Hi,

I often download songs from Webjay. All songs I download get stored in ~/music/From_Webjay. Eventually I realize I have 8 or 10 songs by a single band in there that I\'m really quite fond of, so I move them to ~/music/Some_Band. However, in doing this, I have just reset all of their scores, because scores are associated with a filename rather than a song fingerprint or some other useful data.

I thought it might make more sense to associate the score with a song\'s fingerprint. MusicBrainz uses song fingerprints pretty successfully to identify songs, and this could work well if there are no collisions in a user\'s music collection. If not this kind of fingerprint, is there some other way that score can be associated with a song in a more robust way? Another poster suggested keeping the data in the ID3 tag, which is probably a mistake, but something along those lines would make reorganizing an MP3 collection a lot less painful.

Ethan

i also have the problem that sometimes want to move a song, and therefore getting the score resettet.
i take the way and change the path within the database since my backend is mysql and therefore changing the path via phpmyadmin is quite ok

but a fingerprint, whatever kind, would be a better solution in my eyes.

is there a special reason you habe chosen the file path?

Because there isn\'t a simple quick fingerprint unfortunately.

This discussion is also being made on bug report 99791

Suggestions for TRM (musicbrainz) or checksums are not considered an option because it would be too slow. TRM would also require collision handling. I think it\'s an interesting problem which I hope it could be solved some day...

Please see my posting on that bug report. In short we could create a migration operation with some SQL and move files and update the urls to tracks in the collection at the same time. This way we would not lose stats.

wouldn\'t a simple md5 suffice? might not need to do the whole file either if processor time is still an issue.

In fact, for 1.4-beta1/svn madpenguin8 has the file migration possibility he was talking of implemented.
MD5 sums or whatever hash would be hard to do, as you probably don\'t want to have it depend on the metadata - and thus, depending on the file format, you would have to leave out different parts of the file for check summing.