Compatibility of the AFL with the GPL

[Brian Behlendorf]

Sorry for copying large segments; I have a feeling we're talking past each
other, and I want to try to avoid that.

On Wed, 12 Mar 2003, Lawrence E. Rosen wrote:
> First, as to the Mutual Defense provision and its "compatibility" with
> the GPL:
>
> Person A writes W and licenses it to everyone under the AFL.  Person B
> comes along and, in the true spirit of free software, creates and
> distributes collective work W+X and derivative work W' under the GPL.
> No surprises for B.  He's read the AFL and the GPL and he understands
> that he's doing what's allowed.
>
> Person C gets a copy of W+X or W'.  He knows it is GPL software.  Person
> C now wants to sue Person A for patent infringement by W.  He reads W's
> license and discovers the Mutual Defense provision.  He must evaluate
> his risk of losing rights to copy, modify or distribute W, W+X and W',
> and any other W-based software, if he sues A.  What's wrong with making
> him evaluate that risk before suing for patent infringement?  It's his
> patent, and he's (perhaps) within his rights to sue Person A for
> infringement.  But as the author of an open source license, I don't have
> to make that easy or cheap for him to do.

Nothing is wrong in spirit with the Mutual Defense provision.  That's not
the question here.

When person C gets that copy of W+X or W' from B, they were told that it
is covered by the GPL.  C reads the GPL, understands it, and is led to
believe (barring any situation that B couldn't have prevented, such as a
third-party patent claim on code that W+X or W' implements) that the GPL
completely describes the terms and conditions that C must follow.

When someone downloads Apache software, they read the Apache license, and
are led to believe that's the only license they need to be observant of in
order to exercise the rights the license grants them.  Apache developers
can't guarantee there'll never be patent claims made by third parties of
course, and we indemnify our liability in that event.  But the Apache
developers are very careful to make sure there's no additional clauses in
third-party packages we bundle into tarballs people pull from apache.org.
No one likes surprises.

I think this describes the common interpretation of what it means to say a
particular piece of software has a particular license.  That license
should include all clauses relevant to the exercise of the rights it
claims to give.  If there are additional licenses one must also follow, it
should be included by reference.

John used the example of the Apache license.  What allows a vendor to pick
Apache up and incorporate it into a proprietary work is that we do not
require derived works to also be licensed under the Apache license.  The
GPL, though, *does* require derived works to be licensed under the GPL.

> Perhaps the LICENSE file of any GPL-licensed work that contains an
> AFL-licensed component should contain a warning notice:
>
>    WARNING: Your license to this work may automatically terminate
>    if you sue Person A for patent infringement.  That is because
>    this work contains a component that is licensed to you by Person A
>    under the AFL.  You get to decide whether a patent infringement
>    lawsuit against Person A is worth the loss of your rights to
>    copy, modify or distribute Person A's contribution to this software.

So in what way does this language not constitute an "additional
restriction" as defined by section 6 of the GPL?

Every time someone pops up on license-discuss and says "I have a new
license, it's just like the GPL but it adds a clause that says you can't
stand on your head and count to ten.  Is it OSI-conformant?
GPL-compatible?"  We may hem and haw over whether being able to stand on
your head and count to ten constitutes a discrimination based on field of
endeavor, but it seems that people always say, "adding clauses, anything,
makes it not GPL compatible."

> You're right in suggesting that the GPL has fostered a spirit of license
> trust, and that is wonderful.  I'm seeking compatibility between the AFL
> and the GPL because I want to share in that good will and to encourage
> people to release source code that can be incorporated into GPL-licensed
> programs -- as well as into proprietary and other open source programs.

We share the same objective - I'd love to see the Apache license get there
too.  But we hit the same brick wall, and we can't just pretend it doesn't
exist.

> The mutual defense provision of the AFL doesn't detract from that goal.
> It just causes those who would sue free and open source software for
> patent infringment to do their homework first and to recognize that it
> is no longer cheap and risk-free to do so.

Again, no one is criticizing the intent, though I've brought up before
that I'm not sure patent lawsuits are always about evil patent holders
vs. the small guy hero.  That's besides the point.

> They only have to "crawl through the source code" if they elect to sue
> OSI Certified open source software for patent infringement.  What's
> wrong with that?

They have to crawl through the source code of an application that they
otherwise were led to believe would have no bearing on their lawsuit.
That's what's wrong.  It seems to be legitimizing the concept of hidden
clauses so long as it works to our advantage.  I don't like that at all.

> ***************
>
> Second, as to the trademark provision of the AFL:
>
> I recognize that there is inevitable confusion when one adds a new
> concept to the world of the GPL.  But I've been following open source
> businesses for a while -- not as long as you, I admit! -- and I now
> recognize that the real source of success and profit for open source
> contributors and businesses is their trademarks, not just their
> copyrights.  Contributors to free software create brand names and
> personal reputations [Linux (and GNU/Linux!), Apache, Red Hat, JBoss,
> OSI Certified, Free Software Foundation, RMS, Behlendorf] that we need
> to protect.

I wouldn't call it "the real source of success and profit".  I would call
it "one of the few real exclusive assets the authors of a project build".
In an electronic world, your name and reputation is *everything*, and
unlike source code, reputation *is* rivalrous.  Another company misusing
the Apache or Linux trademark *does* hurt the ASF or Linus.  And let me
say it for the Nth time - I don't think the intent to protect trademarks
is bad at all.  I *love* the fact that Apache has been able to protect
theirs using the club of copyright law rather than trademark law.

I genuinely, sincerely hope RMS opens up something along these lines in
GPL v3.

> That's not contradictory in any way to the GPL.  You found such a
> trademark provision was helpful to the Apache Foundation, and quite
> frankly, I knew to write a trademark provision because of your lead.  So
> please don't try to talk me out of doing what you already did so
> successfully.

I'm not trying to talk you out of it, certainly.  Let's advocate that it
be allowed in GPLv3.  But let's not pretend it's allowed in GPLv2,
especially if Stallman explicitly says it's not.

	Brian

--
license-discuss archive is at http://crynwr.com/cgi-bin/ezmlm-cgi?3