Subject: Re: termless copyright and patents
From: Thomas Lord <lord@emf.net>
Date: Thu, 05 Oct 2006 09:13:19 -0700

Stephen, you are confused (re "I think it sucks").   Your
confusion leads to some clarity about just how bad the
"essential patent claims" language is in GPLv3 draft 2.


We have two proposals on the table for what a patentee should
covenant when conveying a GPL work.  It's handy to give these
names:

    The "essential claims covenant": Patentee promises not to
    assert any claims arising from the making, use, or sale of
    the program conveyed against uses of that program or
    derivitives of that program.  (Some people say that this is
    the meaning of GPLv3 draft 2).

    The "suggested claims covenant": Patentee promises not to
    assert any claims arising from the making, use, or sale of
    any program under GPL if the claimed infringement is
    suggested by (made obvious by) teachings in any GPL works
    conveyed by the patentee.  (This, I say, is what is
    consistent with the four software freedoms the GPL aims to
    protect and promote.)


Stephen's objection (spliced from two messages):


    Stephen> I think it seriously sucks, [...]

    Stephen>  ["suggested claims"] proposes to make *every*
    Stephen>  infringement case involving the GPL into a contest
    Stephen>  equivalent to an attempt to demonstrate invalidity
    Stephen>  of a patent.


That is terribly misleading.

The truth is that that infringements protected by an essential
claims covenant are *also* protected by a suggested claims
covenant and, most importantly:

         A valid defense that invokes an essential claims
         covenant *is also* a valid defense given a
         suggested claims covenant.

Consider this thought experiment: everyone in the court room
except the attorney for the defense knows that a suggested
claims covenant is in place.  The attorney thinks an essential
claims covenant is in place.  The attorney validly applies tests
which show the infringement is protected by an essential claims
covenant.  *Those exact same arguments* establish protection
under a suggested claims covenant.   The defense wins.

The converse is not true.  A suggested claims covenant protects
more infringements.  There can be cases where a valid defense
invoking a suggested claims covenant *would not* be a valid
defense given only an essential claims covenant.  Yes, in these
cases, the court has to be asked to apply the obviousness test
in a non-trivial way.  But so what?  In such cases, given only
essential claims protection, the defendant would simply lose:
here, she's got a fair chance.

There is also a way in which "suggested claims" is simpler.
"Suggested claims" is built on concepts that the court already
understands:  obviousness.   In interpreting and applying a
suggested claims covenant, courts can draw on all the existing
precedent about how to evaluate obviousness.

On the other hand, an essential claims covenant is built on
a novel construct: "essential claims".    These covenants don't
ask a court to decide, merely, if there is infringement.   Rather,
these covenants ask a court to imagine the set of all possible
infringements which can arise from the exercise of GPL rights
regarding a conveyed program -- and then to ask if the infringement
being challenged relies on claims which are subset of the
essential claims.    How the heck is a court supposed to do that?
The concept of "essential claims in" a work doesn't exist outside
of Moglen's and Stallman's heads.   It isn't found in law or precedent.
There is no clear, definitive, simple test for it.  It could mean
just about anything supported by the ordinary english words
in their context and it could mean nearly nothing at all.

Here is an example of how bad "essential claims" can be:

Alice releases a version of `sort' that contains her patented
string-sorting program.    This version of sort always sorts
newline-terminated lines.

Bob sells a modified version of Alice's sort.   In his modified
version, it always sorts fixed size records of 64 characters each.

Nothing in any of Alice's patent claims speaks at all about
lines vs. fixed-size records.

Nevertheless, the infringement that arises from Bob's program --
the actual invention that we plunk down on the table next to
the patent -- is a "making, use, or sale" of Alice's patent which
*could not arise* from the exercise of GPL rights *of the program
which Alice conveyed*.    Don't be confused that Alice has
promised not to assert against derivatives: she has only promised
not to assert infringements that could arise from the originally
conveyed program.

Now, you, me, RMS, and Moglen all know what we meant.
We expect Bob's program to be protected by the covenant.   But
our belief that it is depends on our interpretation of "essential
patent claims" and our interpretation of "arise from the exercise
of rights [under the gpl] [of the conveyed program]".   Alice
doesn't share our beliefs.   Why must a court?

A judge could quite sanely say:

    Ok, Bob,  you say that your program is protected by
    Alice's covenant.   Simple enough:  please show me
    how the program Alice conveyed can replace yours.

Of course it can not.   Bob's change from lines to fixed length
records is not protected, regardless of the fact that Alice's
patents make no such distinction.[*]

-t

[*] Yes, of course one could write a shell pipeline, using
    Alice's program, and the *pipeline* would sort fixed length
    records -- but in such a pipeline, Alice's invention is
    being practiced to sort newline-separated lines (her
    "essential claim").  That isn't obviously a defense of Bob's
    program -- it is a work-around that respects Alice's patent.