useful operation between cidr and inet is the << (\"contains\") operator.\n> Ordering withing cidr and inet be defined in terms of their bit\n> representation, as is the case now. The << family of operators should also\n> be removed for the inet type -- a host cannot \"contain\" another host. What\n> you probably wanted is `inet1 << network(inet2)'.\n\nThen let's define that as the meaning of \"inet1 << inet2\" i.e. define\nthe << operator between inet types as meaning \"tell me if inet1 is in\nthe same network as inet2.\" In fact, if we define << as only allowed\nbetween inet and cidr (or cidr and cidr?) then the implied cast will\ndeal with it if that cast causes the host bits to drop as suggested\nabove.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n", "msg_date": "Tue, 4 Jul 2000 22:13:06 -0400 (EDT)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: Repair plan for inet and cidr types" }, { "msg_contents": "Thus spake Peter Eisentraut\n> network and a host are not the same thing. To construct a cidr value from\n> an inet value, you'd have to use some sort of (to be created) network()\n> function, e.g., network('127.0.0.5/16') => '127.0/16'. IMO, there is no\n\nOh, I forgot to mention:\n\ndarcy=> select network('127.1.2.3/24'::inet);\nnetwork \n----------\n127.1.2/24\n(1 row)\n\nThere is also a host and netmask function and note:\n\ndarcy=> select host('127.1.2.3/24'::cidr);\nERROR: CIDR type has no host part\n\nBut I still see no reason why that can't be implicit if we assign the\n\"'127.1.2.3/24'::inet\" value to a cidr. In other words let \"select\n('127.1.2.3/24'::inet)::cidr\" give the same output.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n", "msg_date": "Tue, 4 Jul 2000 22:19:49 -0400 (EDT)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: Repair plan for inet and cidr types" }, { "msg_contents": "Peter Eisentraut <[email protected]> writes:\n> There's apparently a lack of understanding of what exactly are these types\n> are supposed to do. Therefore, instead of addressing each bug\n> individually, let me first state what I reconstructed as the specification\n> of these types, and then add what is currently wrong with it.\n\nThis sounds good offhand, but then I never paid a whole lot of attention\nto the details originally. Did you go through the original inet/cidr\ndesign discussions (the threads where Paul Vixie was participating)?\n\nI don't believe Paul is subscribed here anymore, but I'd feel a lot\nhappier if you can contact him and get him to sign off on the clarified\ndesign. Maybe this is what he had in mind all along, or maybe not.\n\n\t\t\tregards, tom lane\n\nPS: You do know who Paul Vixie is, I assume ;-). I can think of few\nbetter-qualified experts in this domain...\n", "msg_date": "Tue, 04 Jul 2000 22:31:25 -0400", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Repair plan for inet and cidr types " }, { "msg_contents": "\"D'Arcy J.M. Cain\" wrote:\n\n> > This has also been a source of problems. I propose that cidr and inet are\n> > not made equivalent types at any level. No automatic casting either. A\n> > network and a host are not the same thing. To construct a cidr value from\n> > an inet value, you'd have to use some sort of (to be created) network()\n> > function, e.g., network('127.0.0.5/16') => '127.0/16'. IMO, there is no\n> > reasonable way to construct an inet value from a cidr value.\n> \n> I'm not sure I understand why this is necessary. I can see not allowing\n> cidr ==> inet conversions but inet ==> cidr can be done as it is a matter\n> of dropping information - the host part.\n\nGiven that only contains is a reasonable cidr vs. inet comparison, we\nshould not cast automatically - both the network >> network and network\n>> host comparison are legitimate. \n\n> > * Operators\n> >\n> > Because the types are equivalent, the operators have also been bunched\n> > together in confusing ways. I propose that ordering operators (>, +, <)\n> > between inet and cidr be eliminated, they do not make sense. The only\n> > useful operation between cidr and inet is the << (\"contains\") operator.\n> > Ordering withing cidr and inet be defined in terms of their bit\n> > representation, as is the case now. The << family of operators should also\n> > be removed for the inet type -- a host cannot \"contain\" another host. What\n> > you probably wanted is `inet1 << network(inet2)'.\n> \n> Then let's define that as the meaning of \"inet1 << inet2\" i.e. define\n> the << operator between inet types as meaning \"tell me if inet1 is in\n> the same network as inet2.\" \n\nIf you do not mean \"network(inet1) <<= network(inet2)\" but rather\n\"host(inet1) << network(inet2)\", yes.\n\n> In fact, if we define << as only allowed\n> between inet and cidr (or cidr and cidr?) then the implied cast will\n> deal with it if that cast causes the host bits to drop as suggested\n> above.\n\nRight. The containing side must be a network, so we can implicitly cast\nthat. The contained side has to be taken as-is - if a network vs.\nnetwork comparison is intended, the user must explicitly cast it. \n\nSevo\n\n-- \nSevo Stille\[email protected]\n", "msg_date": "Wed, 05 Jul 2000 12:38:25 +0200", "msg_from": "Sevo Stille <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Repair plan for inet and cidr types" }, { "msg_contents": "On Tue, 4 Jul 2000, D'Arcy J.M. Cain wrote:\n\n> I'm not sure I understand why this is necessary. I can see not allowing\n> cidr ==> inet conversions but inet ==> cidr can be done as it is a matter\n> of dropping information - the host part.\n\nAutomatic casts should not lose information. How would you feel if floats\nwere automatically rounded when you store them into int fields? I think\nthis is an important principle in any type system.