Weird definition?

Hi, Yves! We do plan to add contributions to tracks as well as to albums. However, specifically for compositions, we have a Composition type that would handle that; you would have composed an instance of a Composition, and various Musical Tracks would be recordings of that composition.

Performance contributions (such as conductors) would then be a little easier to track; sometimes, the same recording ends up as multiple tracks (alternate versions on box sets, e.g., but that’s a relatively rare case and I think it’s not too much trouble to capture the performance contributions separately for those.

Hi Crism!

Just a quick comment re. "the same recording ends up as multiple tracks" - well, that happens a lot! (for example singles/albums, which often have at least one track in common).

Also I think you should capture the notion of Performance, between a Composition and a Track, exactly for this sort of reason?

Anyway, we have a mailing list for this sort of discussion - it would be great if you could join us!
http://groups.google.com/group/music-ontology-specification-group

So far, the data model that raised from discussions there is something like:
Composition (---> Arrangement) --> Performance --> Recording ---> Track, record, etc.

(btw, have you heard about the linking open data project? there is a big music-related part in it, and we already interlinked quite a few datasets - Jamendo, Musibrainz, Magnatune, DBPedia, John Peel sessions, etc. etc.)

Cheers!
y

Yves,

By multiple tracks I meant substantially different audio artifacts. If the single is the same as the album track, then it should be the exact same object in Freebase. If, however, the single contains a shortened version, it should be a different track.

We’re not trying to replace MusicBrainz, so I am not sure that capturing recording sessions (i.e., performances) is necessarily the right thing. MusicBrainz’ next-generation schema should be capturing that level of detail; we will have to see if our community demands it from us as well. (I’ll put you down for one vote yea. (-:)

We have Composition, Arrangement, and Musical Track, flattening the intervening layers. We could also have relationships between tracks as needed for alternate takes, radio edits, etc. (analogous to MusicBrainz’ “advanced relationships”). Do you think that would suffice?

Do you have a link to the “linking open data project”?