Dealing with "Editions" for classical works

Tags: #<Tag:0x00007f05081da6c8>


This is motivated by situations that can occur with Classical Music, but may potentially have a wider impact.

I’m looking at Bruckner’s 7th Symphony as an example. There are three common performing editions, the Gutman, Haas, and Nowak. Of the three, the Haas is the most common. In the Musicbrainz DB, the Hass and Nowak editions are listed as revisions of the Gutman edition. This can all be seen here:

When adding a new release of a recording of this Work to the Musicbrainz DB, you are obliged to select one of these three editions to associate the recordings to. This gives rise to two problems. First, it is often the case that you cannot readily determine which edition is the one that was used on the recording, and if you don’t know what the edition is, you can’t choose the correct one, and there isn’t an option for “Unknown Edition”. Second, the process for making the association to the MBID of the selected Work is such that if you don’t actually know that there are a selection of editions that you can choose from, then you end up choosing one of the editions randomly.

I think two things need to be done about this. The first is that there needs to be a “Generic” version of every Work which is the “Parent” Work in the Musicbrainz DB. Then, if multiple editions exist, those individual editions can be listed as revisions of the Generic Work. If the actual edition used in the recording is unknown, or to be more accurate, is not specifically known to be one of the listed editions, then it should be associated with the “Generic” or “Parent” version. Only recordings known to be of a listed edition should be associated with the revision of listed edition. This is what I would like to see a wider discussion of.

The second thing that needs to be done, and this should really be left to a separate discussion once the First question is resolved, is this. If a User wishes to associate a recording to a Work, how do we alert that user to the existence of multiple revisions? How can the UI be modified to accommodate this?

Thank you for your attention :slight_smile:


Based on this idea, I have floated a trial balloon with changes to the 5 existing “revisions” of the Work Bruckner’s 8th Symphony, with the “vote” checkbox checked. I felt this was necessary because the relationships in place were either wrong or absent. It is a good illustration of what I had in mind.


I wonder what the Wikidata linking should be like; if they only have one entry, should it link to just the generic entry, or all of them?


This sounds fine to me (and is kinda what I did with Fratres back in the day). It should probably only be done as a “generic” when there’s no one version which is much more commonly performed than the others - otherwise, I think the way it’s done with Tchaikovsky’s Romeo and Juliet is good enough.

My personal preference would be to have a generic “Work” level and a second, separate “Arrangement” level (which could also be used for editions, revisions and whatnot), where a user would first select the work, and then select an arrangement of it if needed/known. Sadly, that would add even more complexity to an already complex system (probably confusing people) while only being useful in rare cases for non-classical music (maybe something like "is this a Gloria Jones style cover or a Soft Cell style cover of Tainted Love). So it’s unlikely to be added anytime soon, and possibly ever unless we have an epiphany about a very simple and obvious user interface for it.

If the names are already the same (or very similar) then all of them should show up in the same search. Otherwise though, I don’t really have any great ideas, other than them knowing that it’s there.

One thing to keep in mind though is that if they’re using the relationship editor (which is probably the most common way to add works to recordings) any revisions or whatnot will actually show on the work’s relationship list there, so you should be able to see “oh, this has a generic version as well!”.


I wonder what the Wikidata linking should be like; if they only have one entry, should it link to just the generic entry, or all of them?

My feeling is that it should link to the Generic only. If there are Wikidata links specific to the revisions of level they should appear there.


Yes, that sounds ideal to me. It doesn’t add complexity in the sense that the “Arrangement” level is already there, and for the vast majority of Works won’t have a need to use it, and so won’t have any entries there.

Do we have editions of? I haven’t seen that. I hope not, that would be even more confusing. In Classical music there is a reasonably clear distinction between what is meant by a “Performing Edition” and an “Arrangement”. A performing edition is generally intended as a definitive version of the Work, whereas an arrangement can depart from the original Work in ways from slight to extreme, but are recognized as not being definitive versions of the Work. So I am happy with performing editions being classified as revisions of but not as arrangements of. I have already seen one instance in MB of a performing edition classified as an arrangement of instead of a revision of, and I have submitted an edit for consideration.


