Parsability requirements of PART tag (Was: Re: [vorbis] Vorbis Comment question)

[Benjamin Weste Pearre]

On Thu, Jul 03, 2003 at 08:26:42PM -0700, Jonathan Walther wrote:
> On Thu, Mar 13, 2003 at 10:36:47AM -0800, David K. Gasaway wrote:
> >Note: I dislike the parsability requirements for the PART tag spelled 
> >out at reactor-core.
> 
> I didn't realize there were any parsability requirements for the PART
> tag.  Could you refresh my memory, and tell me what you would like to be
> different?  Any problems are probably artifacts of the original first
> version, and didn't get hashed out adequately before discussion of other
> aspects led to it being forgotten.
> 
> Jonathan

The parsability requirements are these:

Many pieces of music have different parts: the movements of a string
quartet (often about 4), the bits of an opera (usually about 40), etc.
Note that since often there will be two symphonies on a CD, or one
opera on 3 CDs, TRACKNUMBER does not effectively communicate what
order to play things in.  Moreover, tracknumbers are an artifact of
CDs that we don't want to take with us from CD to pure information.

Here's an example: please put the following four tracks in the order
that will actually make sense musically:

Verdi - Requiem - Dies Irae.ogg
Verdi - Requiem - Libera Me.ogg
Verdi - Requiem - Sanctus.ogg
Verdi - Requiem - Requiem & Kyrie.ogg
Verdi - Requiem - Agnus Dei.ogg
Verdi - Requiem - Lux Aeterna.ogg
Verdi - Requiem - Offertory.ogg

Here, we have something like

COMPOSER=Verdi
PIECE=Requiem
PART=Sanctus

or something like that.

Now try this:

PART=(04) Sanctus

...or...
PART=Sanctus
PARTNUMBER=4

Note that on the source CD, the tracks (while in the right order)
started at 4 and went up to 10 (PART=(07) Sanctus, etc).  It would
have been more fun if I'd used a multi-disk example, since then there
would have been no useful order information at all in the track
number.

Since building knowledge of the correct order of the parts of a
requiem into every Ogg player doesn't make a whole lot of sense, I
suggested that there be a tag to represent the order that the tracks
should come in.  For some reason that I still don't understand, I was
shot down.  However, they graciously compromised and wrote that
gobbledygook that basically says "The PART tag is where you hack in
information about what order the tracks should be in".

I still think that it was the wrong decision, but I don't feel like
going up against reactor-core, since I think that a bad standard is
better than none (=too many).

One possibility is to usurp TRACKNUMBER.  This is useful because
tracks are mostly irrelevant once we've left the realm of the CD.
However, it is not useful (ie player software can't try to use it)
unless a majority of people agree to it.

The argument against machine-parseable tags is something like "the
metadata will handle that; tags are just for people and shouldn't be
restricted".  Since there is no metadata yet, this argument translates
to "It is not yet the software's responsibility to try to put a bunch
of pieces into the right order", which means that everyone who needs
this feature (most anyone who cares about classical music or Pink
Floyd) will need to choose between the awkward PART hack and some
personally-devised solution.  So much for standards...

Oops - I didn't mean to write all that, but it's a good, um, what's
that word... "unbiased" summary for newcomers!  Oh, and did I mention
that the infidel American troops are fleeing from our righteous wrath?

Cheers :)
-Ben


-- 
Ben Pearre         http://hebb.mit.edu/~ben        PGP ID: CFDA6CDA
Stop sending insecure email!                   http://www.gnupg.org


-------------- next part --------------
A non-text attachment was scrubbed...
Name: part
Type: application/pgp-signature
Size: 190 bytes
Desc: not available
Url : http://lists.xiph.org/pipermail/vorbis/attachments/20030704/49f32c95/part.pgp