Why have classic track artist fields?

classical
soundtracks
composer
artist
Tags: #<Tag:0x00007fcb4c8bbf40> #<Tag:0x00007fcb4c8bbe00> #<Tag:0x00007fcb4c8bbcc0> #<Tag:0x00007fcb4c8bbb80>

#21

That decision was made a long time ago, and I suspect by a very small group of people, what may have been right then doesnt mean still right now.

  • Surely it is MusicBrainzs responsibility to have a sensible database and currently
  • from a Database point of view using the way you use trackartist/recordingartist is wrong, it is inconsistent and it means not capturing info that can be captured for Pop/Rock.
  • from Database point of view storing Composer as track artist is pointless anyway since already have composer relationships.
  • The UI to the database make it hard work to enter relationships easily.
  • Mediocre taggers are not your responsibility, but if the relationship data is not there because the UI is poor then that is MusicBrainz responsibility. and as previously discussed (and I think you agreed) if MusicBrainz provides no way for taggers to identify if released entered CSG then again that is the responsibility of MusicBrainz not the taggers.

It is not just me who thinks this, its a common complaint on Roon, I don’t know what you know about Roon but they do have users who understand Classical very well. There does seem a great resistant to accepting there is even an issue, despite the evidence


#22

Right now there are officially “genre: classical” tags in the web service, FWIW, which should be a good start to know that at least for releases / RGs tagged that way it makes sense to assume CSG. Not saying that’s a perfect substitute, obviously, but it’s a start :slight_smile:

Does it? It does, admittedly, make it non-obvious that they’re a thing you should add, but I don’t think adding them once you know this is difficult. Linking to works is the only slightly more difficult part (without userscripts) - I’d be happy to see something like the @loujin Guess Works userscript adapted to include in the default UI.

I’m not really claiming the current way we store things is ideal. It might very well not be ideal: as you probably know, my personal preference would be not to even have track and recording artists as a concept for classical, which is probably quite different from your personal preference, but means I also don’t think the current situation is the best it could be.

I just haven’t seen anything that suggests changing what’d certainly be over a hundred thousand releases to adapt to a different track/recording artist choice would be useful enough to be a good use of anyone’s time. The same time being used to add proper relationships to releases missing them would be infinitely more useful for MusicBrainz itself and for the use any tagger wanting to take classical music seriously. And if the time wouldn’t be taken to change them all (or most of all) properly, we’d end up with another in the middle situation which would be bad for whoever wants what we have now and bad for whoever wants the alternative.


#23

Yes because the tracks page with track artist (that gets propagated to recording artist) is a standard part of the release editor, and you can you seed a release to add the releases with tracks, but you cant use it to add relationships.

I think the trouble with generating from relationships isn’t it doesnt allow you to capture the joining words and order, but I may be wrong about this.

To do it manually would be quite ridicolous, but to amend the guidelines and programmatically set track artist to be same as recording artist would be a good start, and that would fix track artist where the recording artist is set correctly

Any progress on MBS-211 would help in so many situations

Then it would be clear where recording artist/track artist is not set correctly and manual editing required.

But the point is not really how it is done, I’m struggling to convey the problem with the recording/track artist because replies from MusicBrainz are along the lines of there is no problem rather than there is a problem but we don’t want to fix it.


#24

The thing is, you’re bringing up multiple problems, but their solutions actually conflict with each other.

You seem to want the track artist to be the performers, which is a perfectly valid option, if different from the one we currently chose. It does, admittedly, match the definition of “artist” tagging fields more closely. It also hides composer information more while bringing more visibility to performer information, which can be desirable (in one-composer releases with multiple performers), equally unimportant (on releases where there’s only one composer and one set of performers) or less useful (in one-set-of-performers releases with multiple composers). I do not think this is necessarily a worse choice than composers, but I also do not think it’s a clearly better choice.

At the same time, I understand that you want the track artist and recording artist to follow the way it’s done in the rest of MusicBrainz. You seem to think that means the same as the previous point, but as @derobert explained earlier, that’s not true. What that would mean is that we would credit for the track and for the recording whoever the tracklist has listed, meaning:

  • For most releases with multiple composers, the track artist would still be the composer, since they’re who is usually specifically credited for the track on the tracklist.
  • For most releases with one composer and multiple sets of performers, the track artist would be the performers alone, since they’re who is usually specifically credited for the track on the tracklist.
  • For most releases with one composer and one (set of) performer(s), the track artist would be effectively the release artist, since each track is unlikely to have an artist marked - so composer and performers both.

