More Like, does it work? (at least, as advertised)
I'm guessing their "Acoustic fingerprint" spiel is BS, pure and simple.
Assembling enough of a file from a datastream to then fingerprint any audio within it and then run checks against a database of known fingerprints (and just how many song fingerprints will this database contain exactly?) all takes CPU, memory etc, and I'll bet they don't have it.
I suspect what they'll really be doing if they are just looking for music, will be based on exposed filenames, traffic to-fro known havens of illegal filesharing (hello rapidshare et al!) and looking for and decoding any tag info found within any intercepted datastreams.
I'm doubly calling it BS, as I'm bloody certain this DPI is going to be used for more than trying to guess what percentage of their p2p traffic is illegal music downloads as there are other ways of guestimating this figure. (Hint: bet certain pdf files I've seen in the wilds of p2pdom will be flagged up for the attention of the Spook squad goons behind the running of this if seen in transit on the network..especially in those areas of the country with a large Muslim population)
The music angle is a smokescreen.
Encrypting the source?, yes, there are a number of Eastern European sites where the torrents pointed to are password protected/encrypted (passwords lurking within the Cyrillic on the websites), but bear in mind that as the *majority* of the users of p2p networks are still pretty clueless when it comes to things like encryption and passwords, there's the issue of distributing the keys/passwords in such a manner as to make it usable for them with minimal pain for everyone (less users , less seeds, slower downloads).
It would be nice if everyone used encryption and password protection on archives and not-so-bloody obvious file names, makes indexing a wee bit harder, but makes snooping harder still.