The basic ID3 tagging scheme is very weak for non-popular music.  For example, in the classical world, "Artist" is much more reasonably the performer, whereas the composer is relegated to "Original Artist" or a composer comment (for FLAC, OGG etc.)  In folk or world music, the performer or composer is often not known, especially for field recordings.

Amarok makes a very good attempt at making some standard extended comments easily accessible (such as composer, for example, though I'd like more - a different topic though!) on the browser screen, in playlists and so on.  However, I would like to be able to make that same information available to the audioscrobler support that returns listening information to last.fm.  Currently, last.fm has no idea of composer (or at least, not that I know of) yet that would be a better guide for taste matching than performer, which is what the artist tag contains  I don't want to loose the performer details, only to tailor the information that is returned to last.fm.

Is this possible?

Is it a good idea?

Can it be scripted or otherwise introduced as a plugin or some such?

Seems like a problem last.fm should solve.

No, I don't think so.  Last.fm takes submissions via its little web service where the submitter declares what it wants to be known as "titlle", "artist" etc.  The usage of the standard ID3 tags by any individual or more gerenarrly, by all music types does not lend itself equally to the usage made of it by last.fm.

The sort of thing I had in mind, is something like this:


submitTrack (t) IS:
    title = t.title
    album = t.album
    if t.genre in (CLASSICAL, MAYBE_OTHERS)  AND isDefined (t.composer)
        artist = t.composer
    else
        artist = t.artist

    submitToLastFm(title, album, artist


The thing is, this is a usability feature - it means my usage of tags can be adjusted to the assumptions (which are based on popular music usage in the main) made by last.fm (or generally, other consumers of that information).

Last.fm prefers the artist submitted to be the artist shown for the track at MusicBrainz when it's listed there. Submitting anything other than that artist will most likely cause long-term issues when last.fm gets even more MusicBrainz data integration done in the future.

If your files are properly MusicBrainz-tagged, then the artist in the Artist ID3 field is what should be submitted.

Interesting.

What Amarok shows me is pretty close to what I want to see.
The word "properly" is rather subjective - for one thing, for the music I have, the effect of using Musicbrainz is to put the composer in the Artist field and include the performer as a parenthesised component to the title field.  In this case, I loose many features Amarok gives me for selecting by composer, artist and title.

Secondly, I have (quick count ... 734 albums and 8450 tracks not including the ripped vinyl, live recordings and so on, mostly ogg, some flac, very, very few mp3) a reasonable collection which Amarok looks after quite well, it seems very restrictive to be tied to Musicbrainz.  I did hear that Amarok developers were looking to extend the database so that extended tag information could be held there.  If that was done, a SELECT that provides data for Musicbrainz and another for the GUI would also seem to provide the flexibility required.

Come to think of it, it would not be hard to add an additional table or two for musicbrainz tags, populate it automatically by scanning the connection and inserting musicbrainz lookup data for the same albums that are playing.  It seems a bit of duplication since I think, with the right queries on the current data model. this could already be accomodated, but how else would you know that the data in my database is the same as that in the musicbrainz database for any given album?

Finally (and off topic) I've just seachred for the first 10 albums on my current playlist - only one was found on Musicbrainz.  Some years ago, when Muscibrainz was new, I opened an account so that I could submit music information to their database, and did so.  Nonetheless, it still seems to be very small to me when compared to less accurate alternatives like freedb.