From @tp @end|ng |rom p|@kor@k|@com Tue Feb 11 12:43:17 2003 From: @tp @end|ng |rom p|@kor@k|@com (Andrew Piskorski) Date: Tue, 11 Feb 2003 06:43:17 -0500 Subject: [R-sig-DB] license issues R vs. PostgreSQL Message-ID: <20030211114317.GA98963@piskorski.com> Hello, Joe Conway recently announced on the PostgreSQL hackers list that he will soon be releasing PL/R, R embedded as a procedural language inside Postgres: http://archives.postgresql.org/pgsql-hackers/2003-02/msg00142.php For code maintenance and enhancement reasons, it seems that both Joe and the Postgres developers would like to have PL/R in the Postgres main source tree. However, the Postgres people are insist on a BSD license for anything in that repository. (I'm not sure why exactly.) Now, Joe apparently is happy to license his code under BSD, but he used libR, which I think is only licensed under GPL, *not* LGPL. Which means that Joe can't use the BSD license, right? Basically, resolving this license issue is the only thing preventing Joe from releasing PL/R, something I and many Postgres users would definitely like to see! So, do you have any suggestions for possible solutions, or whom Joe should talk to? May Joe use the BSD license on his PL/R code as things currently stand with R? What license changes would need to happen in R for Joe to license PL/R under BSD? Would the R project be willing to make such changes? I also posted another version of this same question with some additional information here: http://openacs.org/forums/message-view?message_id=78404 -- Andrew Piskorski http://www.piskorski.com From ripiey m@iii@g oii st@ts@ox@@c@uk Tue Feb 11 13:50:05 2003 From: ripiey m@iii@g oii st@ts@ox@@c@uk (ripiey m@iii@g oii st@ts@ox@@c@uk) Date: Tue, 11 Feb 2003 12:50:05 +0000 (GMT) Subject: [R-sig-DB] license issues R vs. PostgreSQL In-Reply-To: <20030211114317.GA98963@piskorski.com> Message-ID: This is not really a matter for r-sig-db, and neither this group nor R-core are going to give advice on licensing issues (we are not lawyers, for a start). You could consult FSF for the meaning of GPL, but I am pretty sure that the contamination principle would apply. libR is GPL-ed, and that is unlikely to change. Since linking against libreadline is considered to need GPL, I expect the same arguments to be applied to libR. Already R has interfaces to (and has been embedded into) several DBMSs, included being previously embedded into PostgreSQL (by Duncan Temple Lang, with sources formerly at least on www.omegahat.org, and still accessible on CRAN/Omegahat, so it may well be still on www.omegahat.org). I am afraid I don't see what makes PL/R (which we have not seen) a special case. On Tue, 11 Feb 2003, Andrew Piskorski wrote: > Hello, Joe Conway recently announced on the > PostgreSQL hackers list that he will soon be releasing PL/R, R > embedded as a procedural language inside Postgres: > > http://archives.postgresql.org/pgsql-hackers/2003-02/msg00142.php > > For code maintenance and enhancement reasons, it seems that both Joe > and the Postgres developers would like to have PL/R in the Postgres > main source tree. However, the Postgres people are insist on a BSD > license for anything in that repository. (I'm not sure why exactly.) > Now, Joe apparently is happy to license his code under BSD, but he > used libR, which I think is only licensed under GPL, *not* LGPL. > Which means that Joe can't use the BSD license, right? > > Basically, resolving this license issue is the only thing preventing > Joe from releasing PL/R, something I and many Postgres users would > definitely like to see! > > So, do you have any suggestions for possible solutions, or whom Joe > should talk to? > > May Joe use the BSD license on his PL/R code as things currently stand > with R? What license changes would need to happen in R for Joe to > license PL/R under BSD? Would the R project be willing to make such > changes? How can we answer the last unless you tell us the answer to the