I don’t think anyone actually wants this inconsistency for classical, even if it would actually be more consistent with the usage in the rest of the database.

The only change that I feel would bring an actual improvement to both information storage and navigation would be to make track artists basically equivalent to release artists: “Track composer; Track performer 1, track performer 2… n”. It would also make editing more complicated though, by requiring selecting multiple track artists for basically every single classical track - I’m sure a better UI would help with that to some degree, but it would still be annoying. I’m convinced this would be at least somewhat better, but unconvinced the benefits are worth the annoyances, and fairly sure the benefits are not worth the effort of changing everything now.

So, there might be a problem (there’s certainly a problem for some people, there’s certainly no problem for others) but I’m not sure I see a reason to try to fix it in any of the proposed ways.

You bring up another problem that does exist and should be fixed: relationships are not obvious and many people do not enter them, even when editing classical music. I expect once our design work reaches editing interfaces, this will be something we will put extra attention towards. In fact, unless I have left MB for some reason by that time, I can guarantee I’ll work to get attention on this issue, because every release that gets added without relationships is a release I** need to fix later.

** Or one of a smallish group of people who actually clean up existing classical content and do a lot of good work, if you’re reading this you probably know who you are, thanks so much for your work.


#25

Okay I see where you are coming with this, but I dont really agree with @dreoberts analysis of the album, on the front cover its credited to Seattle Syphony and Chorals and Gerald Scharz (conductor). You can argue that is its credited to Howard Hanson, or that is just part of the title., i.e I would read as Howard Hansons Symphony No, 6 & No .7performed by Seatle Symphony and Choir, conducted by Gerald Schwarz,

But on the back it is not individually crediting Howard Hanson as the track artist, its simply repeating the title and showing the movements for each symphony under each.

Basically they havent explicity credited the track artists, but they can be simply derived by listening to the music.

So just like Pop/Rock each piece of music has performers, and often they are same on every track and quite often they are not which is why we need track artists. But the problem is that classical releases they are quite slack about actually crediting at a track level, it often has to be derived.

So taking what is written too literally leads to the three inconsistent meanings you specify.

So my preferred option is to take the RecordingArtist as the basis for TrackArtist but allow for when intentionally on cover they credited artist slightly differently, but when they don’t bother to individually credit at least main performers I don’t think that should lead to no credits. Taking this approach is almost the same as Pop/Rock rather than completely different.

Should we include composer as well as performers as track artist or not, I tend to think not then very easy to keep recordingartist/trackartist mostly the same but I don’t have such a strong view about this. The problem is only having the composer, and therefore not having an easy way to add the main performers when creating the release.

I think this would be a welcome improvement, would it be more complicated to edit though, i don’t think so.
For one thing track artists default to release artist so actually for many releases would start of correct rather than having to modify them to only have composer. And the if do further edit for recording artist we would only have to delete the composer. much easier than having to enter all the performers .

Musings, if we didn’t include composer at track artist or release artist level then things would be much simpler to enter, because in many cases all three values would be the same. But of course composer very important so what we really need for the composer to be able to be added as part of the ui edit so easy to add to fro release and tracks (and then stored as relationship)

And for there to be a Composer Search like Artist Search, and a way to view all releases they have composed rather than performed on. But all that is a very major task.


#26

(I really ought to edit this down some, but I even more really ought to go to bed. So, sorry!)

If the composer is not the person most responsible for what you hear at a performance, then, well, you’re not listening to classical music. The now, what, thousand-year long practice of musicians trying to realize the music envisioned & recorded by the composer is sort of fundamental to the classical tradition.

Or: the composer often isn’t in the room; often isn’t even among the living. But the composer still is mainly responsible for what you hear.

Well, first, some people don’t want to see the composer there. Some people (myself included) do. But you’re talking about tagging. How you map MusicBrainz database fields to FLAC/MP3/whatever tags is up to you (and there is, by the way, a separate performer field, too).

And in the MB schema, there isn’t another place for the composer. It’s the only place we have to even get close to crediting the composer as he/she is credited on the release, the recording artist & advanced relationships often work (or at least well enough) for the performers. And that matters quite a bit; e.g., there are a lot of ways print Tchaikovsky’s name.

