Subject: Re: termless copyright and patents
From: <stephen@xemacs.org>
Date: Mon, 2 Oct 2006 19:13:16 +0900

I think we need a real lawyer here, but I have a couple of comments.

Thomas Lord writes:

 >     A party's "essential patent claims" in a work are all patent
 >     claims that the party can give permission to practice,
 >     whether already acquired or to be acquired, that would be
 >     infringed by making, using, or selling the work.
 > 
 > An example, under Mr. Rosen's reading:  The company Global Runes
 > has a patent on a Unicode string comparison algorithm, that
 > algorithm used to route Jabber messages, that algorithm tweaked
 > for field comparisons in a database, and that algorithm combined
 > with hardware to create a utility for routing Jabber messages.
 > 
 > Global Runes distributes a software-only embodiment of the
 > algorithm in the form of a Jabber router.

And you conclude that Balkanization results.  I disagree.  The
algorithm per se is an essential claim and you can use it for anything
not covered by a "reserved" claim.  So the underlying algorithm is not
a problem here; it has been made available by its use in the Jabber
router (JR).

So suppose the database comparison (DBC) claim is not licensed or
covenanted.  You can't use the algorithm in a database comparison.
Nothing to see here, and definitely not "Balkanization", move along.

Now suppose the DBC claim is convenanted in a different GPL
application.  You sure can put that algorithm in a single library, and
link any application that embodies either the DBC or JR claim to it.
I don't see any "Balkanization" here.  Of course, you can't practice
patented claims that have not been released under the GPL covenant or
similar, but there's nothing new here.  People who want to use the
library simply need to read a list of claims that may be implemented
freely via the library.

Note this depends on a specific definition of Balkanization, see below.

 > If everything is hunky dory under the GPL but Mr. Rosen is
 > wrong, then "arising from your exercise of rights under this
 > License" includes modifying GPL code in the usual way -- so the
 > rights to to the database claim were given away in the Jabber
 > client as were the rights to combine with hardware.

I don't see how you get there, but even if you do, there's no
Balkanization there.

 > I think I understand Mr. Rosen to be warming up to a different
 > battle.   I *think* he wants to say "GPL is a copyright license
 > and therefore it says nothing at all about, for example,
 > combining this software with hardware.   It can not govern
 > rights that properly belong to patent law."

I'm sure he would deny that the first sentence represents his opinion.
The last sentence is moderately accurate IMO, but I expect that he would
phrase it "It cannot govern rights not embodied in the software."

 >  >  > Now, yes, perhaps someone will argue that the GPL covenant is very,
 >  >  > very, weak.   This would be a serious blow to the utility of the
 >  >  > GPL.
 > 
 >  > The GPLv3 covenant is extremely strong, since it basically makes the
 >  > claim available to any use of all GPLed code as soon as its
 >  > implementation is added to libsection11.
 > 
 > Which reading are you using?  Am I free to move the code from GNU
 > Awk to libsection11 without additional permissions from the
 > patentee who conveyed GNU Awk to me?  If not: balkanization.

Yes, you are free to move that code.

People (including you) are not free to use it in certain ways,
however.  This is always true: you can modify *any* code (if
necessary, by a full-text transplant of the sample implementation :)
to infringe *any* software patent claim.  The fact that the
restrictive claim in question here is owned by the person who conveyed
the software would, I suggest, be considered immaterial by a court.

 > If so: the "leaky claims" issue.

The claims will not leak.  If this comes up in court, the attorney for
Global Runes will argue that the "essential claims" are those that are
practiced by the conveyed application, and the attorney for the
plaintiff will argue that the "essential claims" are all claims owned
by GR that somebody deriving from the conveyed work might want to use.
As usual there are two cases.  The first is that the plaintiff is
downstream of GR, claiming that GR is violating their covenant.  But
it's *GR's covenant*, and it's hard to see how the court can rule that
"essential claims" should be interpreted for the convenience of
downstream when it's a pure grant.

The second is that plaintiff is a copyright holder upstream of Global
Runes.  I hope that the court will rule in favor of GR, but if not, GR
gets convicted of infringing the copyright of the upstream entity.
The worst outcome is that the FSF will be awarded damages.  Most
likely GR will merely lose their license to the GPLed software.

In the case of severe damages, Global Runes might prefer to settle out
of court in a way that "leaks" the claims, but you can't get a court
to enforce a "leak" for you.

 >  > If you need a "better" version of a possibly patented
 >  > algorthm, you go and look in the source for libsection11 and
 >  > derive from that.
 > 
 > Does every patentee represented in libsection11 have to, at some
 > time, convey libsection11?

????  Of course not; just the software embodying his own claim.

 >  > your fear of Balkanization is completely unwarranted.
 > 
 > I hope I've succeeded in giving you some doubt about that.

The only doubt I have is about your definition of "Balkanization".  If
you mean that each claim splits the GPL world into two parts, and that
the partition will be different for claims implemented in different
APIs, then I agree with you.  There will be a cost of determining
whether a particular API is available to your application, since that
might involve infringing a related claim.  If you wish to avoid that
cost entirely, just don't link to libsection11.  It is trivial to
determine if you need to pay that cost at all---remove "-lsection11"
from your link line.  This kind of "Balkanization" seems no more or
less expensive than having incompatible copyleft licenses, except that
there are even more patents than there are copyleft licenses. :-(
(Feel free to prove me wrong, I haven't thought very carefully about
it.)  Given the trivial litmus test, I think it's reasonable to leave
the decision up to those who want to practice these patents, and
forget about the "Balkanization" issue.

What I thought you meant by "Balkanization" was that the routine
implementing the JR claim and the routine implementing the DBC claim
must be kept separate.  That is definitely not true by the nature of
patent, but if it were it would make Section 11 almost useless unless
it could somehow pry loose IP that is not embodied in the conveyed
software.