\n\n> Then let's define that as the meaning of \"inet1 << inet2\" i.e. define\n> the << operator between inet types as meaning \"tell me if inet1 is in\n> the same network as inet2.\"\n\nAgain, let the user say what he wants: inet1 << network(inet2).\n\nAlso note that \"is inet1 in the same network as inet2\" is different from\n\"is inet1 contained in the network of inet2\" (which is what it does now).\nThe operator you defined is symmetric (if inet1 is in the same network as\ninet2, then inet2 is also in the same network as inet1), whereas the << is\nantisymmetric. In fact, AFAICT, the operator you defined doesn't exist\nyet, although it perhaps should.\n\n\n-- \nPeter Eisentraut Sernanders vaeg 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Wed, 5 Jul 2000 09:02:17 -0400 (EDT)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: Repair plan for inet and cidr types" }, { "msg_contents": "Thus spake [email protected]\n> On Tue, 4 Jul 2000, D'Arcy J.M. Cain wrote:\n> > I'm not sure I understand why this is necessary. I can see not allowing\n> > cidr ==> inet conversions but inet ==> cidr can be done as it is a matter\n> > of dropping information - the host part.\n> \n> Automatic casts should not lose information. How would you feel if floats\n> were automatically rounded when you store them into int fields? I think\n> this is an important principle in any type system.\n\nIf it was defined well I would have no problem with it.\n\n> > Then let's define that as the meaning of \"inet1 << inet2\" i.e. define\n> > the << operator between inet types as meaning \"tell me if inet1 is in\n> > the same network as inet2.\"\n> \n> Again, let the user say what he wants: inet1 << network(inet2).\n\nI think that's what I meant. I'm just saying that inet::cidr should be\nthe same as network(inet). Allowing that cast makes a lot of operations\nwork intuitively.\n\n> Also note that \"is inet1 in the same network as inet2\" is different from\n> \"is inet1 contained in the network of inet2\" (which is what it does now).\n\nHmm. It is a subtle difference and I did miss it.\n\n> The operator you defined is symmetric (if inet1 is in the same network as\n> inet2, then inet2 is also in the same network as inet1), whereas the << is\n> antisymmetric. In fact, AFAICT, the operator you defined doesn't exist\n> yet, although it perhaps should.\n\nI guess what I was really getting at was this.\n\n host OP cidr\n\nwhere inet would cast to host on one side and cidr on the other. What\nwe have now is \n\n cidr OP cidr\n\nwith both sides casting to cidr. Of course there is no such thing as a host\ntype so I don't know how we would cast such a thing.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.\n", "msg_date": "Wed, 5 Jul 2000 16:46:48 -0400 (EDT)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: Repair plan for inet and cidr types" }, { "msg_contents": "D'Arcy J.M. Cain writes:\n\n> > Automatic casts should not lose information. How would you feel if floats\n> > were automatically rounded when you store them into int fields? I think\n> > this is an important principle in any type system.\n> \n> If it was defined well I would have no problem with it.\n\nThat is certainly not how type systems operate anywhere.\n\n> I guess what I was really getting at was this.\n> \n> host OP cidr\n> \n> where inet would cast to host on one side and cidr on the other. What\n> we have now is \n> \n> cidr OP cidr\n> \n> with both sides casting to cidr. Of course there is no such thing as a host\n> type so I don't know how we would cast such a thing.\n\nI think that while the implicit casting could sometimes be convenient,\nit's also a source of confusion. Consider the statement\n\nselect '10.0.0.3'::cidr < '10.0.0.2'::inet;\t=> f\n\nThis cannot possibly make sense on closer inspection. Firstly, it's\nsemantic nonsense, you cannot order a network and a host. Secondly, it's\nalso wrong. According to the documentation, the '10.0.0.3'::cidr should be\nconverted to '10/8' internally. Then one of two things could have happened\nhere: 1) cidr was implicitly converted to inet and '10.0.0.3' is taken to\nbe a host, which is completely wrong. Or 2) inet was converted to cidr.\nBut then we're looking at '10/8' < '10.0.0.2/32', which should be true.\n\nSee also\n\nselect '10.0.0.2'::cidr = '10.0.0.2'::inet;\t=> t\n\nwhich is wrong for similar reasons.\n\n\nThen let's look at the << family of operators.\n\nselect '10.0.0.2'::cidr >> '10.0.0.2'::inet;\t=> f\n\nAgain, there are two ways this could currently be resolved:\n\n\t'10/8'::cidr >> '10.0.0.2/32'::cidr\twhich does return true\nor\n\t'10.0.0.2'::inet >> '10.0.0.2'::inet\nwhich doesn't make any sense.\n\nOn closer inspection, the inet << cidr case is completely misbehaving:\n\nselect '10.0.0.5/8'::inet << '10.0.0.0/16'::cidr;\t=> f\nselect '10.0.0.5/24'::inet << '10.0.0.0/16'::cidr;\t=> t\n\nThis is not what I'd expect.\n\nConcretely, the cases\n\tinet << cidr\n\tcidr << cidr\nare not the same:\n\n\t'10.0.0.5/8'::inet << '10.0.0.0/16'::cidr\nshould be true\n\n\t'10.0.0.5/8'::cidr << '10.0.0.0/16'::cidr\nshould be false, if you allow the left-side value in at all, which I\nwouldn't.\n\nWhat this tells me is that the cast from inet to cidr is not well-defined\nin the mathematical sense, and therefore no implicit casting should be\nallowed.\n\nSo the bottom line here is that these two types are, while from a related\ndomain, different, and the user should be the one that controls when and\nhow they are mixed together.\n\n\n-- \nPeter Eisentraut Sernanders v�g 10:115\[email protected] 75262 Uppsala\nhttp://yi.org/peter-e/ Sweden\n\n", "msg_date": "Fri, 7 Jul 2000 18:16:26 +0200 (CEST)", "msg_from": "Peter Eisentraut <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Repair plan for inet and cidr types" } ]