It’s only wrong with the definitions you’ve made up for the fields, which aren’t the ones used or documented by MB. Your alternative means we instead couldn’t capture other info (the composer). And… if it’s different than pop/rock, well, so what? They’re different music traditions with different requirements.

That’d be awesome, if we had three-way relationships to give artist credits per-release even on works & recordings (e.g., a work-release-artist relation).

The joining words are hardly ever there/important for classical releases. Order is—rarely—but probably we could have ordered relationships there (like we do for works). (Grouped ones would be more useful, really).

I agree there is a problem in that we can’t store the credited-as field for performers when a recording is shared between multiple releases and the releases credit the performers differently. And I agree your proposed solution would fix that, but I think (as pointed out above) creates far worse problems. @reosarevok’s alternative of putting both composers and performers would work (except for, as he points out, the amount of extra work and the significant transition costs).

Please go fix https://musicbrainz.org/release-group/055be730-dcad-31bf-b550-45ba9c202aa3 then, that’s currently credited to something that’s just on the cover as the title. Seriously, of course, we’ve decided in MB to not treat that as part of the title, and even if were part of the title, it can be a credit too. There are some release where figuring out exactly which words on the cover are the title is hard — this isn’t one of them.

I think you must have very limited experience with entering classical releases. Changing the album credit to just the composer is trivial; you click edit, hit the X button a few times, and check the “change all” box. If an entire release has the same performers for all the tracks, it’d save maybe 30 seconds, tops. But when a release doesn’t (which is common, it applies to that Hanson release, which remember was picked randomly) then the edit relationships page is much nicer. For one thing, you can select multiple tracks to add an artist to at once; you can’t do that in the track editor. The tracklist also has width limitations (the field is only wide enough to hold a few artists; more is common on classical) which the relationships editor does not. And you can more easily save partway through the relationships editor.


#27

I said

even though usually the composer had nothing to do with the actual recording/performance

i.e they were not involved in the recording, they didn’t decide to record this version of their music with this orchestra, surely you must have realized what I meant here !

Ive already explained this, you are using the same database field for different things, and you don’t store what method you are using in the database, this is just database basics.

You seem to be saying we must store the composer in track artist so we can can capture how displayed on the release, but its unimportant to capture how performers are displayed on the release. This seems to be quite a simple understanding of classical, i.e composer v important, everyone else less so.

With the current system you have to enter the composer for each track, but then enter performers for each recording then submit release. Then we have to use the relationship editor to edit relationships for each recording, so there alot of duplication and it is clunky.

The fact is that no-one thinks MusicBrainz is good for classical apart from MusicBrainz editors.


#28

Even in general people have doubts, I said that MusicBrainz was better than Amg here and the reply was

Conceptually: yes, I fully agree. But it still lacks content wise and for a database largely depending on voluntary data provisioning and data maintenance that is a problem. Hopefully more and more people and organisations will find it useful to also add stuff at the metabrainz sites; and maybe the sites will make it easier to do so, too.


#29

Right — the composer is often even dead. But that doesn’t mean they have nothing to do with the performance. Unless it’s under an assumption that track artist must be only performers, in which case it seems like begging the question.

For non-classical, the people most responsible for the music being produced are often present at the recording session. For classical, that person often isn’t. Leading in to…

Unless the track artist credits are for the person/groups principally responsible for the track (i.e., #1 in my post from the 15th), in which case we’re not. It’s only being used for different things if the track artist is defined as something like “exceptions for when the recording artist is credited differently on a particular release” which seems something like how you view the field, but you seem unique in that way.

(Now, it’s probably true that there are a few tracks on classical releases where you could argue someone other than the composer is principally responsible, and really ought to be credited there instead. Or more cases where the release very prominently credits a performer more than the composer. But then you’d be arguing that we should change some classical releases, not all of them.)

I agree that we really ought to store in the database if a release is per CSG or not.

