Makin' More CopiesJulian suggests a possible exception to my argument below about retroactive copyright extensions, and I think he's probably right, though I have doubts about the empirical significance of the special case he describes. Essentially, the creative works in question would need to have three characteristics: (a) they are improvements on existing works, and the improvements have a relatively high cost for the first production; (b) the resulting improved work has a relatively low (zero or nearly zero) cost of reproduction thereafter; and (c) the improved work is not copyrightable in its own right.
There a couple of reasons I doubt the empirical significance of cases like this. First, there are various ways in which condition (c) can be avoided. For instance, if you convert an old document into an html document without any other modification, that's probably not copyrightable -- but if you include your own original annotations, I believe it is copyrightable (again, assuming I'm not mistaken about the operation of copyright law). Similarly, cartographers distinguish their own maps from others' for copyright purposes by adding fictional towns and other unique features to their maps. Second, I have a hard time thinking of good examples that satisfy condition (a) without violating condition (c). It's interesting that Julian gives the example of converting old works into html form and posting them on the web, because Eldred (the petitioner in this case) is a guy who had been doing that very thing with works like The Scarlet Letter without remuneration, and the 1998 copyright extension prevented him from doing the same with more recent works. Point being, the initial costs are often low enough in cases like this that some people are apparently willing to incur them for free, which reduces the significance of the problem (if any). It strikes me that this is probably the same phenomenon that Julian (in a previous blog post that I can't seem to locate) identifies as the reason open-source programming can overcome the incentive problem: as the number of interested parties grows, the likelihood of having at least one person willing to provide a relatively low-cost service purely for reputational gain approaches one.
Nonetheless, I concede that there is a theoretical possibility of cases that meet all three criteria. Perhaps that's why constitutional scholars seem to be placing more emphasis on the freedom of speech challenge. If there is some chance that the law serves some economic function other than enriching the holders of copyrights about to expire, the Supreme Court might be willing to defer to the legislature on the interpretation of the Constitution's intellectual property clause. But a freedom of speech challenge means that the government must satisfy a standard higher than existence of a theoretical possibility in order to justify the restriction on expressive activity. At the very least, the retroactive extensions would seem to fail to meet the "narrowly tailored" requirement: they grant extended monopoly protection to all works whose copyright hasn't yet expired, regardless of whether they fit in the special case Julian has described.