I collect records. I rarely buy anything blindly by name or label; I like to discover new stuff by listening to it first. I usually do this by spending hours in grimy bins with a record player but I also found that sometimes Ebay sellers post sound clips of their records. Specifically, what I noticed is that sellers are more inclined to make the effort of posting a sound clip if the record is rare -- i.e. you can't find a clip elsewhere -- and the record sounds really interesting. It helps them make a sale.
So I made an Ebay API spider that gets all the sound clips and record release data. It's been running like a dream since 2007 and I currently have 89,830 sound clips of rare records (probably some dupes in there). I used to watch the feed all the time but I kept finding really cool records and was spending a lot of money! I got busy. I had a good test suite so anytime there was a Unicode error or some bug I patched it pretty quick.
I steadily accumulated a pretty amazing collection of sound clips seeded by these search terms: soul, funk, reggae, ska, country, breaks, disco, psych, afrobeat, jazz, rocksteady, garage, indie, library (as in library music), new wave, electronic, brazilian, and boogie.
I made a frontend for listening to clips all Ajaxy-like and got it working on mobile phones and major browsers. I was sort of happy with it but ran into some database bottlenecks and got busy again so I never launched it. It was too slow to be usable.
What should I do? I don't really have time or money to finish it out but I'd like people to use it to discover music. Since the clips are short and will link to music for purchase, either MP3 (if it exists) or original vinyl on Ebay/Discogs, there shouldn't be any copyright problems. It's fair use to promote music for sale with a clip.
You can comment here or reach me by email: kumar.mcmillan@gmail.com
http://archive.org/about/contact.php