Well, no, actually I gave a few reasons why if we had to pick one or the other, having the composer info there is more important. I admit it probably wasn’t clear in that rambling post, but to recap:

  1. As MB stores the data, there are two places to store the composer: as a work-artist relationship and as a track-artist relationship. The work-artist relationship is shared between all releases of all recordings of a given work, so it’s very widely shared. So “credited as” is essentially useless there (or can only be used for, e.g., the way the composer was credited on the first printing of the score or something).
  2. As MB stores the data, there are three places to store performers: the two recording-artist relationships (both the advanced relationship and the recording artist field) and the track-artist relationship. Recordings are shared, but nowhere near as widely as works. There are some seriously problematic cases (e.g., Alfred Scholz, but also more sane stuff too) but for the most part, recordings are on a small number of releases (and even fewer release groups).
  3. For many recordings, all the releases credit the artist the same way, or at least close enough (e.g., will differ on including a title like “Sir”). So the recording-artist credited-as field is good enough.
  4. For many popular works, releases credit composers very differently. Russian composers like Чайковский are one prominent example; there are numerous ways he’s credited on releases — at least in English, is seems we all agree on how to transliterate his last name, but not first or middle. I suppose that’s better than Рахманинов, where there is disagreement on the transliteration of his last name, too.

No, the best way to do it is to enter the composer in the tracklist, then use the relationships editor, then use the userscript to generate the recording artist based on the relationships you just entered (and if you take a look at the edit history for classical releases, that’s how its done by pretty much everyone doing it). So the data is duplicated, unfortunately, but its duplicated by quick automated means.


#30

It does mean exactly that in this context, the composer was not involved in the performance, the composer has no knowledge of this performance.

But I’m not against the composer being included as track artist, what Im against is the performers being excluded from the track artist.

Well, when you look at Pop/Rock recording/track artist are 99% the same and the release editor has options on the Recordings tab to Copy all track artist credits to associated recordings. so if I am unique, im basing it on the evidence.

Whilst is probably true that composers credit vary more than performers, I dont think the argument really stands that it is okay to not bother capture performer credited as on release as its good enough. And when it comes to very famous conductors, soloists and orchestra they have the same problems as composers,

Right that is considerably more difficult then it should be, and is not a valid solution for casual editors. So this solution basically prevents anyone adding classical releases unless they have alot of free time, and a deep interest in MusicBrainz. It also prevents methods using release editor seeding from being able to set up a classical release correct


#31

Oh dear, this is getting to be a TLDR thread. I will try and go through it when I get sufficient time!
I’m fairly agnostic about the issue, but if anything am more inclined to agree with @ijabz on the following grounds:

  1. The CSG was introduced when much software could not handle “composer”. That is not the case now. The result is that the CSG over-rides the more general rule that the track artist should be as shown on the release.
  2. It is a pain when editing to have to use a script to get the performers rather than the composer shown as the recording artists - this frequently results in a non-automatically applied edit.

On the other hand, there would be a need to have a credited-as field for the composer (at track level) unless the proposal is to include composer in track artist (which still results in problem 2).
FWIW, I disagree that

I think that @ijabz has tried hard to address that. And, as regards Picard, I hope my Classical Extras plugin achieves the same (and more :wink: ).
And FWIW some more, I use tag mapping in Classical Extras to overwrite the track artist tag with the principal performers (and I can still get the credited-as for the composer because that’s in the original track artist tag).

Of course the big issue for MB would be the pain of changing all the data, and the possible unintended consequences for those who use it, expecting track artist to be the composer for classical releases.
BUT, I think MB has a reputation (most unfairly, given the quality of the database) of not being “classical-friendly”.


#32

I agree with this one, but as mentioned earlier, overriding this just makes the track artist an inconsistent mess. I wouldn’t mind that too much though, since I feel it is fairly superfluous, but I’m not sure most people would consider it an improvement.

Agreed, though I feel the solution to the second part of this issue is to make more edits apply automatically. The solution to the first is to either implement the script in the MB code, or let users select recording artists separately - the former seems more reasonable.

Is there any user-edited database that has a classical-friendly reputation in your experience? I wouldn’t mind having a look to get some ideas :slight_smile: Allmusic is ok, but it’s not user-edited (so it’s easier to tell people to do annoying things to store the data to make it easy to use, because they’re getting paid for it), and they have a completely different structure for classical (even more than we do) and I still feel a lot of their data is mediocre - they do have a lot of it though. Discogs is… ok? Mostly because they have a lot of relationships. But they have very little useful data at a higher level, even after adding compositions. Who is there who does these things well and we could look into?


#33

No. My personal view is that MB is by far the best there is. There are a number of closed databases (e.g. SonataDB) but it’s difficult to peer behind the curtain.
I have a sneaky feeling that the reputation is more to do with how other software uses MB than MB itself, although it may be related to the interfaces MB provides. I haven’t investigated the mechanisms used by, for example, ripping software, to get the MB data, but I do know that the results are inconsistent. dbPoweramp frequently doesn’t know about MB entries and doesn’t provide MBIDs. CueRipper is much better, but doesn’t get all it could out of the database.
IMHO it would be a good idea for MB to work with the developers of tools such as these (and Roon) to get the best out of the best open database there is.


