This dovetails so nicely with Jimbo's post on the Italian kidnapping/rendition case: The Joint Special Operations University's publication on "manhunting." This is not just targeted killing or detention, but encompasses a range of behaviors (see p. 37).
The author is right to point out that we do this already in the current conflict zones (what he calls "permissive" environments, i.e., where we have permission from the host nation and can operate freely). He's right to point out that these kinds of missions are an important part of keeping up the pressure on insurgent/terrorist groups. Having a dedicated, interagency group focused on this might be a worthy idea, and he offers some useful suggestions on such issues as the ideal size for such teams, and some useful historic parallels (I appreciated the bit about Daniel Boone).
The key problem really is the one that the Italian case brings up: how do you handle the framework for a larger-scale, permanent version of this kind of operation? How do you keep these operations legitimate, in other words, so that they aren't interpreted as illegal kidnapping?
The discussion of that issue begins on page 14, but it's only a sketch of the problem. (There is also some errata; for example, he mentions privateering as evidence that the US can lawfully license seizures, though privateers were generally empowered to wage war against enemy nations' property rather than to carry out attacks on people; further, the treaty he mentions as banning it is one to which the US was not a party, and indeed, American privateers have been used twice since then, by the Confederacy during the Civil War, and -- oddly enough -- to issue a letter of marque to a Goodyear blimp for anti-submarine operations in WWII).
In addition, his treatment appears to elide past the issue of guilt v. innocence; it's perfectly correct to say that a member of a terrorist group has set himself outside the norms of human behavior, and ought to be liable to killing, but we have to have clear standards for proving that our intelligence on them is correct (and not just the word of rival tribes or factions who would like to use us as a weapon against their enemies; or worse yet, planted by the enemy itself to cause us to drive a wedge between ourselves and a tribe that might be swinging our way).
What really needs to be done is to generate a clear framework for these operations that is acceptable at least to key allies like Australia and the UK -- the ones who reliably support us, and therefore must not have their political will to do so undermined. Ideally, we could bring in other nations that might wish to conduct similar operations, including at least France, and possibly China and Russia.
The model for this would be the conventions on piracy, which essentially give all nations the right to deal with pirates anywhere. As a way of handling ungoverned or difficult-to-govern spaces, at least, it makes perfect sense to treat terrorist groups in the same way.
How you handle snatching people off Italian streets, however, will require more thought in the future. The CIA is explictly licensed to break foreign laws, so long as it doesn't break American ones. That license is necessary in some times and for some problems, but you have to consider carefully how far you want to extend it (or, as Froggy pointed out in the comments to the piece below, if the CIA is still the right place to locate that license).