[vorbis] Vorbis Comment question

[Segher Boessenkool]

Beni Cherniavsky wrote:
> On 2003-03-03, Segher Boessenkool wrote:
>>John Morton wrote:
>>>On Mon, 03 Mar 2003 03:53, Segher Boessenkool wrote:
>>
>>A program that sorts streams on tag values is pretty useless --
>>TAGS ARE NOT GENERIC METADATA.  Sorry for shouting.
> 
> Web sites don't contain generic metadata, so Google must be pretty
> useless.  Sorry for exagerating your position :-).

Google has available some hugely more structured data than
Vorbis tags are.  You'd need some pretty strong AI to sort
streams in a meaningful way using only tag values.

>>Programs that mangle tag values are retarded.
> 
> Except when you know what you are doing and want the program's
> mangling happens to fit your needs.

It would need AI here, too.

>>>And I quite like the way that vorbisgain calculates and stored the replaygain
>>>values for tracks and albums, and various players can put those numbers to
>>>use, automatically.
>>
>>Info like the replaygain values belong with user preferences in
>>a player program, not with the stream.
> 
> ReplayGain is actually a deterministic function of the stream (at
> least the per-album value); storing it in the tags is actually a dirty
> hackish way to cache it but it works and creates few problems.

True -- although I think of replaygain values as a user preference,
not an attribute of the stream /an sich/.

>>>I disagree - I like being able to see my pick of the artist, title, track
>>>number and album in my media player. I think the tag=value model is just the
>>>right level of machine-readable formalization for the task.
>>
> Let me put it this way: the vorbis tag model is a quick hack to impose
> minimal conventions onto the "quick notes".

True.  And it has been useful, and will be useful, until we get
a real metadata format.  But we should *not* try to put further
restrictions on tag formatting/contents to try to use it for more
formal purposes; instead, that generic metadata format should be
developed.

 > Creating real generic
> metadata and doing it right is awfully hard; 99% percent of the people
> would never care to enter the data correctly for such a system.

That's why they should download the metadata off some centralized
service.

 > Yet
> most of them want something that helps them extract information from
> these notes automatically, even if it only a heuristic that works in
> 70% of the cases.

That's completely fine with me -- but don't try to do anything
to increase that success rate -- that will only hide the real
problem, causing it to not be solved any time soon.

 > If it weren't for these minimal conventions, it
> would harly work in 10%, when you mix tags written by different
> people.

True.  This ARTIST= etc. bla was a nice stepping stone to at least
get functionality equivalent to ID3 into Vorbis, until something
better will be developed.  As it is now, Vorbis tags already is
better than ID3; but if we put further restrictions on the tags,
it will become *worse* than ID3.

> When you want to store the composers for many tracks in your computer,
> the freedom to write "Composer: Bach", "COMPOSER=Bach" or "Composed by
> Bach" gives you little benefit; agreeing on one of them hurts little
> and gives practical value.

It misuses the tags for something more formal.  I don't want to
think about tag names when I jot a note.

> Again, it's just a quick hack, a heuristic, like using grep on English
> text files - its imperfect but does the job pretty well, especially
> when you only have this text.

Da.

<p>Segher

<p>--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.