Re: GPL'ed Patent Pool

On Sunday 12 September 2004 04:09, Richard Stallman wrote:

> A few patent holders have signed such licenses.  This isn't really a
> pool, in the usual sense, because each patent is individually licensed
> to GPL-covered software.  If 10,000 patents had already been licensed
> in this way, that would not put any additional pressure on patent
> holder number 10,001 to participate.

They've signed variants of such licenses, but there is no standard license 
text to attach a patent to (L)GPL software, in a way that the patent license 
applies only to (L)GPL software (and not BSD or proprietary software) and 
terminates if the (L)GPL terminates.

> Therefore, even if everyone that likes free software licenses patents
> in this way, their having done so would not pressure anyone else to do
> so.  I don't see how this idea could lead, even if it is very
> successful, to putting any pressure on Microsoft (for instance) to
> join the pool.

Okay, three things here.

First, right now everybody using GPL software is infringing upon more patents 
than they can even enumerate, and everybody hopes nobody's stupid enough to 
sue.  Imagine if the Canopy Group had bought SGI, and hadn't been stupid 
enough to start with an attack on IBM first thing, but instead went straight 
for shaking down individual users.  Caldera used to be a friendly member of 
UnitedLinux, but then they had a management change.  This has nothing to do 
with Microsoft.

Right now we can't even publicly list what the patents ARE for fear of 
painting a target on developers' backs and tripling potential damages, but if 
somebody like OSDL could lobby just the friendly ones to clean up the mess a 
bit, list the patents we could defend ourselves with while protect us from 
future management changes that would turn those patents against us, that by 
itself would be worth doing.  This could address the white elephant nobody 
can talk about just now.

Secondly, there's clarification.  IBM promised not to sue Linux over patents, 
which falls under estoppel but is not a license.  Red Hat tried to make this 
kind of license already, but without making it a group effort.  The RTLinux 
guys have a similar but not quite the same license on their patent that's an 
attempt to defend themselves against non-GPL appropriation of it while not 
suing anybody who uses it under the GPL.  (They're just horrible at 
explaining it, you have to drill into mailing list archives to find out what 
they're talking about.)  Getting an actual license to these things, attached 
to the GPL, would be a lot clearer than the current ad-hoc mess.  Get in 
writing what has been granted and what hasn't been granted. 

So two good reasons to do it are to move beyond the gentlemen's agreement to 
something binding on the parties involved, and clarifying everything in 
writing.  Those are easy to accomplish.

A third goal, which is harder, is defending against a big evil company like 
Microsoft (or potentially Sun, or Oracle with a Canopy style management 
change).  And this group of patents would give us the same thing any other 
group of patents does: something to countersue with.

Now admittedly most of the big companies have already cross-licensed 
everything, in a way that leaves the little guys vulnerable.  IBM's patents 
wouldn't help anybody against Microsoft the way they can beat down SCO, and 
if SCO had directly sued Red Hat or end users, it's not clear IBM would have 
stepped in pro bono to defend them.

This is why it's important to get small companies like SGI and Red Hat to step 
in with their contributions, contributions that are not already licensed to 
the big companies, to prevent somebody like Microsoft from doing a divide and 

Of course Microsoft wouldn't put their patents in.  You can't force anybody to 
participate in a patent pool any more than you can force them to participate 
in a copyright pool.  (We can't make Microsoft release software under the 
GPL, either.)  The point is to make them think twice about suing software 
covered by the pool.

There are three aspects of a patent pool that make them nice to join.  One is 
knowing that you can't be sued over patents in the pool, and if this was to 
be a free patent pool, that one must be available to people who haven't put 
patents in the pool.  So I agree, that's not a reason to join an existing 
pool, but we have social reasons to replace that one.

Second is termination.  If you sue somebody in the pool, your rights to use 
patents in the pool terminate.  The GPL already has a termination clause, and 
all we'd want the patent license to do is tie into that.  Demanding a 
non-free patent license as a condition of using GPL software is an 
"additional restriction" that should terminate the suer's rights under the 
GPL.  All the patent license should do is reinforce that, and a widely used 
patent license that explicitly interpreted things that way might help sway a 
judge that this was the correct interpretation of matters.

Third is that I believe patent pool members can enforce other people's patents 
(defensively weilding the whole pool).  I don't know the details of the legal 
mechanism here, I guess that's why there's usually an enforcing consortia.  
If a free patent pool made it so that contributing a patent to the pool let 
you countersue with the entire pool (but only if you were sued first), that 
would be a heck of a reason to scrape up at least one patent of your own to 
put in the pool.  (And then in for a penny, in for a pound...)  This is 
something I'd want to ask a lawyer about...

The best defense against Microsoft using patents is if there were a pool of 
patents that Microsoft had no other access to, and which made it more likely 
that if they sued over these patents' use in GPL code, a judge might make 
them stop using GPL code internally.  The goal isn't to absolutely stop them, 
it's to make them reluctant to sue, and to provide an anti-FUD bulwark we can 
point to "if they do this, we have a ready response".  Not a perfect 
response, but a coherent and believable story about what might happen next 
that would be bad for them and good for us, that we can tell the press and 
nervous users pondering whether to jump on board.

> Most ideas for a patent pool try to arrange for the patents in the
> pool to pressure other patent holders to join the pool.  That requires
> licensing the patents in the pool to whoever joins.  But it is not
> easy to work out the details.  People have been looking at the idea of
> a software patent pool for about 13 years now, but no one has ever
> been able to get one started.

IBM's statement that they won't sue Linux provides estoppel against, 
basically, their entire patent pool.  They have nothing to lose by granting a 
license.  Red Hat already tried this, they just didn't explain it clearly or 
explicitly try to make a unified group effort out of it.  If those two made a 
joint effort, Novell/SuSE wouldn't want to be left out.  And at that point, 
there would probably be enough momentum to get SGI to jump on board.

It's not perfect.  But it's better than what we've got now.

> The problems are very hard.  There are
> many different ways to organize such a pool, but no one has come up
> with a way that does the whole job.

The perfect is the enemy of the good.

> Also, it is not clear what sort
> of patent pool you might be able to find the first members for.
> I'd suggest first to look for a company that has a lot of patents and
> wants to join a software patent pool.  Then we could talk with that
> company about what kind of pool to set up.

We're working on it.

I'm swamped until October 10th due to Linucon, but Eric and Cathy Raymond will 
be visiting Austin due to that, and Linas lives in Austin too.  I'll try to 
get everybody together and hammer out a proposed license text while they're 
in town, and then maybe bother the IBM Linux Technology center with it.  Alan 
already gave me the email of a Red Hat guy, I'm just not going to approach 
him until I have something specific...

-- Linux Expo and Science Fiction Convention
October 8-10, 2004 in Austin Texas.  (I'm the con chair.)

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]