[mb-style] What real-world entity does a Track represent?

[Jim DeLaHunt]

Olivier: 

Thanks for your response. It's taken me a while to reply.


Olivier-10 wrote:
> 
> 2008/1/14, Jim DeLaHunt <from.nabble at jdlh.com>:
>> Where this definition makes a difference for recent discussion is in what
>> ARs relating to Tracks mean. It's easy to see that a "Performed By"
>> relationship between artist A and the Track is meaningful.
> 
> It depends. Sometimes it's meaningless. For example, if a unique
> performance for some reason is splitted into multiple different
> tracks, I definitely think it makes more sense to link the artist to
> the release rather than to all individual "tracks".
> 

I don't follow your logic.  If, for example, a performance is recorded in
one take, but is published split into Track 1, Track 2, and Track 3, and if
Artist A could be heard during the parts of the recording published as Track
1 and Track 3, but was silent during the part published as Track 2, then I
believe there should be a Performed By AR with Track 1 and Track 3, but not
Track 2.  It seems it would be less precise to link a Performed By AR with
the Release.


Olivier-10 wrote:
> 
> You may argue that this is due to current db limitations, but that
> doesn't really help - a track is not necessarily equivalent to a
> specific work, or even to a specific performance, as you apparently
> (from above) want it to be.
> 

You have a good point.  I don't think a Track is equivalent to all of a
specific work, or limited to just one work. Nor is it limited to all of a
specific performance.   This may mean that the definition of Track needs to
be more precise. Maybe it should say that it refers to a "part or whole" of
a specific audio recording.  

I think it will be hard to get the definition right.



Olivier-10 wrote:
> 
> 2008/1/14, Jim DeLaHunt <from.nabble at jdlh.com>:
>> It seems to me
>> that a "Liner Notes written by" relationship between artist W and the
>> audio
>> recording of a performance is not so meaningful -- the liner notes are
>> part
>> of the Release, not the Track.
> 
> The tracks titles themselves are always part of releases, not tracks -
> obviously, the same audio recording may have different titles
> depending on the release they are on...
> 

Good point!  According to http://wiki.musicbrainz.org/DatabaseSchema/Tracks
(which may not be current), the track title is stored in the track table,
not the albumjoin table.  This means that if two Releases were ever to point
to the same Track, then they would have the same Track title.  As you point
out, different Releases could put different names on the same Track.



Olivier-10 wrote:
> 
> So, going by that train of thoughts, nothing actually relates to
> tracks... and I don't think this is really going to be helpful...
> 

I disagree.  Those relations marked "Type 1" in the list of ARs at
http://wiki.musicbrainz.org/BarryPlatt all relate to "a part or whole of a
specific audio recording of a specific musical performance". e.g.
"{additionally} {guest} performed", "{additionally} conducted", "has/had
{additional} {instrument} samples taken from", etc.



Olivier-10 wrote:
> 
> Back to your example:
> What if a single track is released over the internet with liner notes
> in the comment tag?
> 

If we accept that the Track is "a part or whole of a specific audio
recording of a specific musical performance. A particular track is
transformed to various recorded data formats for distribution to listeners",
then then that music file you mention, containing both a transformed copy of
the recording and some liner notes, is a Release, with one Track.  The liner
notes are still a property of the Release, not of the audio recording.


Olivier-10 wrote:
> 
> Or not even going that far, what if a release has different liner
> notes each by different artists, one for each track?
> 

I would argue that such a Release has should have an AR for each author of
each liner note, all attached to the Release record rather than any Track
record.

Suppose a CD has two or three liner notes, and nine or ten tracks?  Would
you attach ARs describing those liner notes to Track records?  Which ones
and why?

Suppose that one of the liner notes talks about the artist's biography, not
about the specific recording?  Would you attach ARs describing this liner
note to Track records?   Which ones and why?

Suppose that some of the same exact master recordings are release on a
budget CD which has no liner notes?  If MusicBrainz could represent track
masters, should the Release records for the original CD and the budget CD
point to the same track master record?  If so, and if there was an AR
describing liner notes from the original CD attached to Track records,
wouldn't that AR now appear to describe the Track on the budget CD? Wouldn't
that be incorrect?



Olivier-10 wrote:
> 
> Just because almost every piece of metadata is release dependent (just
> like track titles are) doesn't mean they ought to *link to release*
> only...
> 

I disagree.  I think MusicBrainz describes the music and published works
better if those pieces of metadata that describe the Release should be
attached to the Release record.  

Anyone else have any comments?

-----
     -- http://jdlh.com/ Jim DeLaHunt , Vancouver, Canada  • 
http://wiki.musicbrainz.org/JimDeLaHunt

-- 
View this message in context: http://www.nabble.com/What-real-world-entity-does-a-Track-represent--tp14814457s2885p14943865.html
Sent from the Musicbrainz - Style mailing list archive at Nabble.com.