It's basically an open-source Shazam back-end: you feed their server chunks of audio and it tells you what songs are in it.
So, you know what would be awesome, is to run that over the archived DNA webcasts to generate actual playlists after the fact.
Except, it seems to barely match anything, even the hits of the 80s. I was testing it on Depeche Mode's Some Great Reward, because hey, not exactly obscure, but when feeding it successive 30 second chunks of each song, it seems to get a match about one time in 50.
Here's my rough draft of the code: echoprint.pl.