[mb-users] songs vs. tracks

[Daniel Bumke]

On Wed, 2007-02-21 at 12:41 +0100, Lukáš Lalinský wrote:
> On St, 2007-02-21 at 11:29 +0000, Daniel Bumke wrote:
> > Am I right in saying that tracks attached to songs would be identical
> > (as in "is a release of" AR)? That is, even a minor change such as a
> > radio edit (a 'master' in NGS terms, I guess) would be a different song?
> 
> Yes, the relations would be:
> 
> track 'Foo (radio edit)' -> work 'Foo (radio edit)' -> work 'Foo'

Ah, one concept/term, many uses. I like.

> > If this were the case, PUIDs could probably be attached to songs, rather
> > than tracks. Unless we want to use PUIDs to verify that two tracks are
> > the same song (different PUIDs could mean different audio).
> 
> I'd probably rather keep PUIDs assigned to tracks for now. Mostly
> because people can easily do mistakes and I think assigning a wrong PUID
> to ~100 tracks in one edit would be really bad.

Yes, agreed.

> > But a release still contains tracks, not songs, right?
> 
> Yes. Tracks are objects with titles, lengths and PUIDs (nothing more).
> One track be be assigned to one work/song and one release.
> 
> > Calling it 'song' might not be a good idea, if only because that is the
> > most fuzzy and broad concept in NGS whereas this is quite a specific
> > instance, closer to a master.
> 
> I actually wanted it to be fuzzy. I think Frederic's suggestion to call
> it 'work' is better. That way it can represent a compositions, remixes,
> medleys, symphonies, etc.

I see what you mean now: we can have as many layers as we like, but they
are all referred to as a 'work', differentiated by ARs.

Some examples, to see whether I really get it:

Take
http://musicbrainz.org/track/c54a4411-f1c1-4693-ad23-89093534ab9c.html
"Come Out and Play" (track) (artist: The Offspring, album: Smash) ->
"Come Out and Play" (work) -> is covered by -> "Come Out and
Play" (work)