Subject: Re: termless copyright and patents
From: <stephen@xemacs.org>
Date: Tue, 3 Oct 2006 15:24:15 +0900

Thomas Lord writes:

 > There are simpler questions that we should not need to speculate
 > about.   Here is an example:
 > 
 >         A patentee, Alice, conveys a GPL program, Prog1,
 >         whose source consists of files foo.c and bar.c.
 > 
 >         A hacker, Bob, is working on Prog2.  Initially,
 >         Prog2 does not infringe on any of Alice's patents.
 > 
 >         Prog2 contains a file, zap.c, whose external interface
 >         is identical to that of bar.c.  Indeed, if Bob replaces
 >         zap.c with bar.c, copied from Prog1, then Prog2 runs
 >         three times faster.
 > 
 >         Bob makes such a replacement and conveys NewProg2,
 >         a modified version of Prog2 that contains bar.c
 >         instead of zap.c.
 > 
 >         Question: under any circumstances whatsoever,
 >         can Alice sue Bob, users of NewProg2, or sellers
 >         of NewProg2 for violating a patent of hers?

Yes.  For example, if they also use Alice's entirely proprietary
patented techniques in Prog3 in an unlicensed verbatim copy of Prog3.

 > If the answer to that question is "yes", then very clearly the
 > GPL has failed to protect Stallman's freedom's 0, 2, and 3
 > (copying, modification, sharing of modifications).

It damn well better fail in court, because of third party patents.

 > The GPL had double failed by permitting Alice to release Prog1
 > which carried these extra restrictions.

See my example of Prog3.  I do not think there is any reasonable
middle ground between restricting GPL-enabled practice of a patent to
the claims actually embodied in the conveyed work, and full
expropriation of all patents owned by the conveyor of a GPLed work,
because it is trivial to derive a program from Prog1 and Prog3
simultaneously.

All you're really saying is "A patent system that actually does what
it claims to do sucks.  We must destroy it".  Nothing new about that.

 > A related question would ask:  suppose Bob merely studies
 > Alice's program and uses techniques found there?  What then?

If you don't know how to derive the answer to that from the answers to
your questions above, you have forgotten what a patent is.  There's
yet another family of questions: suppose Bob doesn't know Alice from
Adam or her works from all of Creation?  The answers don't change.

 > The examples in the questions obviously stand at the top of a
 > slippery slope.

No, they do not.  That's precisely what Larry Rosen likes about
patents.  There's a cliff, and either you fall off or you don't.  It
really doesn't matter if the cliff face is greased or not.

  For example, if bar.c contains a string
 > comparison routine and, instead of just re-using the file, Bob
 > cuts and pastes individual characters to transform it into a
 > memory allocator, and the memory allocator happens to infringe a
 > patent of Alice's that Prog1 doesn't even come close to
 > infringing, then it would be odd to say that Alice's
 > infringement suit indicates a failure of the GPL.

You're darn tootin' it would be odd.  Nevertheless, that's precisely
the consequence of what you're advocating.

 > I honestly can't tell where on that slope the GPL is.  It sounds
 > like some people believe that Bob can be sued.

That depends on whether he restricts practice of patents to those
covered by Section 11 or not.  If in the opinion of the court he
practices other claims not embodied in such software, he will lose a
suit and be forced to stop practicing or get a license.  I don't think
it should matter whether those "reserved" claims (ie, not covered by
section 11) are owned by a third party or by Alice.