It should be noted, however, that "the use of language X" has in fact been frequently legislated against, usually in an attempt to banish ethnicity and enforce some goal of social uniformity. And such laws do tend to be strongly enforcible, even though they often don't succeed in totally eradicating the banned language, which may re-emerge when the ban is lifted. But forbidding use of a language in official business, schools, and in public or among strangers is very effective at suppressing it.
Similarly, one might make the devil's advocate argument that banning "traffic which looks like BitTorrent" is certainly possible, especially as it lowers the bar for prosecutions from "being certain infringing material is being exchanged" to simply "being certain the protocol is BitTorrent". And, like the "use of language X" case, this can effectively shut down "talk among strangers" and the use of the protocol "in public" (ie servers)—not to mention "official use" (ie for legitimate content). (Note that open peer-to-peer systems require "talking to strangers", who you can't be sure aren't going to rat you out for using a banned protocol—even if you wrap the connections in SSL or some such to prevent third-party snoops. It is possible to use anonymizing techniques to allow "talking to strangers" without revealing your identity—but then we can legislate against "wearing masks in public" as many places do; ie ban the "anonymity" protocol.)
One should note that a difference between BitTorrent and, say, Norwegian, is that language is fairly difficult to change, while software updates to change a protocol might be easier. The users would still be playing a risky cat-and-mouse game with the law, though: I'm fairly certain a judge would rule that "BitTorrent over port 80" was still a use of the (hypothetically banned) BitTorrent protocol. Likewise for most other 'trivial' changes; one might also consider banning whatever "update protocol" allowed a popular client to adapt to legislative fiat.
I agree with Prof. Felten's main hypothesis—going after the protocol is more like whack-a-mole than it is a sound legislative strategy—but I don't think the whack-a-mole alternative is so obviously impratical that it won't get legislative consideration.
Consider, for a moment, if legislation established a list of 'banned protocols', with an administrator in charge of adding to this list (ie not requiring an act of congress). It takes time for a user community to change protocols. Even if it took 6 months (say) for a "new" peer-to-peer protocol to be "discovered" and added to the ban, it might still prove very effective in limiting the community size (and thus use and usefulness) of peer-to-peer technologies. It would also strongly hurt legitimate uses of peer-to-peer technologies (since the protocol is banned regardless of content), which is something we might worry about but the legislators (and their MPAA/RIAA/etc lobbyists) might regard as a "necessary consequence of fighting crime".