The Non-Revocable GPL (Groklaw)

The GPL is a legaly-binding promise anyone getting downstream code is automatically granted a
license to do all sorts of stuff with it by the original authors. So you're right the licenses
stem from the authors not the intermediaries. However given these authors made the promise to
automatically grant them in the future with no time limit, they can't elect to stop issuing
them later. The law is very careful about people not reneging on past promises.

Those 100 people have licenses, and that license permits them to distribute it to more people,
under the GPL, and so on and so forth. Yes, you can't get new copies *from the original
author*. But who cares, get them from a single person who did get it from the author a ways
back. In this way a GPLed source can never be revoked, _unless_ somehow all copies vanish from
the internet.

Have you raised this concern with the FSF? If it's a legitimate concern, (a) they need to hear
about it, and (b) they must have thought of it themselves.

However, I think you're wrong, and here's why:

The GPL, as applied by the author of a work to that work at the point of distribution, allows
anyone who receives the work to redistribute it under the terms of the GPL. Not to relicense
it under the GPL, but merely to redistribute it under those terms.

The terms of the GPL, of course, allow any recipient of the software to redistribute it under
the terms of the GPL.

There's an infinite loop built in there, and that's what stops it being a one-step-only
process. If the first redistribution is possible, the inherent loop makes every subsequent one
allowable.

I suggest you talk a long session with a lawyer on this. I have never ever heard of someone
succeding in revoking a license if terms in said license hasnt been reached. The GPL license
is very clear and do not contain any terms that can be used to revoke the license from the
issuers point.

The only thing the issuer can do is to dual license or license new bits under a different
license. You cant take something you have already given away or sold back without something
explicitly stating you are entitled to do so in a contract/license. 

Sorry, I can't cite anything. A discussion of a related (but not identical!) topic on Advogato
five years ago was also inconclusive: http://www.advogato.org/article/606.html

The difference is that the previous discussion was about retracting distribution rights from
current licenseholders, where this is about whether those rights necessarily extend to new
recipients even if the owner says no.

The above mode of argument has a name.  It's called "begging the question".  At the core, its
pattern is "Assuming (1) I'm right, then (2) ... (n) therefore I'm right."  

You cannot reason about how the law will treat a license, a licensor, and licensees based
solely on the text of the license.  The law determines much of the meaning.  The copyright
holder has rights under the law that aren't stated in the license, and that may override what
is stated in the license.  That's the topic under discussion.  If you haven't got that far,
you haven't got anywhere.  

The problem is that the law isn't clear.  You can wish all you like that the law were both
clear and agreeable with your prejudice, but the facts are otherwise.

"Thus, they voluntarily give up the right of revocation in return for ..."

That's the theory, but it begs the question: have they, in fact, given up that right?  Have
they, furthermore, caused all subsequent owners to give up that right, even the ones who don't
want "the greater good"?  I don't know, and I have every reason to believe you don't know
either, wishful thinking notwithstanding.

Please try to keep up.  

If you transfer ownership of the copyright, it is not at all clear that the new owner is bound
by the promises you made.  Sure, *you* still won't assert your privilege under copyright law
to restrict publication, as promised (indeed you can't, any more), but the new owner never
made any such promise.  The contract is only between the former and new owner of the
copyright, which needn't involve licensees at all.  Sometimes there's no chance to involve
licensees' rights in the transfer, as in death without a will, or confiscation after
bankruptcy, and there's no legal requirement for it.

I agree, Mr. Yellman.  In addition, it doesn't matter (in the legal sense) whether they get
anything or not.  Licenses do not require exchange of consideration as contracts do.  They are
a grant of privilege, possibly with stipulations of responsibility attached.  The
responsibilities attached don't necessarily convey any advantage to the licensor--there are
motivations besides those of the economic variety.  In addition, it also doesn't matter
because no action is occurring at all on the part of the author; the license text merely
states that anyone acquiring a copy of the software has, already, by explicit stipulation in
the license, the same rights and responsibilities as everyone else, with no requirement to
inquire of the author to verify this.  The rights of usage, redistribution and modification
are "inalienable", so to speak, for anyone with access to the software.

They have that access anyway.  Therefore, it's not an exchange, therefore not a consideration.

Wishful thinking is muddy thinking.

I don't (yet) believe that no-one can cite anything on this subject.  It is certainly possible
that there is no case law that would bear on this subject (is permitting multiple levels of
distribution a grant of permission, or simply an unenforceable, any-time revocable notice),
but I would not trust that until I heard it from a lawyer specializing in copyright law, after
that lawyer had done the research they felt necessary.  And a statement like that, if someone
has done that research, could be cited.  

Until someone posts either a case, or, even better, a statement from a copyright lawyer, we
DON'T KNOW; there may be a risk, but there may not be.  At a minimum, someone should put in a
request to the SFLC to opine on this subject.