Hi, my apologies if anyone gets this twice, just trying to increase coverage.. We've updated the version of TRM that's included in the MusicBrainz client library. It's now much more robust against normalized audio tracks -- so it does a lot better than the previous versions at matching random tracks downloaded from unreliable sources. =) This update's passed through our internal test suite and it's now time for more expanded, external testing before it goes live on the real MB server. So, if you're interested in helping make sure that MB's audio recognition's the best it can be, here's how to help (sorry 'bout the unix-centric directions): Grab the snapshot of the MB client library from: ftp://www.freeamp.org/pub/musicbrainz/musicbrainz-1.0.0-mbtest.tar.gz Compile and install as normal -- if you're using CVS, you can get it by checking out the sha1 branch of 'mb_client' (something like: "cvs -z3 co -r sha1 mb_client"). Then, grab and install current FreeAmp cvs.. The URL below is a current snapshot of CVS, and can be used instead: ftp://www.freeamp.org/pub/freeamp/src/freeamp_mbtest.tar.gz Using FreeAmp CVS isn't required to test the new code, it's just a bit more stable than the previous release, and creates a debugging file of the GUIDs/filename pairs that are generated during the signaturing process. Once everything's installed, you're all set to begin testing. All you need to do is enable the Relatable features as normal in FreeAmp and have it find and signature your music collection.. If you're using CVS freeamp, the file 'guid_mapping.txt' in your ~/.freeamp directory is a list of everything that's been signatured. Basically, we're mainly interested in files getting the same GUID that shouldn't be -- ie., mis-matched songs. Things that should match but don't are also of interest, but there _will_ be more of those, and those are a less important problem than things matching when they shouldn't.. It's relatively easy to scan the guid_mapping.txt file for duplicate GUIDs -- you can just import the file into a spreadsheat or whatnot and sort the lines.. If that's too much effort, you can email the guid_mapping.txt file to me ([EMAIL PROTECTED]) and I'll scan it for you.. In fact, just send me the guid_mapping.txt files, that's easier =) If any mismatched tracks are found, then it gets a bit more fun.. We'd like to get access to the first 30 seconds of both of the tracks. So, we've put up a publicly accessible upload area at ftp://www.relatable.com/incoming If people email me their guid_mapping.txt files, I'll just send out mail asking for particular tracks, etc.. We hope to have this (and the SHA1 acceptance algorithm Robert keeps mentioning) in use on the main MusicBrainz server in as little as a week, so any help anyone can give in testing will be much appreciated. Thanks, Isaac _______________________________________________ [EMAIL PROTECTED] http://www.freeamp.org/mailman/listinfo/freeamp-dev
