LWN: Comments on "PostgreSQL 8.0.2 released with patent fix"

PostgreSQL 8.0.2 released with patent fix

alvherre — Wed, 13 Apr 2005 23:14:02 +0000

FYI, this LWN.net article and this comment have been quoted by ZDnet UK (including a link to the relevant PTO entry and Elein Mustain's article on the subject.)

PostgreSQL 8.0.2 released with patent fix

nconway — Tue, 12 Apr 2005 12:29:27 +0000

The wire protocol did not change. All that happened is that the major version number of the libpq .so has been bumped. If you don't want to recompile any applications, just keep the 8.0.1 libpq around as well.

PostgreSQL 8.0.2 released with patent fix

Wol — Tue, 12 Apr 2005 12:11:47 +0000

As Khim said.

IBM is perfectly happy with a BSD licence.

Read IBM's patent pledge, and read it *CAREFULLY*. Provided the software *stays* BSD, the pledge continues to apply. But change the copyright licence, and you can lose the patent licence.

That said, I do understand this conflicts with the BSD principle that anyone should be free to use the code in any way they choose (including going proprietary). But can you blame IBM for not wanting people to use their own patents against them?

Cheers,
Wol

PostgreSQL 8.0.2 released with patent fix

nconway — Tue, 12 Apr 2005 12:05:25 +0000

There were a few reasons why we didn't talk to IBM. We would have needed IBM to license the patent to PGDG (the open source project), as well as any commercial entity that wanted to modify and redistribute PostgreSQL. Since those commercial entities include companies like Fujitsu and Pervasive that compete with IBM, most people thought IBM wouldn't be willing to do this.

Also, there might have been more merit in talking to IBM if the patented algorithm could be shown to be significantly better than any unpatented alternative. However, all the tests we've run have indicated that 2Q has about the same performance as ARC -- if you have benchmarks that indicate otherwise, please post to the pgsql-hackers list about it. Going through a bunch of legal headaches for an algorithm that doesn't improve performance doesn't make a lot of sense. It's also worth noting that this entire subsystem has been rewritten in 8.1 for better locking performance -- because of the way the new locking scheme works, neither 2Q or ARC can be used, so the point is moot in the long run. That's why I think that a simple technical fix that doesn't hurt performance and doesn't involve lawyers makes the most sense.

Cheers,

Neil Conway

PostgreSQL 8.0.2 released with patent fix

khim — Tue, 12 Apr 2005 08:19:27 +0000

IBM is perfectly happy with BSD license. It will grant license for BSD-licensed software. Of course. "As published by OSI". Obviously.

That is: you'll get your license (with paper trail even!), but... you'll be unable to change terms of license later (and you'll be unable to make proprietary version of PostgreSQL).

FSF hacked copyright system to create copyleft. IBM hacked patent system to make non-copyleft open-source licenses copyleft as well. Pretty ingenious if you'll ask me.

Think about it: by granting big pile of patents for open source projects IBM makes "MySQL business model" possible for unrelated projects! If you are open-source developer - you can just forget about patents and hack freely: you have written license to do so. Once you'll try to make program proprietary... bam: you'll get letter with royalty demands pretty soon. Kinda defeats reason to license something under BSD license, but... Brilliant, just brilliant.

P.S. What I really do not get is why IBM waited for so long to make this move. I was sure they'll do something like this years ago, but I guess moment was wrong.

PostgreSQL 8.0.2 released with patent fix

Anonymous1 — Tue, 12 Apr 2005 04:54:45 +0000

Postgres uses a BSD License and IBM is unlikely to grant everyone free
use of its patent, so there wasn't a lot of point in asking IBM for
an exception.
This might have made sense for a GPL licensed project.

PostgreSQL 8.0.2 released with patent fix

jwb — Tue, 12 Apr 2005 04:34:46 +0000

Am I reading this incorrectly, or did they in fact break the wire protocol? If I move to 8.0.2, do I
need to recompile applications on totally unrelated machines? Or are they saying that I will need to
replace applications that run on the database server, due to the soversion increasing? if the latter,
why can't I just keep my old .so?

PostgreSQL 8.0.2 released with patent fix

dlang — Mon, 11 Apr 2005 22:46:07 +0000

frankly I think they made a mistake by not even asking IBM about the patent.

as commented on in the more detailed article it's unlikly that IBM would immediatly take legal action, especially as they were showing willingness to change it as nessasary.

so not postgres users are faced with the choice of getting a batch of fixes and loosing performance, or doing without the fixes. this is a bad position to put anyone in.