Attached is the IRC chat log from this morning's (PST) chat. Thanks to everyone who participated.
Paul
--- pbryan has changed the topic to: Agenda: Works: qualification, granularity, hierarchy, opera/classical. <pbryan> Well, it's 11:01 PST, so time to get started. Thanks everyone for attending. The objective is to discuss works and try to establish some foundations for a guideline to be written. --> jdlh (~j...@66.183.100.133) has joined #musicbrainz-style <pbryan> So far, there's been a lot of varying opinion on what works should be, and a lot of open questions and I was hoping an interactive chat would help move things along. So, topic #1: what should be and what should not be a work? <jdlh> Hello all! <pbryan> Hi jdlh. <ruaok> pbryan: start simple and easy, why don't you? :) <pbryan> :-) I think there's general agreement that it should be some composition. <jdlh> A work should be a database record in which we can hang Classical Style track titles and album titles. (It may be other things besides.) <kepstin> In my opinion, a work should be an entity that represents new creative input. <pbryan> jdlh: Classical track titles are usually individual movements, and sometimes contains more than one movement. kepstin: So, your definition would include a remix of a previous composition? <kepstin> pbryan, in many cases, I believe it would, yes. <pbryan> Would it be safe to say that every work has at least a composer? <kepstin> that's a bit more tricky, because it would then be a work derived from recordings, derived from original works which, by transitivity, means that there is a composer involved eventually <pbryan> If the answer is yes, then works should be limited to musical content only. * nikki doesn't see why works can't be used for things like audiobooks too <pbryan> In other words, do we rule-out audio books, oratories, etc? <caller_6> and poems? some lyrics started as poems <kepstin> yeah, In that case I don't think a composer would be a requirement for a work. <pbryan> Okay. So, is there agreement works represent more than music? <ruaok> I agree with kepstin on both the remix and not needing a composer requirement. <caller_6> what about, instead, "writer"? every work has a (fuzzy) writer. <kepstin> all in all, some human input is probably required, whether it be a composer/lyricist/writer/producer, or whatnot. <pbryan> A work represents a piece of recorded audio content that is authored in some way. <jdlh> In Classical Music, I can think of at least 3 things that might be called "works": <ruaok> pbryan: +1 <ruaok> I think a work requires some "createive" work. <pbryan> Okay. <kepstin> pbryan, A work represents the authoring of a piece of recorded audio content. <jdlh> 1. A complete composition, e.g. a symphony or concerto. Sometimes assigned "opus numbers" by the cultural tradition. <Mineo> caller_6: at least one :) <ruaok> but it doesn't really have to be a composition. a remix or mash-up is ok. <pbryan> jdlh, That's topic #2: granularity. <caller_6> pbryan: a work must be recorded? <pbryan> caller_6: Well, that's a good question. Do we want to track non-recorded works? <kepstin> caller_6, that's more of a musicbrainz thing, we're primarily representing works that have been recorded. <nikki> I think in general they should be recorded, but I don't think it should be a hard and fast rule <kepstin> it's not really part of the definition of works; more of a side-effect of how people use the database <jdlh> 2. A movement (or part) of a complete composition. because complete compositions are long and the cultural tradition accords great significance to the movement as a segmentation of the whole work, and because movements are often the granularity of tracks in recordings. <caller_6> I think Brian has mentioned wanting to list "lost" classical works. <ruaok> +1 on nikki <pbryan> Are works meaningful in MB if they're not associated with recordings? <ruaok> no. <pbryan> Is that an objective of works? <ruaok> since they are tied to other things with ARs only... <caller_6> pbryan: point taken <jdlh> 3. A work (concerto, etc.) in the #1 or #2 sense, but altered by additional composition to make a related but distinct creative product. <pbryan> Okay, if we're okay with #1, I'd like to move onto #2. (and so does jdlh, I think! ;-) Granularity! <jdlh> Hold on, I think my #1 #2 #3 are part of answers to your first question, pbryan. <pbryan> Okay, holding... <jdlh> But maybe I'm not understanding what you meant by your question 1. <pbryan> I think jdlh that you'd agree a work is at it's most basic a a piece of recorded audio content that is authored in some way? s/a a/a/ <jdlh> " a piece of recorded audio content