I agree with how you handled both.


For sure. If your workflow causes you to open the Work’s MB page, then everything will be crystal clear. But, for example, if you enter the Work-Recording relationship as part of the Editing > Add Release dialog the process of forming the relationship involves choosing a selection from a drop-down box, and that’s where the opportunity for confusion is greatest. Confusion means bad data in MB. Someone who is coming in after the fact and editing out mistakes is approaching it from a different mindset.


I have discovered this web site which appears to contain a comprehensive list of which recordings of Bruckner Symphonies use which performing editions. I’m going to use this to review all of the fully tagged performances of Bruckner 8 in the MB database and see how many are in fact correctly tagged :slight_smile:


With Mozart’s Requiem we have editions/arrangements linked with the original fragment. With this work the current structure makes more sense than having all these editions linked with some generic type. For unknown editions we’ve been using “catch-all for unknown editions”.


Thanks for taking time to respond.

I see where you’re coming from with this. To me, this makes sense when you are compiling an academic database for purely reference purposes. What we need, though, is for it to make sense from the practical perspective of real-world users. As far as most people are concerned, the “(Süßmayr Edition)” is the “Generic” version of Mozart’s Requiem. I am therefore quite comfortable seeing this at the top of the tree. In this regard, Mozart’s Requiem differs from Bruckner’s Symphonies, where you tend to have multiple editions in widespread use.

I understand that, strictly speaking, the other Editions listed are not actual revisions of the Süßmayr Edition, but I think it would make the most sense to the most people if they appeared under that relationship in MB. Those informed enough to understand the distinction will be able to live with it comfortably, and I don’t see how it can cause any practical problems. Likewise the (fragments) which currently occupy the top-level position would also function perfectly well as revisions of (again, not strictly speaking accurate) the Süßmayr Edition.

The vast majority of recordings are of the Süßmayr Edition, and few MB editors will have even heard of any of the other editions. So the one thing you want to avoid is having a regular Süßmayr Edition recording classified as some other edition - or even as an “unknown edition” - out of confusion. This is the “most probable problem” scenario, and as such is the one we most want to avoid.

I’m not even sure that this work actually needs a “(catch-all for unknown editions)”. I’m wondering what, specifically, would end up in there, and why. You have a recording, you don’t know what it is, but you do know it’s not Süßmayr?

I guess I should point out my interest here. I’m working on a Music Server App, for a company that is one of MB’s partners.


This catch-all type is for editors who aren’t able to identify the difference between editions (or might not have access to recordings). When you don’t know which edition to choose you will choose this catch-all type. Most of us aren’t even able to identify the Süßmayr edition so we just link works with catch-all type and hope that someone who knows better will fix it later.

Some editions are based on Süßmayr Edition but some include nothing written by Süßmayer. Wouldn’t make sense to link any of them as revisions for it.


Regarding the Mozart Requiem, all of the Editions are categorized as arrangements of. I’m wondering why that is. It seems wrong to me. An arrangement implies a significant departure from the original version. Revision seems more appropriate.

Are there guidelines for when one should use Revision, and when Arrangement? And when something else?


I would be interested to get your take on the equivalent situation with Bruckner’s Symphony No 7. There are 3 editions included in the Musicbrainz DB, the Gutman, Haas, and Nowak. The data is organized with the Gutman as the root, and the Haas and Nowak as revisions of.

Actually, I have submitted a trial balloon edit, creating a “Generic” root, with the Gutman, Haas, and Nowak as revisions of. The generic root is disambiguated to provide for all those recordings for which the performing edition is not known. But that’s beside the point right now, since those changes have not yet come into effect.

Looking at all of the recordings of Bruckner 7 in the MB DB, there are approximately 35 distinct recordings which have been associated with the Work. Of those recordings, 23 are credited to the Haas Edition, 10 are credited to the Nowak Edition, and the remaining 2 to the Gutmann Edition. However, of these recordings, 18 have definitely been wrongly attributed, and one remains uncertain because we don’t know what the actual performing edition was.

My take-away from this is that the majority of Users are assigning the work to an edition randomly. I think the reason for this is that the process for selecting the Work does not do an adequate job of highlighting the availability of multiple editions/revisions/whatever, and therefore does not provide a consistent basis upon which to present the available choices.


