Book of carefully-edited songs, with CD: what Works?

Could I please get some help deciding what Work entities and Relationships to use for “26 Italian Songs and Arias” (medium high voice) (Release/e2b8c)?

What’s notable about this Release is that it’s a book of carefully edited scores of arias and songs from the classical music tradition. The editor goes to pains to explain how the changes to the notes and accompaniment result in a score which is truer to what the composer intended, and is based on careful scholarship. So, it seems relevant to note the editing and scholarship, not just the music performance, and to credit the editor’s role.

Why is a book in MusicBrainz? Because bound into the book is a CD. The CD has recordings of a pianist playing the instrumental part of each song or aria in the book. It is a CD release, but packaged in a large book, instead of being packaged in a plastic box with a small booklet.

The CSG/Works guidelines seem to say that an arrangement Work should only be created if there are at least two Recordings of the new Work available, and if the differences in the arrangement change the sound to the non-specialist ear. Changing the instrumentation counts as “arrangement”. Transposing the piece does not. Presumably correcting the notes and the dynamic markings does not count. And, “If in doubt, do not create a new work.”

That says I should relate the Recordings in this Release to existing Work entities. But then how do I record the basic facts about the carefully-edited scores which are the whole point of this Release? How do I note that Paton edited the score? How do I say that Dunn translated the Italian lyrics into English? How do I say that Alfred released this edition?

The scores have both the original Italian lyrics and a translation of lyrics into English. But the accompanying CD has instrumental-only performances. The translator of the lyrics gets a credit in the book. Should they get credit in MusicBrainz, even if we can’t hear their work? If so, how?

As you look at the track list for “26 Italian Songs and Arias” (Release/e2b8c), you will see that they are presently a hodge-podge. In some cases (tracks 1-9, indeed any track which says “publisher: Alfred Music Publishing”) I created a new Work to represent the arrangement. In other cases (tracks 11-19, and any track which says “publisher: G. Schirmer”) I linked this Recording to an existing Work.

Part of the problem is that the MusicBrainz entity named “Work” doesn’t align well with what I understand a musical work of composition to be. MusicBrainz has style guidelines for representing audible differences that can be traced back to musical notation (e.g. arrangements for different instrumentation). It doesn’t seem to have guidelines for representing differences which aren’t easy to hear, but that still can be traced back to notation (e.g. subtle edits, different editions, etc.). The Work entity doesn’t seem to relate directly to editions of printed scores. Putting “catch-all” in the disambiguation has the feel of an expedient work-around for a complex reality. This has the feeling of getting into a part of music information for which MusicBrainz is not yet ready to be the “ultimate source”. So maybe any choice will be messy.

If I link these Recordings to existing Work entities used by other Recordings, then how do I note that for these Recordings in particular, the editor and arranger is known? An Artist-Recording “Arranger” relationship?

Should these works be related with Publisher to a Label (as, say, Work/f27cd and Work/3e33c are)? Since a Work entity can correspond to an unknown set of different editions of published music scores from different publishers, in what circumstances is a Publisher relationship between Work and Label meaningful? I love being able to identify one Work as “the Schirmer edition” and another as “the Alfred edition”, but I fear Work entities don’t have that meaning.

One more layer on the cake: the same editor and scholars and performers made a companion book, “26 Italian Songs and Arias” (medium low voice) (Release/c1fc5). This Release is also a book of carefully-edited scores, with a CD bound in. However, these scores are transposed a few steps lower than the scores in Release/e2b8c. Should the recordings of Release/c1fc5 point to the same Work entities as Release/e2b8c?

Thanks for helping me figure this out. Once I understand how to handle this Release, I’ll go back and clean up the Work relationships.

I’m going to confess to more or less ignoring (or, let’s say, following the spirit, not the letter :grin:) of large parts of that style guideline… It does even say up top that “lost and/or unrecorded music can be entered if necessary for correct relationships and/or to make a composer’s complete catalogue available.”

After a composer writes down sheet music, he sends it to his publisher who cleans it up, typesets it, and prints it (all this was of course done by hand). Mistakes happen—both in the autograph manuscript the composer scribbled out, and in the publication process. Sometimes a piece is performed before being widely published, and the composer decides that wasn’t what he wanted, and does his best to eradicate the 1st version (for example, Tchaikovsky’s 2nd—which BTW, I put works in for, and some of those only have one recording). With some composers (e.g., Bruckner) there are numerous versions of each work.