second? -- Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595 From m@|| @end|ng |rom joeconw@y@com Tue Feb 11 18:05:56 2003 From: m@|| @end|ng |rom joeconw@y@com (Joe Conway) Date: Tue, 11 Feb 2003 09:05:56 -0800 Subject: [R-sig-DB] license issues R vs. PostgreSQL In-Reply-To: References: Message-ID: <3E492D74.2070500@joeconway.com> ripley at stats.ox.ac.uk wrote: > libR is GPL-ed, and that is unlikely to change. Thank you. That explicit confirmation was what I've been looking for. > Already R has interfaces to (and has been embedded into) several DBMSs, > included being previously embedded into PostgreSQL (by Duncan Temple Lang, > with sources formerly at least on www.omegahat.org, and still accessible > on CRAN/Omegahat, so it may well be still on www.omegahat.org). I am > afraid I don't see what makes PL/R (which we have not seen) a special > case. You haven't seen code yet because I didn't want to make a public posting with the wrong license. PL/R isn't special compared to REmbeddedPostgres, but they both are quite different than, say, RMySQL. The standard DBMS interfaces are from R to the database using the databases' client libraries. Those libraries are relatively stable interfaces to their respective databases. On the other hand, here is Duncan's note from the top of the REmbeddedPostgres INSTALL file: ********************************************************************** Note: Before compiling this package, you will _have to_ have the source for Postgres. The regular header files that are installed when Postgres is installed are not sufficient since this package does not use the client interface, but the internal facilities of the Postgres server. ********************************************************************** The same is true of PL/R. In this case the interface is from Postgres to R using R's standard interface library, which I presume is intended to be reasonably stable. Because of this issue, I'd be amazed if REmbeddedPostgres would even compile against current Postgres sources, because there have been huge changes since PostgreSQL 7.0.2 that Duncan wrote to. PL/R (and REmbeddedPostgres) could be said to be loosely coupled to R, and tightly coupled to Postgres. The advantage of having such in the Postgres source tree is significant -- as the backend code evolves, PL/R comes along for the ride. But I guess all of the above is moot, because I'll need to release PL/R under GPL, meaning it will not be accepted into the PostgreSQL source tree (I have already asked). So I'll change the license blocks in my code from BSD to GPL and get on with it. I'd very much like feedback from some R developers before a general release -- would anyone be willing to give it a look if I post a link? (after changing the license to GPL of course ;-)) Thanks for your time, Joe From ro@@|n| @end|ng |rom b||ndg|obe@net Tue Feb 11 18:28:50 2003 From: ro@@|n| @end|ng |rom b||ndg|obe@net (A.J. Rossini) Date: Tue, 11 Feb 2003 09:28:50 -0800 Subject: [R-sig-DB] license issues R vs. PostgreSQL In-Reply-To: <3E492D74.2070500@joeconway.com> (Joe Conway's message of "Tue, 11 Feb 2003 09:05:56 -0800") References: <3E492D74.2070500@joeconway.com> Message-ID: <87of5iohf1.fsf@jeeves.blindglobe.net> Joe Conway writes: > I'd very much like feedback from some R developers before a general > release -- would anyone be willing to give it a look if I post a link? > (after changing the license to GPL of course ;-)) Yes. best, -tony (I'm on vacation until the 17th -- replies will be terse). -- A.J. Rossini Rsrch. Asst. Prof. of Biostatistics U. of Washington Biostatistics rossini at u.washington.edu FHCRC/SCHARP/HIV Vaccine Trials Net rossini at scharp.org -------------- http://software.biostat.washington.edu/ ---------------- FHCRC: M: 206-667-7025 (fax=4812)|Voicemail is pretty sketchy/use Email UW: Th: 206-543-1044 (fax=3286)|Change last 4 digits of phone to FAX (my tuesday/wednesday/friday locations are completely unpredictable.) From m@|| @end|ng |rom joeconw@y@com Tue Feb 11 20:21:36 2003 From: m@|| @end|ng |rom joeconw@y@com (Joe Conway) Date: Tue, 11 Feb 2003 11:21:36 -0800 Subject: [R-sig-DB] PL/R alpha posted (was: license issues R vs. PostgreSQL) In-Reply-To: <87of5iohf1.fsf@jeeves.blindglobe.net> References: <3E492D74.2070500@joeconway.com> <87of5iohf1.fsf@jeeves.blindglobe.net> Message-ID: <3E494D40.9010407@joeconway.com> A.J. Rossini wrote: > Joe Conway writes: >>I'd very much like feedback from some R developers before a general >>release -- would anyone be willing to give it a look if I post a link? >>(after changing the license to GPL of course ;-)) > > Yes. > Thanks! OK - license changed to GPL, and source is available here: http://www.joeconway.com/plr/plr.0.1.1.alpha.tar.gz The documentation, including preprocessed html, is in the tar ball. I've also posted the html docs here: http://www.joeconway.com/plr/index.html Installation: Place tar file in 'contrib' in the PostgreSQL source tree and untar. Then run: make make install You can use plr.sql to create the functions in your database of choice, e.g. psql mydatabase < plr.sql plr.sql is installed to /usr/local/pgsql/share/contrib on my system, but YMMV. This all generally works best if you're logged in as user postgres, and postgres owns the source tree. You may need to su to root in order to do `make install` Also note: make installcheck will create a datbase called regression, and run the regression test. In addition to the documentation, the plr.out file in plr/expected is a good source of usage examples. PL/R should build cleanly with PostgreSQL 7.3.x and cvs HEAD. It was developed using libR from R 1.6.2 under Red Hat 7.3 & 8.0 -- I've not tested against other versions of R or different OSes. Please let me know how it goes. Thanks, Joe From dunc@n @end|ng |rom re@e@rch@be||-|@b@@com Tue Feb 11 20:47:40 2003 From: dunc@n @end|ng |rom re@e@rch@be||-|@b@@com (Duncan Temple Lang) Date: Tue, 11 Feb 2003 14:47:40 -0500 Subject: [R-sig-DB] license issues R vs. PostgreSQL In-Reply-To: <3E492D74.2070500@joeconway.com>; from mail@joeconway.com on Tue, Feb 11, 2003 at 09:05:56AM -0800 References: <3E492D74.2070500@joeconway.com> Message-ID: <20030211144740.E5051@jessie.research.bell-labs.com> Joe Conway wrote: > > On the other hand, here is Duncan's note from the top of the > REmbeddedPostgres INSTALL file: > > ********************************************************************** > Note: > > Before compiling this package, you will _have to_ have the source for > Postgres. The regular header files that are installed when Postgres > is installed are not sufficient since this package does not use the > client interface, but the internal facilities of the Postgres server. > > ********************************************************************** > > The same is true of PL/R. This is something that should be addresed to the Postgres developers. If they want any procedural languages to exist outside of the Postgres source tree, they are going to have make the relevant header files available in their development libraries. It would be far more advantageous to urge them to install the necessary header files rather than dealing with each author. -- _______________________________________________________________ Duncan Temple Lang duncan at research.bell-labs.com Bell Labs, Lucent Technologies office: (908)582-3217 700 Mountain Avenue, Room 2C-259 fax: (908)582-3340 Murray Hill, NJ 07974-2070 http://cm.bell-labs.com/stat/duncan From jgr@m||ch @end|ng |rom p|ocon@com Tue Feb 11 21:50:31 2003 From: jgr@m||ch @end|ng |rom p|ocon@com (Joshua Gramlich) Date: 11 Feb 2003 14:50:31 -0600 Subject: [R-sig-DB] postgres? Message-ID: <1044996631.4197.57.camel@localhost.localdomain> Sorry folks if I'm double posting here, I was having some trouble with my smtp server...anyways... How does one connect R and postgres? I've got Rdbi, but it has no documentation as to how to use it. Can anyone help? Joshua Gramlich Chicago, IL