Re: Questions about legal theory behind (L)GPL
On Mon, 17 Jan 2005 13:48:23 -0500, Raul Miller <email@example.com> wrote:
> > > The GPL is a license document, and "automatically receives" is a
> > > license grant. The GPL doesn't need to be law to grant license --
> > > granting license is what copyright licenses do.
> On Sun, Jan 16, 2005 at 10:48:55PM -0800, Michael K. Edwards wrote:
> > "The GPL isn't law" was in response to "the GPL doesn't say this is an
> > authorization to sublicense". Under US law as I understand it,
> > there's no other way to implement the purported license grant
> > indicated by "automatically receives" other than the sublicensing
> > paraphrase that I gave.
> Why would direct licensing not work?
As I understand it, generally speaking, a contract has two parties --
offeror and offeree. To the extent that it binds other persons or
entities, it does so through the doctrine of agency -- either party A
declares that non-party B will fulfill some of A's obligations as an
agent of A, or A agrees, acting as an authorized agent of B, to commit
to conduct on B's behalf. The GPL appears to me to fall under the
latter, authorizing the licensee to offer a sub-license to all
copyrights in the incoming GPL work. To get the same effect with
"direct licensing", you'd have to read separate offers of contract
from each copyright holder to the recipient into the single act of
passing her a modified work, which is a little far-fetched.
> > IANAL, and I can't say for certain how a court would weigh the GPL
> > drafters' intent (which I agree is reasonably clear on this particular
> > point) against precedents like Everex v. Cadtrak -- especially if two
> > copyright holders differ from one another on the interpretation.
> I imagine that (where two copyright holders differ from one another in
> their interpretation) the judge would look at the history of how these two
> copyright holders have acted. If one has recently changed their intent
> then the judge would need to consider their previously expressed intent.
> If there is no such change, then the judge would probably look at how
> the situation developed, to determine which parts of the copyrighted
> work belong to which party.
On the question of sub-licensing, I doubt that you would be able to
find evidence of either copyright holder's stance in advance, and it
wouldn't matter much anyway, since as a matter of law (in the US)
ambiguities in contracts must be construed against the offeror and
there's no way to demonstrate the licensee's intent in a
non-negotiated, "standard form" contract. (That isn't necessarily
true if there's a history of correspondence between the parties and it
can be demonstrated that both interpreted the contract in the same
My guess (IANAL) is that a court would find that, when A offers
Project X under the GPL, B modifies and distributes it, and C accepts
license in the modified version, B and C have formed a contract and
A's participation is limited to the agency for sub-licensing purposes
implicit in the contract that it offered B. This is especially likely
to hold in a situation where B is Debian, since most users deal
directly with Debian for updates, bug reporting, etc., and can
reasonably claim that as far as they are concerned their license came
from Debian and the rest is between Debian and the upstream(s).
> > Suppose Ms. X contributes some code to Kaffe and then sues Debian for
> > distributing Kaffe and Eclipse together. Then suppose that the FSF
> > files an amicus brief saying that Debian is OK because GNU Classpath
> > has a special linking clause and Ms. X's code is part of an
> > interpreter, while the main copyright holder on Kaffe files an amicus
> > brief saying that as far as he is concerned the GPL doesn't propagate
> > across linking boundaries and that if Ms. X says different then she's
> > failing to extend the same license to Debian that he extended to her.
> > Whose interpretation wins? The answer could depend critically on what
> > implicit terms the court construes in order to implement the implied
> > authorization to sublicense -- or some other way around the problem
> > that I'm not seeing.
> Most likely, the judge would say that Ms X doesn't have standing.
How could that be? Factually, her copyright has been infringed unless
Debian (reachable through SPI and/or as a list of named defendants
plus a stack of Does and Roes) can demonstrate that it acted under
> Eclipse is not a module of Kaffe.
I don't understand what legal significance you expect this to have in
this situation. I have argued that no derivative work containing
Eclipse and any part of Kaffe or Classpath is created at any stage,
since a "derivative work" is by definition an "original work" unto
itself, and the interpretation and linking processes don't create
"original works". But there is no question that both the Debian CD
and the system on which Eclipse and Kaffe are installed infringe on
Ms. X's copyright in the absence of a valid license to Kaffe.
> In the unlikely event that she did have standing, I'm sure the judge
> would ask her what she thought people would use Kaffe for, and why she
> contributed the code.
Why would that matter? But suppose she did, and Ms. X answered that
it was her understanding that her contributions could only be used by
GPL applications, based on an argument similar to Grzegorz's. I think
that the defendants could successfully argue that they relied on the
actual text of the license, with its use of terms with a definite
meaning in copyright law, and a reasonable review of the relevant case
law, to conclude that the GPL doesn't cross published API boundaries
irrespective of the technical details of linking, interpreter
For the record, that's how I approach the issue -- relying on license
text and law, not the FSF's FAQ or the opinion of any particular
copyright holder -- and that's the defense that I think I would offer
if I were ever in the defendants' position. In the hypothetical
situation I have described, other lines of defense might be plausible,
including reliance on the Kaffe FAQ or the FSF interpretation, or even
an argument that Ms. X is estopped from pursuing the claim because it
amounts to an attempt to assert an implicit term that conflicts with
an implicit term in her license from other Kaffe copyright holders,
and thus voids her license.
> [Also, if the FSF did get involved, I imagine they'd be able to cover
> a lot more ground in that brief -- I don't think they'd limit the scope
> to classpath.]
Read the FSF's brief in MySQL v. Progress Software -- if you ask me,
it is less than scintillating as an exposition of the legal foundation
for their position. The FSF, in recent years, seems to be much more
interested in protecting its position from competent scrutiny and in
prying source code out of infringers by force than in creating a
climate in which the ground rules for use of the GPL are well
understood and well founded in law. I probably don't have to tell you
that, outside of communities that are ideologically committed to free
software, the GPL is widely feared and distrusted -- and the FSF seems
to like it that way. To a pragmatic fellow-traveler like me, this
doesn't feel like freedom.
(Based on the interpretation I have given, it is my usual practice in
my private capacity to cross my fingers and convert the LGPL and all
other GPL variants to the GPL before I "accept" the license. I intend
to rely, in the event of legal proceedings, on precedent and judicial
competence rather than on special exemptions. Nevertheless, it is not
my habit to use or advise the use of FSF copyright code in ways that
are contrary to their interpretation of the GPL. As I've said before,
I think that doing so is discourteous and, if one has something to