FSF response to the Python 2.1 license

Today, I received the following email from Eben Moglen, the Free Software Foundation's attorney. (See also a postscript.)

Subject: Re: Python 1.6.1 and GPL compatibility
From: Eben Moglen <moglen@columbia.edu>
To: Guido van Rossum <guido@digicool.com>
Cc: "Bradley M. Kuhn" <bkuhn@gnu.org>, rms@gnu.org
Date: Thu, 19 Apr 2001 07:44:11 -0400 (EDT)

On Wednesday, 18 April 2001, Guido van Rossum wrote:

  Please get me an answer before Thursday's over.  I can't hold up my
  response to the Slashdot questions much longer.

What follows is our official response to the 2.1 license.

The removal of the acceptance ceremony takes care of the primary
dissonance between GPL and the Python 1.6.1 license (but see below).
The other issue over which we spent much time in negotiation with CNRI
is the applicable law provision in paragraph 7.

The GPL doesn't use an applicable law provision.  We are concerned
about a number of problems that can arise.  You see the simplest of
them already in your license stack: one license applying the law of
California and another the law of Virginia.  That could happen many
times over with free software projects, with each module and some
patches within the modules treated under different rules of law, which
would create almost intolerable confusion.  It does not happen with
proprietary software because there is a single owner, who does not
permit derivative works.

In addition, we need to be worried about Unfreedonia, the jurisdiction
that treats the GPL as a nullity, or even worse interprets it to make
code unfree.  Then anyone can receive a GPL'd program, modify it, and
rerelease under a GPL with an Unfeedonian-law-applies clause.  We're
screwed.

So we have never allowed a license with a choice-of-law clause to be
treated as fully compatible with GPL.  Virginia is the worst of all
choices, because that state has passed the UCITA law, which adds a
whole new range of risks and burdens in the distribution of free
software.

We explained all this to CNRI, but it was the other subject on which
they were immovable.  So we settled for a clause that specifically
exempted any GPL'd code in Python from that clause, except as it
applied to the warranty disclaimers.  This was a kludgy solution, but
we wanted to make a deal, and we expected to have the code ultimately
released under a different license.

But this situation shows why we hate doing that: you've taken over
their language.  If we now agree that this is GPL-compatible, we'll
have to live with it forever, because other people will copy it from
your license.  And it still contains more risks for the future than we
think it is wise to incur.

So let's take paragraph 7 and reduce it to the no-joint-venture
clause, which we also think is unnecessary, but which is harmless.  It
would then read: 

   7. Nothing in this License Agreement shall be deemed to create any
   relationship of agency, partnership, or joint venture between PSF and
   Licensee.  This License Agreement does not grant permission to use PSF
   trademarks or trade name in a trademark sense to endorse or promote
   products or services of Licensee, or any third party.

Now about paragraph 8.  By removing the ceremony of acceptance you
eliminate the incompatibility with GPL, but you are still telling your
user that installation and use of Python constitutes acceptance of
your license.  This may or may not hold up in court, in the event that
you ever need to enforce, but we wonder why you would want to sue
someone who had installed and used Python with his fingers crossed
about your license, but who never modified or redistributed it.  The
GPL says that you don't need to accept the license, but you can't
modify or redistribute unless you do, because nothing gives you the
right to do those things except the license.  We strongly recommend
that approach: it loses you nothing except the hypothetical right to
enforce against someone you wouldn't care about anyway, and it doesn't
commit you to claiming this sort of "implicit acceptance by use" that
the lawyers don't believe will work, which is why they use acceptance
ceremonies, .....

You will notice that we're about simplifying here.  Most lawyers
operate from caution: anything that might ever possibly help they
throw into the license.  Every place they do business they have a
local license adapted to the conditions there, with every little
detail thrown in.

We're about elegance.  We try to use the absolute minimum, because
we're offering one license for use everywhere, giving the maximum
amount of freedom to every user.  So the simpler things are the more
compatible they are with our needs.

Conclusion: in order to be "compatible" in the strict sense, we need
paragraph 7 to remove the choice of law clause.  We would also
recommend that paragraph 8 be changed to say that copying,
modification or distribution constitutes acceptance of the license,
but we don't have to have that change to agree that the license is
fully "GPL compatible."

I hope this helps.  If you have any further questions I can try to
respond during the course of Thursday, but my schedule is extremely
full.

Best regards.

-- 
 Eben Moglen                       voice: 212-854-8382 
 Professor of Law & Legal History    fax: 212-854-7946       moglen@
 Columbia Law School, 435 West 116th Street, NYC 10027     columbia.edu
 General Counsel, Free Software Foundation   http://moglen.law.columbia.edu

Postscript

In accordance with the requirement above, the PSF board has decided to change the license, by replacing the original paragraph 7 with the paragraph 7 suggested in Eben Moglen's letter above. The first release with such a GPL-compatible license is Python 2.0.1; Python 2.1.1 will soon follow. (Added 14-June-2001)