I did use probably just a one second of my time when adding relationships between these works so there’s no more meaning of it. All I wanted was to link these editions somehow quickly together. There’s no guideline related to these relationship types. I guess “revisions of” would be more suitable with this work.


I think it might be better if all of the editions of Bruckner’s 7th would be linked as revisions of original 1883 version. That version isn’t published (and I believe performed) but we still got data like premiere event to link with it.

I don’t know these works that well but based on Wikipedia for his edition Haas attempted to remove the influence of Nikisch, Schalk and Löwe in order to retrieve Bruckner’s original conception of the symphony. For me it seems incorrect to link Haas Editions as revision of Gutmann when the main point of this edition was to remove changes to original version.

By following edit history you’ll notice that many incorrect relationships are caused by disambiguation update for Haas edition (disambiguation wasn’t earlier mentioning any edition).

My experience is that our editors aren’t that careless. When there’s a catch-all work the amount of incorrect relationships is low. For Bruckner’s 7th there hasn’t been such a work and such a mess shouldn’t be a surprise.

In a dream world we could get system level support for editions and arrangements but in reality I believe that’s not going to happen.


Yes, that would work fine for me. I also like having all of the other editions as revisions of that. What I really don’t want to see is a cumbersome tree of revisions, sub-revisions, and sub-sub revisions that attempt to reflect the arcane subtleties of the inter-relationships that they share.

That actually explains a lot. I did look into the edit history, but didn’t pick upon that.

Where do you stand on the distinction between revision of and arrangement of? It seems to me that these are used somewhat interchangeably. We have revisions, arrangements and orchestrations as our list of available relationships. In my own mind I have a reasonably clear idea of where the distinctions between those concepts lie, and I think the style guide is reasonably fair reflection of that. But in the case of the Mozart Requiem, I really don’t like the use of arrangement of to describe the performing editions. Is there some reason why someone hasn’t changed that?


Could be because of the help page for “revision of”: “Revisions are usually made only by the creator of the original work. Do not use this relationship to link to arrangements, completions, reconstructions or any other work by a third person”. For me it’s unclear what would be the correct relationship to use with completitions.


Right now, I’d probably just go for “version of”, but it might make sense to get a more precise option for these cases if we can think of one. I wouldn’t say they’re really a “revision”, but “arrangement” is probably not ideal either.

On that matter, it would probably make sense to add a “completed by” relationship to complement “reconstructed by”.


The problem here is twofold. First, no matter how far you go to extend the available relationships with “clarifying” options, there are always going to be edge cases which fall into the grey areas between two or more choices. Second, the deeper you dig into the arcana which define these relationships, the less relevance they are likely to have for the people making practical use of the MB database.

In the classical music world, “Performing Editions” are generally created, intended, and held to be legitimate published versions which present themselves as an official representation of the core work. Sure, they do include variations, revisions, and arrangements of the composer’s original material, and there are often valid reasons for the existence of multiple editions, but their core objective is nonetheless to be definitive. IMHO, for MB, revision of is just fine for performing editions, but I suppose you could also create “edition of”…

“Arrangements” are situations where somebody has made an acknowledged effort to depart from the original piece, and the result is generally presented as being distinct from the composer’s original work. Sometimes the arrangement is subtle, like Mahler’s arrangement of Beethoven’s 9th Symphony, sometimes dramatic like Ravel’s orchestration of Pictures at an Exhibition. But the former, though subtle, is not considered to be a Performing Edition, whereas the latter, though drastic, is not considered to be a Version (and is not given a Ravel Op. number). Performing Editions are not Arrangements.

“Versions” are an awkward situation, and sit somewhere between “arrangements” and “editions”. These are usually created by the original composer, for performing scenarios where the original composition would not be appropriate, usually for reasons of length, the number (or availability) of performers required, or setting (such as a ballet, with or without dancers). Sometimes (see Prokofiev’s 4th Symphony) the rationale is too complex to put in a neat box. It is normal practice for “Versions” to be treated as separate Works, and given separate Opus numbers.