It seems to be a common pastime of musicologists to attempt to rediscover the composer’s true intent. Or even original intent before he changed his mind. Lacking any talent for magic (or at least not willing to admit it), musicologists resort to more mundane means: going over piles of autograph manuscripts, original printings, letters—anything they can find in the archive, really.

It sounds like that’s what your book is. So I think it’s actually something that could go into the works database even if it didn’t include recordings.

I would suggest putting them in as new works.

I’m not sure what the best way to link the two is, as the documentation is somewhat of a mess. There is a “revision of” relationship, but that documentation says to never use it if its done by a different person. But then there is also an “revised by” artist, which would clearly be entirely unneeded (as it’d be always the same as the composer). That’s odd. Seems like this would be a reasonable use of those two relationships, at least if they’re small changes attempting to discover the composer’s intent.

The style guidelines, though, and some (but not all) existing practice, add the editor as an additional composer. (It’s not, as you observe, an arrangement). Then often don’t even link the works :frowning:

I think the style guidelines could use some clarification (and probably changes) in this area.

You should use disambiguation comments to identify the revision. E.g., “ed. EDITOR NAME” or something. A year in there as well, if the editor may have done a few versions.

edit: notice I missed some questions in there

  1. Translations get another work entry. There is a checkbox for that in the has later version relationship, I believe. The translator gets credited there. There won’t be any recordings of that work, but it strikes me as “necessary for correct relationships”.

  2. “The Work entity doesn’t seem to relate directly to editions of printed scores.” I disagree. It does (at least as far as the scores are different musically—I don’t think we’d put in a new work because someone decided to print the exact same music on thicker paper, not doubt for twice the price or because they made a second printing to fix a typo in Allegro). The catch-all for unknown versions is because we often don’t know, because the CD doesn’t say. If we do know, we use the correct work entry.

  3. The transposition—I don’t know. The style guidelines seem to say that’s the same work. I guess that makes sense to me if it’s an essentially mechanical process (“yep, moved every single note down one octave”). If it’s a creative process, seems like it should be a new work. No idea if anyone else feels that way.

I’m afraid there’s no better place than release or recording annotation for this data. When editions aren’t well known (or considered important) releases don’t usually mention about them. There’s no use for edition works if editors aren’t able to identify them. I would recommend adding only works for editions which can be easily identified. Same reason for not adding separate works for transposed versions.

It would be nice to be able to record every tiny detail related to music but we need to be practical. I believe one of the main functions of MB works is to group recordings of the same work together. There’s quite little use for works if popular works are split to hundred different versions.

1 Like

confusion. I think the editions of all the works Jim DeLaHunt is asking about can be easily identified. He’s holding the critical edition published scores…

They’re still linked together as long as we link the works to each other. It’d be nice to figure out how we want that done—when I looked briefly at the Bruckner symphonies, for example, it doesn’t seem to have been done.

I suggest maybe we should have all editions/reconstructions/etc. have some version-of AR to the catch-all (in addition to any other ones, as sometimes we know that Bob’s reconstruction is based on John’s, and should note that as well). This would also let anyone looking at the catch-all work quickly see all the other versions available, aiding finding the others. (And if looking at one of the specific ones, the full list is only a click away).

(Also: Recordings would only wind up split if it turns out the folks releasing the music consider it important enough to note which edition is being performed—most of the time they don’t, and 99% of the releases would go under the catch-all work. Except of course for a few composers & works where it really matters and “serious” release will say which edition.)

What I ment was that people aren’t able to identify these based on recordings. How I could tell if my recording is the same as the one added by Jim? It’s not that big deal with non popular works but with more popular works there could be a problem. Another problem is our user base who like to link whatever edition they find first.

When service is more ready we could consider adding printed score releases to BookBrainz. I believe it would be the best place to list different editions.

Editions of Bruckner symphonies are commonly known and often releases mention about the version. These aren’t that hard to identify thanks to

Lots of works that exist in different versions already have separate works per edition, plus a “catch all” arrangement. I agree that the interface could be better at presenting these different works to make it easier for inexperienced editors to pick the right one, but I wouldn’t want to see them all merged into one work either.

Most of these works also follow “two different recordings available” rule and are commonly known. I have no problem with that. Most of the libraries don’t see any reason to standardize names for all rare editions and I believe there’s no need for us to do that either. Editions of Beethoven symphonies for example aren’t commonly mentioned on releases. Most of the editions could be listed as score releases (BookBrainz) instead of new works.

Thank you for the validation. That’s what I’m leaning towards also.

