I thought I'd take a different tack for a minute and write about things that Raul and I (and other current debian-legal participants) seem to agree on about the GPL, and seem to think (though most of us are not lawyers) are well founded in law.
1. GPL release is not release into the public domain. There is appellate precedent for this in the US and civil court precedent in Germany. Copyright in works offered under the GPL is retained by the author, his or her employer, or an assignee, and ultimately it is this retention of copyright that gives GPL enforcement efforts their teeth. 2. GPL release on the Internet, if done under legitimate authority in the first place, is hard to undo. Caveat: there is not really very strong language in the GPL to indicate a "perpetual" term, and there are jurisdictions where a contract with no explicit term may be terminated at will. However, it's going to be relatively hard for a copyright holder to claim that they did voluntarily license works under the GPL, and did seek their wide distribution via the Internet, but did not intend to create a basis for reliance on a perpetual term for the license to copy and to create derivative works. 3. There is a large category of "derivative works" that the GPL legitimately offers license to create and distribute, solely on the conditions of source code release discussed in Section 3; we generally agree that the Debian-packaged version of a given GPL work is such a derivative work. There is also a large category of "collections containing GPL works", authorized by the "mere aggregation" clause or otherwise, which are not obliged to be offered exclusively under the GPL; we generally agree that distro CDs are in this category. When such a "collection" is offered, the GPL obligation to offer source has full force with respect to the individual GPLed components. 4. There are some "moral rights of the author" that are more or less universally recognized (such as the right to truthful attribution) and others that apply in some jurisdictions but not others. Wherever these rights exist, they are reserved to the author (not the copyright holder) and cannot be contracted away. They are, in a sense, outside the copyright calculus, and may provide an independent mechanism for obtaining redress for some abuses. Applying them to software is, however, a bit of a stretch, both in the EU and in the US (where they are in any case technically recognized but only quasi-implemented; see http://aic.stanford.edu/jaic/articles/jaic36-02-006_3.html ). 5. There is some cause for long-term worry about termination of license by a primary author's heirs under US copyright law. I worry about this less than some others seem to, since as I read it (for example) the only person who could obtain a valid registration of copyright on the Linux kernel in the US is Linus Torvalds; see Aalmuhammed v. Lee ( http://caselaw.lp.findlaw.com/data2/circs/9th/9955224.html ) for why. Note that this is not necessarily inconsistent with Harald Welte's success in obtaining copyright registration in Germany for the netfilter subsystem and prosecuting it successfully, as civil law countries seem to handle "joint works" differently, and in any case the validity of his copyright was probably not contested on "degree of authorship" grounds by the defendants. I hope this will help avoid misunderstandings such as "MKE says that A+B+C is an uncopyrightable collection! Does that mean he thinks copyright on A, B, and C is voided by combining them?" Would someone care to contribute the next few points of agreement? Cheers, - Michael