2010/4/4 Brian Schweitzer <brian.brianschweit...@gmail.com> > > > On Sat, Apr 3, 2010 at 6:10 PM, Frederic Da Vitoria > <davito...@gmail.com>wrote: > >> 2010/4/3 Philip Jägenstedt <phi...@foolip.org> >> >>> On Sun, Apr 4, 2010 at 03:17, Brian Schweitzer >>> <brian.brianschweit...@gmail.com> wrote: >>> > >>> > >>> > On Sat, Apr 3, 2010 at 3:15 PM, Philip Jägenstedt <phi...@foolip.org> >>> wrote: >>> >> >>> >> On Sun, Apr 4, 2010 at 02:14, Frederic Da Vitoria < >>> davito...@gmail.com> >>> >> wrote: >>> >> > 2010/4/3 Philip Jägenstedt <phi...@foolip.org> >>> >> >> >>> >> >> On Sat, Apr 3, 2010 at 14:31, Brian Schweitzer >>> >> >> <brian.brianschweit...@gmail.com> wrote: >>> >> >> > Note: I'm going on the working assumption that, however we define >>> >> >> > works, >>> >> >> > the >>> >> >> > 'artist' for a work will be the principle (ie, non-'additional') >>> >> >> > composer(s), rather than the performer, lyricist, >>> lyricist+composer >>> >> >> > duo, >>> >> >> > etc. >>> >> >> > >>> >> >> > I'm in the midst of cleaning up the text and wiki-formatting of >>> the >>> >> >> > CSGv2 >>> >> >> > proposal. One works problem occurred to me today that we'll end >>> up >>> >> >> > running >>> >> >> > into, likely sooner rather than later: spurrious works. >>> >> >> > >>> >> >> > These are the works that for whatever reasons are (or have been) >>> >> >> > frequently >>> >> >> > attributed to one composer, but were actually composed by someone >>> >> >> > else. >>> >> >> > It >>> >> >> > happens all the time in classical, but I've seen it even for >>> modern >>> >> >> > music >>> >> >> > (Kurt Cobain and Trent Reznor, to name just two). For anyone >>> who's >>> >> >> > not >>> >> >> > run >>> >> >> > into these in classical before, Crazee canuck had started a list >>> of >>> >> >> > spurrious Mozart compositions. It's far from complete, but you >>> can >>> >> >> > get >>> >> >> > some >>> >> >> > idea from it: >>> >> >> > http://wiki.musicbrainz.org/CSG_Standard/Mozart/Spurious . >>> >> >> > >>> >> >> > Now, if a work is known to be spurrious, it's already a >>> >> >> > cat-corner-ish >>> >> >> > case, >>> >> >> > musicologically, let alone to us. But in almost every case, the >>> >> >> > composer is >>> >> >> > known or suspected - for those that aren't, we can always use >>> >> >> > [unknown]. >>> >> >> > >>> >> >> > But how then to link that back to the 'often attributed' artist, >>> and >>> >> >> > what >>> >> >> > about when the correct artist isn't 100%? >>> >> >> > >>> >> >> > It seems that work-artist ARs could help here. What about >>> something >>> >> >> > like >>> >> >> > this: >>> >> >> > >>> >> >> > 1) For spurrious works, they should be listed under the correct >>> >> >> > composer, >>> >> >> > not the 'frequently attributed' one. If the composer isn't 100% >>> >> >> > known, >>> >> >> > or >>> >> >> > if there are multiple possible correct composers, then the most >>> >> >> > likely >>> >> >> > correct one should have the work listing. >>> >> >> > >>> >> >> > 2) Add a new work-artist AR. This is a bit wierd, as it's >>> defining a >>> >> >> > non-relationship, but something like 'Artist was spurriously >>> >> >> > attributed, >>> >> >> > and >>> >> >> > not the composer of Work'? (the wording here could use some >>> >> >> > improvement). >>> >> >> > This way, someone looking at the ARs for an artist would still >>> see >>> >> >> > that >>> >> >> > attribution, and thus hopefully not re-add an incorrect Work >>> listing. >>> >> >> > Perhaps in future we could even get a "Spurrious Works" listing >>> added >>> >> >> > to >>> >> >> > an >>> >> >> > artist's Works listing, which could use this hypothetical AR to >>> >> >> > populate >>> >> >> > the >>> >> >> > list. >>> >> >> > >>> >> >> > Annotations could be suggested here instead, but those won't stay >>> >> >> > up-to-date, for both (or more) involved composers, if the Work's >>> >> >> > title >>> >> >> > and >>> >> >> > or composer gets updated, and honestly, I don't see them really >>> being >>> >> >> > updated now, for the few similar 'misidentification' lists which >>> have >>> >> >> > been >>> >> >> > created in existing artist annotations. (It also could easily >>> >> >> > explode >>> >> >> > the >>> >> >> > size of the annotations for some artists; just a single line per >>> work >>> >> >> > could >>> >> >> > easily add 100 or more lines to Mozart's annotation.) >>> >> >> > >>> >> >> > 3) Add 2 new attributes to the composer artist-work AR, >>> 'possibly' >>> >> >> > and >>> >> >> > 'very >>> >> >> > likely'. This would allow some granularity when we assign a >>> composer >>> >> >> > AR >>> >> >> > to >>> >> >> > a work. 