This afternoon I got a call referring me to a guy in Louisiana who wants to build Linux jukeboxes. Not something to put in your living room, but a replacement for the CD-containing jukeboxes in bars and clubs. Two hours later people in four states were discussing the project, particularly the non-technical complications (USC 17, state licensing, BMI/ASCAP, and ... the power of the Mob in Louisiana). Nobody's pretending it's a particularly novel idea (we had a fairly novel set of designs and business applications sketched out over a year ago, and we were hardly the first), but the guy on the ground seems to be a real underdog (without the "...get some VC hire 300 people to run the website and ..." mentality), so we're taking a look at it again.Along those lines (automating weird extraction and classification tasks), I keep coming across whizbang! labs (they also have a research site...). Ullman's advising them. The guy who wrote WordPerfect is in deep. They run flipdog.com, have a particular fancy for spidering my home page, and seem to cite a lot of my old co-workers in their research papers. I like some of their ideas so I think I'm going to implement those I can deduce from their datasets and papers, then wait and see if anybody comes out of the shadows to throw a lawsuit at me :-)
Noticed that musicbrainz is doing a cddb replacement (I knew they were doing a metadata database but I didn't put 2 + 2 together I suppose). Unfortunately their idea of "P2P" involves people being the peers as they review album entries by hand. I've had a submarine project in the works for a while now which will be a direct "competitor" to musicbrainz from what I can tell, but which will be automated. I always assumed p2p peers did computation. Then again I always assumed people who said "p2p" too much were marketroids.