Re: GPL and linking (was: Urgently need GPL compatible libsnmp5-dev replacement :-()
On 5/8/05, Jakob Bohm <firstname.lastname@example.org> wrote:
> On Sun, May 08, 2005 at 08:34:17AM -0400, Raul Miller wrote:
> > On 5/8/05, Michael K. Edwards <email@example.com> wrote:
> > > The "work" (subspecies "copyrightable collection", or "collective
> > > work" in 17 U.S.C. 101) known as Debian sarge CD #1. You simply
> > > cannot read C not to include this "work". Or you can, but I can't
> > > follow you there, nor do I believe that any court of competent
> > > jurisdiction in any Berne Convention country can.
> > I assert that that work satsifies the terms of the GPL (unless
> > we've done something really stupid, like include programs which
> > mix GPLed code with non-GPLed code).
> > The GPL's mere aggregation clause makes this easy.
> I believe that this is true (in the hypothetical case that the
> GPL FAQ is right, which is the topic primarily being discussed
As written, the "mere aggregation" clause, if it has any legal
significance at all, applies only to Section 2. The phrase "work
based on the Program" is defined in Section 0; and even if you want to
bend over backwards to extend the "mere aggregation" clause to Section
3 (since it references Section 2 several times), there's no way you
can apply it to Section 6. So as far as Section 6 is concerned,
you're going to have to find a definition of "work based on the
Program" in Section 0; and if you want to ignore grammar and common
law principle to argue that the legal meaning of "derivative work" is
irrelevant, you're going to have to argue for the other noun phrase in
that sentence. That phrase is: "a work containing the Program or a
portion of it, either verbatim or with modifications and/or translated
into another language".
This would mean that, if the Debian CD set (indisputably a
copyrightable work) is distributable at all (which requires Section
6's extension of terms from the original licensor to the recipient),
it is distributable solely under the terms of the GPL. In this
reading, no one who receives a Debian release in the form of a CD set
can use the Debian packages on it under any terms other than the GPL,
even if they are labeled MIT/X11, BSD, etc -- not unless they can
document having received them separately, verbatim (not just the
upstream bits), via an audit trail not colored by the GPL at any
stage. And since Debian main certainly contains works that are
DFSG-free but not GPL-compatible (OpenSSL is on CD #1 even, right?),
it would not be distributable at all under this interpretation.
Oh, you can believe personally that you want to read the GPL as
internally consistent and as having the implications that the FSF FAQ
says it does, and read between the lines accordingly. And you can
argue that Debian acted on that belief in policing shared-library
dependency relationships for GPL compatibility while bundling packages
labeled with various different licenses, some non-GPL-compatible, into
one CD image. But you aren't going to have much success, I think,
arguing that a _licensee_ is obliged to cut-and-paste text around
mentally in order to backport some other meaning to Section 0 and thus
to every usage of "work based on the Program".
I would venture to say that Debian, to the extent that there's a legal
entity by that name, is estopped by conduct from arguing that the "any
work containing the Program" interpretation applies, simply because it
constructs and distributes CD images that contain GPL and
non-GPL-compatible material. And the FSF (along with MySQL and other
copyright holders on major Debian packages) is probably estopped by
laches from trying to enforce that interpretation on anyone who got
their GPL licenses via Debian, since they have been aware for over a
decade that Debian ships mixed-license CDs. So the only restrictive
interpretation that anyone with standing can argue in a common-law
court is the one in the FSF FAQ, for which the textual basis in the
GPL is absurdly weak.
> > Reading between the lines, I believe you're ignoring the GPL's
> > "mere aggregation" clause. I also believe you are trying to
> > bend the definition of "work based on the Program" to account
> > for the treatment of that clause in the context of this license.
> The way I read it, Mr. Edwards is starting from a different
> meaning of "work based on the program" and related terms, and
> then concluding that the "mere aggregation" clause is a no-op
> under that interpretation. Although this is one of the less
> elegantly phrased parts of his messages so far.
I don't think I'd call any of my phrasing "elegant"; I feel like I'm
using a sledgehammer to drive nails into the wall to hang pictures on.
But yes, the "mere aggregation" clause in Section 2 is a no-op under
the "W = E" definition of "work based on the Program", which it is
clear to me (if not to others) is the unique legally valid way to read
> > Furthermore, you seem to think that the legal phrase "derivative
> > work" means something less than what it means (otherwise your
> > example of the Debian CD set wouldn't make sense within the
> > context you've been expressing).
> I think that Mr. Edwards has been stating very clearly up front
> that this is precisely what he thinks and precisely the basis
> for most of his other conclusions.
Precisely what I think is that the legal phrase "derivative work"
means exactly what it means under the relevant jurisdiction's
implementation of the Berne Convention, no less and no more. My
example of the Debian CD set is a reductio ad absurdum based on the
counterfactual stipulation that W = C instead of E.