'Possibly' to cover those cases where there's multiple >>> >> >> > possible >>> >> >> > correct composers, and 'very likely' for when there's one other >>> >> >> > likely >>> >> >> > composer, but not yet any definitive evidence to make that >>> composer >>> >> >> > attribution 100% definite. Same note here re: annotations. >>> >> >> > >>> >> >> > (There might even be a reason for the third level of granularity >>> in >>> >> >> > #3 I >>> >> >> > typically see used in classical; 'doubtful' - useful for when >>> there's >>> >> >> > no >>> >> >> > alternate composers really worth mentioning as possible >>> composers, >>> >> >> > but >>> >> >> > there's also good reason to believe that the 'frequently >>> attributed' >>> >> >> > composer wasn't actually the one who composed the work.) >>> >> >> >>> >> >> Why associate works with an artist at all, other than through ARs? >>> The >>> >> >> UI side of things could simply show any work with a composition or >>> >> >> lyricist AR under their works, and any other creative ARs there may >>> >> >> be. This would avoid the whole problem, and seems quite natural. >>> Are >>> >> >> there any real reasons for requiring a main artist for a work, >>> other >>> >> >> than perhaps the fact that we use a relational database which isn't >>> so >>> >> >> well suited to representing it in the natural way? >>> >> > >>> >> > I don't understand your comment. AFAIK, the only association between >>> >> > Works >>> >> > and Artists in NGS will be ARs. Did I miss something? I agree >>> applying a >>> >> > generic Artist concept to works would be a mistake. >>> >> >>> >> On Sat, Apr 3, 2010 at 14:31, Brian Schweitzer >>> >> <brian.brianschweit...@gmail.com> wrote: >>> >> > Note: I'm going on the working assumption that, however we define >>> works, >>> >> > the >>> >> > 'artist' for a work will be the principle (ie, non-'additional') >>> >> > composer(s), rather than the performer, lyricist, lyricist+composer >>> duo, >>> >> > etc. >>> >> >>> >> What is the 'artist' for a work being talked about? If we only use >>> >> ARs, how could the problem being discussed exist? >>> >> >>> > >>> > Works do have an artist (or rather, artist credit) associated with >>> them. >>> > I'm not sure what the mistake intended by "I agree applying a generic >>> Artist >>> > concept to works would be a mistake." would be... >>> >>> Why have an associated artist? Wouldn't ARs be enough, and avoid >>> having to try to define which artist a work really "belongs" to? >>> >> >> It must be getting too late, but I still don't understand. But now I am >> not sure who I am not understanding. Here is what I (think I) understood: >> work W was believed to be composed by composer C1 until it was revealed it >> was C2 who really did it. Of course, there would be a Composer AR between W >> and C2. >> > > Correct so far. :) > > >> Brian suggested creating a new AR to link W to C1 too, which I believe is >> a good idea. >> > > Correct again, plus also the suggestion of 2 or 3 new attributes to link W > to C2. > > >> >> I suggested that to do this, we could use an attribute in the Composer AR, >> to indicate that C1 is not the real composer. >> >> I fail to see why Brian and Philip are discussing an "associated >> artist"... Is Brian's "associated artist" the same as Philip's? If possible, >> I suggest we avoid the word Artist as much as possible in this discussion, >> because this word has an unfortunately strange meaning related to it's usage >> in mp3 tags. >> > > I think there is some misunderstanding here, starting back in Frederic's > "AFAIK, the only association