Pandora's Jar - distributed quality-assurance of song names

[ColdFusion]

The post that was here is old news, for the all-new-and-improved version see:

http://www.hak5.org/forums/viewtopic.php?p=51318#51318

********Here's what was here before:*****************

Notice:

If you wish to get a sneek preview of the latest stable build or help out in any way - please PM me and I'll send you a link to the jar and sources.

-----------------

Hi guys,

There were two things that annoyed me:

1) Some files are saved under the wrong name (and with the wrong id3-tag obviously)

2) I'd like to be able to replay files from the mp3 archive that have similar fetures, as if I'm listening to a station on pandora

the obvious solution to #2 would be to tag the files with the "station name", but since that is not available - I've done the second best thing I could think of - I'm appending the file name to a playlists - selected from the web-page, and whenever I change stations I also select an appropriate playlist to append. when listening to "cole porter radio" or "glenn miller radio" I select "jazz.m3u" playlist, and when listening to "The Corrs Radio" I select "theCorrsRadio.m3u" and so on.

This way, later, when offline - I can play the playlist 'corrs' (as is or in random order) and be sure I'm not going to get any jazz songs.

now back to the more important feature, It is extremely annoying that I start "Norah Jones/Norah Jones-Come Away With Me.mp3" and get Frank Sinatra's New York New York... as this issue cannot be solved deterministically with the current way the application operates - I've added a feature that calculated the MD5-sum of every file downloaded, and "votes"on it in a central repository, only if the artist/title/md5 combination submitted is the CURRENT highest ranked match for this md5-sum the application saves the file into the mp3 archive directory, otherwise it saves it under a "mismatches" directory - for later inspection.

I'm also working on an application that will scan the mp3 archive and mismatches directory and will fix any discrepancies.

This is right now at final QA stages, and running pretty smoothly.

From experiments about 2% of the downloaded songs are named incorrectly, with my solution - this can be reduced to 0%.