On 6/11/05, Michael K. Edwards <[EMAIL PROTECTED]> wrote: > P. S. Note, however, that the Linux kernel is a derivative work of > works by some other authors, such as netfilter/iptables. I don't mean > to say that no one but Linus can file a claim of copyright > infringement or breach of contract against someone who is doing > genuinely inappropriate things with the kernel and friends (cf. the > Sitecom and Fortinet cases). But in order to demonstrate standing, > they're going to have to identify a work of authorship, not totally > subsumed in the Linux kernel development process, on which their > contribution rises to the level of co-author, as Harald Welte did with > respect to netfilter/iptables.
Let me explain a little further why I think "derivative work" is the right theory for the relationship between netfilter and the kernel. I will start with an excerpt from the House Report's commentary on 17 USC 201: <quote> Two basic and well-established principles of copyright law are restated in section 201(a): that the source of copyright ownership is the author of the work, and that, in the case of a ''joint work,'' the coauthors of the work are likewise coowners of the copyright. Under the definition of section 101, a work is ''joint'' if the authors collaborated with each other, or if each of the authors prepared his or her contribution with the knowledge and intention that it would be merged with the contributions of other authors as ''inseparable or interdependent parts of a unitary whole.'' The touchstone here is the intention, at the time the writing is done, that the parts be absorbed or combined into an integrated unit, although the parts themselves may be either ''inseparable'' (as the case of a novel or painting) or ''interdependent'' (as in the case of a motion picture, opera, or the words and music of a song). The definition of ''joint work'' is to be contrasted with the definition of ''collective work,'' also in section 101, in which the elements of merger and unity are lacking; there the key elements are assemblage or gathering of ''separate and independent works ... into a collective whole.'' The definition of ''joint works'' has prompted some concern lest it be construed as converting the authors of previously written works, such as plays, novels, and music, into coauthors of a motion picture in which their work is incorporated. It is true that a motion picture would normally be a joint rather than a collective work with respect to those authors who actually work on the film, although their usual status as employees for hire would keep the question of coownership from coming up. On the other hand, although a novelist, playwright, or songwriter may write a work with the hope or expectation that it will be used in a motion picture, this is clearly a case of separate or independent authorship rather than one where the basic intention behind the writing of the work was for motion picture use. In this case, the motion picture is a derivative work within the definition of that term, and section 103 makes plain that in a derivative work is independent of, and does not enlarge the scope of rights in, any pre-existing material incorporated in it. There is thus no need to spell this conclusion out in the definition of ''joint work.'' </quote> The "elements of merger and unity" are not lacking in the kernel, so there's no way it can be called a collective work (except as regards the firmware blobs, which are not part of the kernel and not under the GPL, no matter what the headers in the drivers may say). Oh, maybe a driver maintained exclusively by a vendor isn't "merged" enough to avoid "collective work" status; but the history of periodic rework is so clear that I don't think a judge would have a hard time accepting that the kernel is every bit as much a single work of authorship as Spike Lee's "Malcolm X". So if netfilter were not accompanied by a component outside the scope of the kernel, and were intended from the beginning to be "absorbed or combined into an integrated unit" together with all other kernel contributors' work, then the only theory on which Harald Welte would have any claim to authorship would be as co-author of the kernel as a joint work; and as I wrote before I don't think that claim could survive the analysis of Aalmuhammed v. Lee. But since netfilter+iptables also exists as an independent work of authorship -- at any given time, an update to the netfilter subsystem of the kernel plus a tightly interlocked userspace tool for control and status reporting -- Harald does have a claim of co-authorship on that work. Recall that there was no particular distinction between what are now known as "collective works" and "derivative works" under the 1909 Act, which was perfectly happy with this clause (added in 1947): <quote> 7. Copyright on compilations of works in public domain or of copyrighted works; subsisting copyrights not affected Compilations or abridgements, adaptations, arrangements, dramatizations, translations, or other versions of works in the public domain or of copyrighted works when produced with the consent of the proprietor of the copyright in such works, or works republished with new matter, shall be regarded as new works subject to copyright under the provisions of this title; but the publication of any such new works shall not affect the force or validity of any subsisting copyright upon the matter employed or any part thereof, or be construed to imply an exclusive right to such use of the original works, or to secure or extend copyright in such original works. </quote> Why, then, does the category "derivative works" warrants its own definition in the 1976 Act? Judging from the House Report, largely for the sake of the 203(b)(1) / 304(c)(6)(A) exception to license termination. As discussed previously, this exception was largely intended to remedy situations in which a large work requiring a large financial investment to produce (such as a film) was held for ransom by the copyright holder on a small work embedded within it (such as a song). In fact, under the 1909 law there were apparently instances "in which third parties have bought up contingent future interests as a form of speculation", which is why an author cannot under the 1976 law offer a new license to anyone other than the original licensee (or the licensee's successor) until the effective date of termination. The relationship of a song (or perhaps rather an orchestral score) to a film is precisely the relationship of netfilter to the kernel -- an independent work of authorship, perhaps written specifically for the purpose of accompanying the larger work, but with some independent merit as a creative work (perhaps together with other material that is part of the independent work but not used in the larger work). That's the sort of situation for which the "derivative work" category was conceived. Cheers, - Michael
