SFWA (in the form of its agent, Andrew Burt) was clearly in the wrong when it requested that a large number of items posted to scribd.com be taken down. Scribd.com was clearly in the wrong when it complied with this request from an incompetent party. Of the two, scribd.com seems the more culpable in that they took action without investigating the competence of the requestor of the action or considering the form of the request.
It's interesting, and perhaps indicates something about the attitudes of the people involved, that SFWA and Burt are taking the brunt of the blame -- this is in contrast to the reaction to removal of material from publication on LiveJournal by Six Apart, where the blame fell on the actor, Six Apart, rather than the various incompetent requestors of removal.
Reading the two e-mails from Burt as "epiracy@sfwa.org" leaves me with the impression that he is incompetent in this matter in more than a legal sense. I have the impression that he meant well, and from what he writes Burt does seem to have intended to negotiate with scribd.com to cooperatively find a way to avoid recurring unauthorized posting of science fiction and fantasy. What got him into the mess at hand is what amounts to a throwaway sentence referring to a list which might otherwise have been taken as a "scientific wild-ass guess" about the amount of material which was likely to have been posted without the approval of the holders of the intellectual property rights associated with that material.
Burt's notion that he had solutions to the practical problems of unauthorized posting of fiction is terribly naive, though, even for a computer science academic. The idea of identifying work by "statistically unlikely phrases" might be a good one, and if it is then it's probably already patented, perhaps by a commercial enterprise named for a Large South American River. In fact, most of the obvious ways of detecting similarities in bodies of text representing prose have almost certainly been thought of, applied, and possibly patented by companies which produce software for the detection of plagiarism.
Requiring declaration of rights clearance by anyone wishing to upload material to scribd.com clearly misapprehends the relationship between customer and vendor: there is really no incentive for scribd.com to make use of their service more onerous for customers than it is required to by law. In effect, the existence of the 12-inch gun of the DMCA is disproportion as a response to the gnat-like inconvenience of individuals posting unauthorized editions of sf novels, but it obviates any business case for checking customers for gnats at the door and holding them back from doing business with scribd.com. Given a legal framework that puts the onus for identifying infringements of rights on the owners of the rights and requiring the act of publication before any of the framework's provisions are in effect, it's going to be hard to find a purely pecuniary prompter of proactive publicatory prophylaxis.
Finding common ground between the purported owners of rights and the owner of a web publication site does seem possible, in particular because the after-the-fact remedies of DMCA are complex and overpowering. There might be a good role for SFWA and other writers' organizations in vouching for the identities of their individual rights-owning members. In particular, it could be much cheaper and more convenient for a site like scribd.com to provide a complaint interface if an organization like SFWA did the work of associating author names and pseudonyms to a set of identity codes. This might allow scribd.com to accept a complaint about misappropriation of Padgett's work from some guy named Kuttner as actually competent, for example, while screening Doctorow's work from well-intentioned but incompetent protest by Burt.