I agree. It would be nice if that could get a push from this discussion.

OK, I like that. It will get tedious to enter 26 translation-works by hand, however. Maybe this doesn’t happen until there’s a web app feature or a user script to do the repetitious work.

I like the idea of being able to record relationships between music scores. However, we have to define better what it means to be “different musically”. Thickness of paper doesn’t count. Fixing errors doesn’t count, even though it might result in a difference in performed sound?
What about the difference between orchestral parts (played by the orchestra) and a piano-vocal reduction (from which the chorus sings)? And what if the chorus is using a hodge-podge of four different editions of vocal scores? I’ve sung Beethoven’s 9th Symphony with exactly this mix of music score editions.

Agreed, that’s a tough decision. In my case, there are two Releases with two different music scores, each giving the same arias in different transpositions. The transposition is the essential reason for there being two Releases, not just one. Same Work, or different Works?

I like using “editors are able to identify them” as a good rule for Works on MusicBrainz. It resonates with my intuition that MusicBrainz is really “RecordedmusicBrainz” with a bit of “BiographyBrainz” built in.

That’s a reasonable division of labour. We would then probably have lots of situations where one “Work” in MusicBrainz corresponds to multiple “Music Score Editions” in BookBrainz. There’s still the problem of tracking the musical essence of relationships between different scores. I suspect BookBrainz won’t want to be in the business, and will push that back to MusicBrainz. That in turn will lead to MusicBrainz needing to have Work distinctions that aren’t audible. Difficult.

We should remember that a lot more goes into the audible differences between recordings, than the details of which scores they used. The modifications each musician makes to the score for performance, their musical skill, their artistic interpretation, their instruments, their performance practice: all of these are differences that will never be captured by tracking music score editions. I think this is an area that MusicBrainz doesn’t yet attempt to record.

I recall making that same point in another thread :-), but getting strong resistance. I do agree with being practical, however.

Thank you, everyone, for the good ideas. Please keep them coming.

I believe MusicBrainz is slowly becoming more than “RecordedMusicBrainz” but it currently is pretty close to this. I actually have added hundreds of works which haven’t yet been recorded or at least not found from MB. Even some missing works which can’t be recorded. Fact is that MB schema isn’t that good for storing useful data about editions. Some editions are almost necessary to have when every damn release is mentioning about them. I’m just not sure how useful it is if I keep adding works which have nothing more than editor (and maybe publisher) linked with them. How many of you have even heard about different editions of Sibelius symphonies? I’ve never seen editor names for these mentioned on releases, concert programs or anywhere else than on printed scores.

For sure there’s more data we could record about works and arrangements. Some day we might have all necessary data about instrumentation, average durations, etc. But in MB world it’s going to take years (my guess is 10 years). Works are still little used by our editors. My guess is that more than 90% of classical recordings don’t have any work linked with them and majority of them don’t have any performers linked. I hope that we could focus on fixing majority of basic level data before continuing to add more complexity to it. Don’t get me wrong. I’m not against of every new feature or guideline change. New ideas are always welcome (at least if I’m in a good mood).


I think I partly disagree with your disagreement :slight_smile:

MB’s implementation of the Work entity allows you to create a version that corresponds 1:1 to a particular score, but that is not their primary function.

MB-Works were conceived (partly) as an adaption of the FRBR-Work entity, which are higher-level abstractions. They represent the idea of a thing. Because we don’t have an entity that matches the FRBR-Expression, MB-Works sometimes do “double duty”, acting as both, but I think their primary function is still to be abstract and general rather than specific to a score/version/arrangement.

1 Like

IMHO, this has nothing to do with the issue here. Tchaikovsky revised his symphony, so we have a new version, which is different from a new edition. I agree with the opinion that works are abstract entities, Beethoven’s 9th is Beethoven’s 9th no matter wether the performers used a 19th century edition or the latest Bärenreiter one.

A critical edition of a work is just that, I really don’t think we should have dozens of different works for each work.

I don’t think so. A new work would be appropriate in this case, though.

About Perhaps I don’t realize how many changes John Glenn Paton did, but if he just edited the score using the arranger relationship is wrong IMO.

Having lots of details about scores, editions etc. is great, but I don’t think MB is the good places for that. In an ideal world, we could link a recording to an edition in BookBrainz to say that the performers used a particular edition.

I fully agree, this would be awesome.

My understanding is that this relationship is meant for Non-PD works which have only one publisher (it’s very common for CDs to list the publisher of Non-PD music).