#34

Agreed, there are not really any databases significantly better than MusicBrainz, but they appear to users that they might be better. The key thing that makes MusicBrainz look poor is when you look at a release in the Artist field you don’t see the recording artist (the performers) only the Track Artist and since this is just set to composer it is deemed wrong by most Classical buffs.


#35

For this you can click Display Credits Inline button located at the top of the track list.

Not only it will show you the performer names, but their instruments as well.


#36

Ah… I thought from your concern about the track artist being always different than the recording artist in classical (but not pop/rock) that you objected to having the composers in the track artist, and wanted only performers.

In that case… I don’t really have any principled objections to it. Just some practical ones.

It strikes me as something that’d take a lot of time to get there, and would (with the current tools) make adding classical release (especially complicated ones where who performs on each track differs, like opera) more difficult (that seems pretty hard to argue against — you now have to enter the information twice). Better tools could solve all of this, of course.

But if I imagine what those better tools are — I think I’d pick not entering track artists at all when entering a release, and instead generate everything from ARs. ARs that have been enhanced to store per-release credited-as and possibly order as well. At least for how I enter new releases into MB, it’s easier the less I have to enter before getting a release MBID; once you have an MBID you can also start doing things like uploading cover art.

And of course once you’re generating everything from the ARs, you can expand the API to let you request an artist field filled out however you like — composer only, performers only, or both.

(Another reason to pick ARs if you’re only going to do one: ARs include what instrument (etc.) the person played; track artists do not).

Requiring more info to be put in to enter a tracklist makes that worse, not better. Right now, you can enter fairly minimal information and then upload cover art. You can use the cover art to ask for help from someone more experienced. If you had to get all the performer info in before even adding the album, that’d seem worse for casual editors.

I’m with you though that it should be easier. I believe that is a widely shared opinion.


#37

I’m well aware of that, but I am talking about first impressions, and if you haven’t selected that then it is easy to miss the information at the bottom. In contrast it is very easy to compare the trackArtist for each song but unfortunately this only shows the composer, hence the problem that a cursory glance of MusicBrainz would give the impression that only the composer information is stored.

Even if you have selected that then you have the problem of too much information, you can see the information for a particular track but very difficult to work out if all tracks have same performers ecetera.


#38

I do object to the difference, I would like performers in both recording artist and track artist field, and possibly composers in both. I’m not sure if it is right to have composers in the artist field but I can see practible advantages to doing so.

Going back to your track credits example I dont believe there are any track credits for that release. The principal performers are all credited on the front cover but they haven’t gone to the bother of listing the slight differences (Orchestra/Choir) per movement or work. The composer is not credited on the back, that is simply part of the album title.

In the database generating fields from AR’s does seem a good idea but it would definently need order not just credited as, because actually order is important. Isn’t it usually the case that the most well known artist comes first, by default conductor tends to come first but if its a famous violionist they may be top billing, if there are multiple solo performers they would have ordered in terms of standing.

But the UI would have to change so that the ARs was an integral part of the release. The way I imagine it working was you would simply add a series of people to the trackartist, and specify their role as you add them. Then when you went back to edit later you would find that these have all been added as relationships and autogen the artists field.

As a user I want to be able to add the information into the release from the cover, submit the new release and be done. i.e I want o be able to add sufficient detail at initial add release time to consider that okay I havent added 100% of the information, but I have done a good job and if someone wants to go further they can. Taking this further I can very quickly add a Pop/Rock release using seeding to add release from Discogs, and usually needing no further edits, but I cannot do this for Classical.

I do not want to add basic detail, and then have to go back in to fix like recording artist (because should be different to track artist) and then have to add all the relationships. Even worse I don’t want to do as you suggest and ask for help, and then have to liase with someone else. That is far too slow.
.


#39

That’s as untrue for pop/rock as it is for classical, since that also means no relationships are added. It’s just people care less about having proper info for their non-classical releases :slight_smile:


#40

Okay, you are right I didnt phrase that very well. But for Pop/Rock the basic info without relationships is good enough for casual users of MusicBrainz data, but for classical it is not, knowing at least the conductor/soloist/orchestra for each track is basic info that cannot be added as part of initial add.