Subject: Re: termless copyright and patents
From: <stephen@xemacs.org>
Date: Sat, 7 Oct 2006 23:24:30 +0900

Thomas Lord writes:

 > If a court is asked to decide "What are these permissions that
 > Alice gave?" the answer has to be "The narrowest set of
 > permissions that is consistent with the requirements of the
 > GPL."

Precisely.

 > So, again, if Alice is conveying a 'sort' program, she has
 > no obligation to permit her invention to be converted into
 > a general purpose library.

I will take that sentence literally, ie, to refer only to the act of
modifying the code.  This *would* be an infringement of Alice's patent
absent the covenant.

The GPL grants the right to transform a copy of a sort program from a
standalone program to a library, does it not?  Such modification is
therefore "exercise of rights" as described in section 11.

The covenant says that the patent-holder promises not to assert any
claims *in the original work*.  Those claims are *patent* claims, ie,
abstract---it doesn't matter to what they "are" whether they are
embodied in a verbatim copy or a modified copy.  Alice has promised
not to assert those claims against the modified copy because it arose
from the "exercise of rights", namely the right to modify.

 > (If you want to do forensics on that, look very closely at the
 > definition of "essential patent claims IN a work" and then again
 > at section 11.)

I see nothing there that even attempts to change the inherently
abstract nature of patent claims in the way that you are trying to
do.  It is the *abstract* claim that Alice promises not to assert, not
any concrete embodiment.

The definition of "essential patent claims in a work" does not change
the definition of "patent claim" in any way.  It simply says that if
you can't find the claim in the work, then the covenant of section 11
doesn't cover it; while if you *can* find the claim in the work, the
covenant of section 11 covers it in *any* code that was produced as a
result of running, copying, modifying, or distributing as permitted by
the GPL with respect to the work that Alice conveyed.  It further
defines "find" in terms of the well-understood (by lawyers and judges,
at least) process of determining infringement of a claim.

 > You seem to want to believe that if the program Alice conveys
 > is captured by scope assertions in "claim #2" of her patent,
 > that therefore she has forgiven all possible infringement under
 > claim #2 in the conveyed work or any derivitive.

That's right.

 > That would be plausible BUT FOR the way that section 0 defines
 > "essential patent claim".

Section 0 can not and *does not* define "patent claim"; it defines
"essential ... in a work".  That's all it does.

 > Now I find that a certain company is selling the medicine in
 > the form of a gel capsule.   We negotiate a patent license.
 > I DO NOT have to sell a license that says:
 > 
 >      In consideration for .... licensee may manufacture
 >      and sell the compound in any way that falls under the
 >      scope of claim #2 of my patent.

That's right.  You don't.  But a patentholder who conveys a work under
the GPLv3d2 must, or it's not the GPLv3d2.

 > A patentee can not induce infringement and then turn around
 > and sue for the infringement they induced.  This is why, for
 > example, you see some classes of product marked with patent
 > numbers or marked "patent pending".

Highly unlikely, although I don't have time to read 35 USC right now.
(Modulo my temporary ignorance) the difference is not whether you can
sue for infringement (cf the LZW patent), but whether the infringement
will be classed as inadvertent (temporary injunction to stop and/or
payment of back license fees) or deliberate (treble damages).

 > Authoring or conveying GPLv2 code is inducing the behavior
 > commonly known as "open source practices" which includes a

Sorry, but you have no standing to decide what the license induces.
It induces what it says, which is "you may copy the code and
redistribute it" (and that is all that you can deduce from a
*copyright* license), plus anything the rightsholder says.  Copying
the code clearly requires infringing any claims in the copied code, so
that's permitted (asserting those claims after granting a license
would be inducement).  So you are allowed to copy, modify, and
redistribute code embodying those claims if you start with a copy
licensed to you by the patentholder.

GPLv2 has two big ol' holes in it, though.  The first is *really* big,
that GPLv2 doesn't grant permission to *run* the program!  So the
implicit patent license arguably doesn't apply to that!  The second is
that the whole issue of the implied patent license becomes really
murky with respect to anybody who didn't receive the GPL from the
patentholder.  In the case of a verbatim copy, I think you're OK.  But
in the case of a modified copy, the downstream recipient has *not*
received a license to the *claim*, he has received a license to the
claim *as embodied in the work conveyed by the patentholder*, and what
that means a court would have to decide.

Ie, GPLv3d2 makes things clear, while it is *GPLv2* that is subject to
the kinds of problems you're describing.

 > pragmatic use of code borrowing (for other GPLv2 programs) and
 > re-application of lessons learned.

Wishful thinking.  GPLv2 nowhere says that the author or conveyor of
the code *who owns a patent* has any responsibilities whatsoever!
Patent licensees do have an explicit responsibility.  Furthermore, the
GPL explicitly recognizes that licensed GPL activities may infringe on
forms of IP other than copyright.

If I were the judge deciding what GPLv2's implicit license says, I
(like simo and Stallman and Moglen) would probably come up with
something like paragraph 1 of section 11 of GPLv3d2.  But I can easily
imagine both substantially more restrictive readings (including that
the license is only for verbatim copies and personal mods) or more
permissive readings (including your "suggestive claims").