Re: [pgsql-hackers] Patent issues and 8.1

Robert,

> Read the law... willful vs. unknown infringement are two different
> things.

We're not infringing anything, yet. That's a *pending* patent.

> Um... thats the way our legal system works. You could do that to any
> project if you had a patent they were infringing upon no matter how
> stoic they tried to be about it. (By our I mean the U.S. system)

You're not following me. Imagine this:
1) Pervasive/Fujistsu/SRA/Mammoth PostgreSQL steals some big clients from
Obsolete Proprietary Database Company (OPDC).
2) OPDC has someone dig through their piles of patents and finds something
that looks like it might infringe on something PostgreSQL does.
3) OPDC gets a blogger or similar to post something "And in the latest patent
infringment news ..."
4) -Hackers hears about it and we derail development for another 3 months in
order to work around the patent.
Net Cost to OPDC: couple $thousand, to delay a PG release by 3+ months.

What's kept patent litigation from being used against OSS projects so far is
the bad PR that would attach, the potential cost of litigation, the
possibility of having the patent invalidated, and the dubvious prospect of
compensation. But if a competitor can disrupt an OSS project with a
*threatened* patent, then the cost is minimal and the effect is huge.

We will face this situation again -- at least, until software patents go away
-- and both I and Bruce feel that it's important to set a precedent in
dealing with them because you can bet this discussion is being read by people
who are not in favor of the spread of PostgreSQL. This isn't just about
the ARC patent, it's about the next one after that.

> FWIW I've really only been advocating

BTW, my last post wasn't specifically addressed at you, but at the viewpoint
that we should drop everything and work on the ARC replacement to get it out
the door in 4 months.

> that we don't do the change in a
> patch branch, which I'm afraid the "do nothing till the lawyers show up"
> plan would eventually lead to. We wouldn't normally do things that way
> on technical grounds, so I'd prefer not to be forced into doing things
> that way for other reasons; enough so that I think we ought to have a
> plan to address it now.

It's not a choice between doing something and doing nothing; you're
mischaracterizing. It's a choice between:

1) Shall we begin development immediately on an 8.1 which does not include the
ARC code and can be upgraded without initdb, for plans to release this
version in 4 months or less?

2) Shall we work our regular 1-year development cycle, with plans to replace
ARC with an improved memory management approach as part of the features of
8.1, keeping a reversion-to-LRU patch in reserve in case we have to release
it as a patch in the 8.0.x series?

I advocate (2), partly because I don't believe that (1) is really possible for
us. When's the last time we did a fast release? What I do advocate doing
*now* is:

a) someone (Simon? Sean? Neil? Jan?) should start hacking on a
better-than-ARC buffer manager to have it for 8.1, and

b) we should build an 8.0.1 with Neil's Revert-to-LRU patch, upload it to
OSDL, and start hammering on it so that it will be tested in case we need it
(and if there's no loss of performance or stability, maybe drop it in the
update stream regardless of patent status).

I also suggest that we might want to hit up one of the several well-funded
parties involved in PostgreSQL, who have staff attorneys, for an opinion on
the whole business, and some insight into what proprietary software companies
do. That would be Pervasive, Fujitsu (assuming that AU has software patents,
I don't know), OSDL, and CMD. Heck, there will be a panel of OSS attorneys
at the Enterprise Linux Summit; I can ask them but of course it won't be an
actual opinion unless money changes hands.

--
Josh Berkus
Aglio Database Solutions
San Francisco