[vorbis] Vorbis Comment question

[Segher Boessenkool]

Benjamin Weste Pearre wrote:
> I know that some of the developers share that view.  So sorry, but
> asking that no attempt be made to make the tags machine-readable is
> simply stupid:
> 
> * These files will live on computers.  Computers need something
>   parseable.

The only thing computer programs are allowed to do with tags are
displaying the tags to users, and letting users edit the tags.

 >   The paragraph you quoted requests that no attempt be
>   made to make the comment fields useful to a machine - in other
>   words, the files are on your computer, but you are essentially
>   forbidden from writing any program that uses them.

Tag editors and tag display programs are allowed.

 >   Even if you
>   don't share your music, you might be inclined to share a program
>   that manages music, but the "intention" of the spec demands that any
>   such program have a good natural-language parser, which puts it far
>   beyond the cutting edge of machine language research.

Tag data is associated strictly only with the actual Vorbis
stream, not with the original recording, nor with any data
a user interface might need/want.

> * The as yet non-existant metadata stream has been non-existant for a
>   very long time, and we need something _now_, or else Vorbis is
>   simply not ready for release as anything but a beta.

Nonsense.  MPEG Audio never had any "official" metadata defined;
and ID3 (old) was too restricted (so you need an AI to parse that,
as people misuse all fields of it at will) and ID3 (new) is too
complex to be used.  Still MPEG Audio is not in beta anymore ;)

Free-form tag comments are better than anything ID3 ever did,
from a useability standpoint.

> * Will the metadata stream have piece titles, authors, performers,
>   etc?

"We can't comment on unreleased products".  :-)

 >   If so, why dupliclate all of this information?  That's

All 200 bytes of it?

>   inefficient.  If not, then programs are essentially forbidden from
>   having any code that can manipulate the music according to any rules
>   a human would use - say, sorting your music by composer.

That is a task best left to a database -- and Vorbis is not
a database format.  You don't want to have your player program
open and read all your 10000+ music files just to display a
sorted list of composers of music you have Vorbis streams of.

> * When the as yet non-existant metadata stream is finally implemented,
>   if it will in fact contain "composer", "performer", etc, then any
>   already-encoded vorbis files will have to be painstakingly
>   hand-edited, because the information that we need to extract has
>   been formatted without machine-parseability in mind.

No, you will get it from MusicBrainz (or the like), and in the
rare case it doesn't have it yet, you can put it in (for the
benefit of all mankind ;) ).

> Now, the paragraph you quoted justifies all of the points above.  It
> also basically justifies only one comment field: "COMMENT" - a
> free-form paragraph listing composers, CD numbers, dates, etc (to a
> human, "COMPOSER=Bach" is equivalent to "Written by Bach").

I actually think all of this tag business has been a mistake --
comments should be truly free-form.

Unfortunately an =-sign is required in every comment now.

 > But if
> you feel like listening to Bach and want to see what you have, the
> latter is about as useful as a stack of file cards - even though we're
> on a computer, we're not allowed to take advantage of a computer's
> ability to manage information, but need a human to look through each
> file and see who the composer is.

Your music file manager program should help you here.

Also, in my experience, everyone with a (legal) music collection
big enough so that browsing through it would take a significant
amount of time, *knows* exactly what music he has -- no need to
browse at all.

> HOWEVER, the actual comment field spec is far, far better than that
> paragraph would suggest - it has well-defined tags, and the syntax for
> unknown tags is likewise specified.  In fact, I think that it's pretty
> well done (especially the proposed list at
> http://reactor-core.org/ogg-tag-recommendations.html): it leaves the
> tags very readable by humans, while making it possible for programs to
> manipulate them.  In other words, it's useful.

Not any more useful than completely free-form comments would
have been, just very annoying to the user -- he now has to
think about tag names while all he wants to do is jotting down
some notes, or copying what's on the record sleeve.

> It's trivial to teach a human how to read well-designed computer-
> readable text,

Ah great, back to the early 60's.

 > and currently impossible to teach a computer how to
> understand human language.

Fortunately, no program has any business trying to understand
what those comments mean, they just have to display and/or edit
the tags.

 > If you believe that it might sometime be
> useful for a computer to be able to help you search for a recording
> you're looking for, then you must accept some level of machine-
> readability in the comment fields.

Or stop thinking in terms of not-well-thought-out player programs.

> Stop claiming that any argument for machine-readability is Not What We
> Believe In!

Because you bring up religion, let me state once more the
constituents of the "Trinity Of Media Meta-data":

        1) Attributes of the recording /an sich/;
        2) attributes of an owner's copy of the recording;
        3) attributes of the owner's use of the recording.

<p>Also, there are three places meta-data are stored:

        a) In the stream;
        b) in a public database ("The Great Library");
        c) in a local database.

Possible combinations:

1+a is Draycott (Ogg metadata).
1+b is MusicBrainz etc.
1+c is possibly a cached copy of 1+b.
2+a is Vorbis comments.
2+b is irrational.
2+c is useful for some programs.
3+a is not good, as reading tags [in many streams at once] is slow.
3+b is irrational.
3+c is useful for many programs.

<p>> That's way more than I wanted to say, but it just all seemed to belong
> together.  Have fun :)

Much more than I intended to ever again write about tags, too.

<p>You have fun too,

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.