My issue is with the relation attributes. There is any number of attributes that differentiate two different recordings of tracks, see [A] for a partial list. Why we should select just "live" and "acoustic" is rather mysterious to me.
[A] http://wiki.musicbrainz.org/ClarifyExtraTitleInformation On top of that, like you said, there are covers and medleys, and also parodies, translations, karaoke tracks, and I'm sure quite some other things we won't ever think about (on an audio CD, a rendering of a fan-sequenced MIDI of a song is a cover?) Also, rather funny things happen (semantically) when the initial version has one of the attributes. For example we can deduce from a "later version" relation with an "acoustic" attribute that the later version is acoustic, but there's no AR way to deduce that the early version was acoustic too; this is an annoying asymetry, especially since these ARs are intended primarily for automated interpretation. The only solution I see is that we should make a clear separation between attributes that are relevant to the relation and those that are relevant to the tracks: acoustic or live is a track attribute, cover or parody or translated is an attribute of the re-releasing process; a simple way to see if an attribute is a relation attribute is if it is meaningless when there is a single release of a song: live and acoustic still make sense, but cover or translation don't (BUT, _language_ would, if we would store it). I propose we do just that: remove "live" and "acoustic" from the AR. If we can, in the future, we may add such attributes to the tracks themselves in a machine-readable way (we retain mix names and version info as text now). If we keep/add attributes for the "later/earlier recording" relationship, I propose we only keep/add those that pass by test above. Examples of such attributes are "cover", "translation", "partial", "parody", "medley" (though this depends on how exactly we define medley). This is not as straightforward as it seemed (to me) at first: *) some of these attributes can be automatically added (some now, some in the future): for example, cover (now: different artist) and translation (later: different track languages). *) but it's not that easy: if "Iron Maiden" re-records, say, Bruce Dickinson's "Chemical Wedding", is it a cover? After all, Bruce is still on it... *) other are not really attributes: a parody is not really a re-recording, and you could argue neither is a translation, but they are very closely-related relationships. And what happens when someone releases a track that has the same music but totally unrelated lyrics with a song, or the other way around? Has the possibility of building a real ontology been discussed before? You know, OWL and RDF stuff? -- Bogdan Butnaru — [EMAIL PROTECTED] "I think I am a fallen star, I should wish on myself." – O. On 4/23/06, Simon Reinhardt <[EMAIL PROTECTED]> wrote: > Hi, > > as the title says, I would like to change > http://wiki.musicbrainz.org/OtherVersionRelationshipType to what it really is > about: re-recordings. [...] > I already proposed new link phrases and additional attributes on the page: > - track has later {acoustic} {live} recording(s) track, > - track is a later {acoustic} {live} recording of track [...] > Comments? _______________________________________________ Musicbrainz-style mailing list Musicbrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