between Works and Artists in NGS will be ARs." > comment. As for keeping "Artist" separate, I don't think it's all that > possible (unless we switch to "Artist Credit", which is more correct but, as > a term to use for this discussion, is even more ripe for confusion). > > To clarify (I hope): > > > 1) In NGS, a Work is 'filed' under an Artist, just as Releases, Tracks, > Release Groups are now. > 2) Spurious works exist and will need to be handled. > > Re #1, my only suggestion was that the Artist of a spurious Work be the one > who is the 'real' composer (C2), or lacking that info, the one who is 'most > likely' the composer (the 'maybe' C2s). > > Re #2, entirely separate thing, that'd be where the new W-C1 AR and new > W-C2 attributes would come in. > > We have a spectrum of composers - and this is why I'd suggested 2 or 3 > attributes. In my experience, it's not just black/white "we know the right > composer" or "we don't know the right composer": > > (The rest of this email looks more complicated than it really is; it > basically writes out the various possibilities for 'doubtful', 'possibly', > and 'most likely'; this just breaks it out for an attempt at clarity.) > > 1a) we know the right composer is *not* C1 (C1 at 0%) > OR > 1b) we suspect that the right composer is *not* C1 (See 2f) > > 1a cases: > > 2a) we know that the right composer actually is C2 (100%) > OR > 2b) we're pretty definite that the right composer actually is C2, but > there's not yet enough definitive proof that we can be 100% about it. > (Let's call this one 95%.) > OR > 2c) we think that the right composer actually is C2, but there's not yet > enough circumstantial or definitive proof that we can be sure about it. > (Let's call this one 75%.) > OR > 2d) we think that the right composer actually is C2A, but it also could be > C2B - there's not yet enough circumstantial or definitive proof that we can > be sure about it either way. (Let's call this one 50%.) > OR > 2e) the right composer could be C2A, or C2B, or C2C - or maybe someone else > we don't know about - there's not enough evidence to know. (Say, 20 to 40%, > situation depending?) > OR > > 1b and 2f) it *could* really be C1, but we have good reasons to think it > might actually be C2, not C1. (Anywhere from 1% to 99% C1, the remainder > made up by C2, C2B, C2C, etc, depending on the 'good reasons')
Ah, now I understand why Philip asked "Why associate works with an artist at all, other than through ARs?" I agree with him. It seems to me that using a generic abstract Artist for Works would mean re-implementing the same mistake which was done (for technically and historically understandable reasons) when MB put an Artist field in the Releases. This was a mistake which has triggered a lot of problems (the most obvious was the sterile discussions about who should be the Artist, the performer or the composer). A work has no single artist, and defining an Artist for Works will lead to the same issues, but this time we won't even have the excuse of technical or historical reasons. A work has 0 to n composers, 0 to n librettists or lyricists or whatever (depending on the type and the period), and eventually (not always!) 0 to n performers, 0 to n arrangers... I don't see any reason to give the name Artist to any of these any more than to any other. Either they are all artists, or none are. I don't see any reason to give precedence to one of these over the others. Who is the Artist for Don Giovanni, Mozart or Da Ponte? For l'Histoire du soldat, Stravinsky or Ramuz? For l'Enfant et les Sortilèges, Ravel or Colette? I used French music classical examples, but I could as well cite Reggiani and Prevert (still French ;-) ), Sting and John Dowland... ARs are perfect for this IMO, they just express a factual relation. They don't say an artist is The Artist, they just state who did what. -- Frederic Da Vitoria (davitof) Membre de l'April - « promouvoir et défendre le logiciel libre » - http://www.april.org
_______________________________________________ Musicbrainz-style mailing list Musicbrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