that is authored" agreed. <kepstin> pbryan, I think that definition's a little backwards, because the work isn't a piece of recorded audio content; it's what leads up to the recording (as there may be multiple recordings of a work) <kepstin> that's just a wording issue tho <pbryan> Point taken. I'll work on the wording. I just wanted something to root the rest of the discussion on. <kepstin> anyways, granularity? <pbryan> Note, I'm sure others will come back with traditional or Gregorian chants at some point later, at which point I may need to justify the definition of "authored", at which point I'll do my Bill Clinton impersionation. Granularity: What should be the supported granularit(ies) of works? <jdlh> Work: "A set of authored instructions to performer(s) on who to take a musical performance, which has been recorded". <pbryan> I've heard albums, I've heard tracks, I've heard movements. jdlh: Definitely an improvement to what I've said. <warp> pbryan: I would think that can vary depending on the work you're trying to model <kepstin> jdlh, that's not necessarily accurate; because for many types of modern music, the only "performer" is a computer synthesiser <ruaok> I think releases or tracks. we don't have the concept of a movement, do we? <nikki> for most pop music, the song is the work... for some concept albums, I guess the entire album would also make sense as a work... no idea about classical <pbryan> Well, classical works in general definitely have the concept of movements. <kepstin> ruaok, the point is, we may be able to use works to add the concept of a movement. <ruaok> interesting <pbryan> And certainly movements have been well represented in CSG. <jdlh> @kepstin: ah, but what instructs the synthesiser how to be behave? Programming for that performance, entered by a musician? If so that might be the composition. If the synthesiser is autonomous, maybe it's the composer. <caller_6> Are we still talking about works as nestable entities with different layers of abstraction? <pbryan> caller_6: Yes, it's certainly possible. <kepstin> caller_6, are they? There's really two things that might need some sort of work→work link <pbryan> The model allows for a web of ARs if we wanted. <kepstin> 1. Work A is a new work which derives from Work B <jdlh> Opera and musicals: the entire show is a work, but there are "songs" or "numbers" or "choruses" or "overtures" which are subdivisions that are works in their own right. <kepstin> 2. Work A is larger work which incorporates Work B <pbryan> I think that if song is a work, then probably musical number, movement, etc. should also be considered a work. I think if concept album is also a work, and so is opus, etc. then we're talking hierarchy. We'd want an AR to say work A is a part of work B. <kepstin> we don't want an AR that says "Work A is part of a set, the next Work in the set is Work B". Or do we? <pbryan> Maybe that too... :-( I frown because this can get very complex, very quickly. <caller_6> So, something like Composition>Derivative_Composition>Arrangement>Recording>Mix/Re-mix? <warp> hierarchy \o/ <pbryan> caller_6: That's one type of AR. <jdlh> [ Would like to add to agenda for later: cross-check, how can taggers use this Work concept we've sketched out to generate CSG-style Track Name strings and Album Name strings and Artist strings for music files? ] <caller_6> I guess that's what "granularity" means to me. Asking how many abstraction layers we want. <pbryan> jdlh: Noted, time permitting... So, there's two arms here: 1. Works and sub-works. <warp> caller_6: I don't think we need to define that, it will differ depending on the work <pbryan> 2. Works and derivatives. <jdlh> In Classical, Opera, Musicals, the sub-works are in a definite sequence. We need some way to preserve that ordering. Don't want the DB to list movement IV before movement II. <pbryan> To get concrete here, would there be disagreement Pink Floyd's The Wall is a concept album. <kepstin> for that matter, is a DJ-mix album of multiple tracks a work? <ruaok> pbryan: no. <ruaok> kepstin: yes. <pbryan> Is there any disagreement that The Wall should be considered a work? <ruaok> pbryan: not from me <pbryan> Is there any disagreement that Another Brick in the Wall , Part 2 should also be considered a work? (a track on The Wall) <ruaok> not from me. <pbryan> Anyone else/ ? <caller_6> agreed so far <jdlh> pbryan: not from me on Wall, or on Another Brick. <pbryan> Because if these propositions are true, then we certainly need work->sub work AR. * ruaok nods <kepstin> yes. that seems appropriate. * jdlh nods <caller_6> that would cover mash-ups as well? and the like? <pbryan> What about a run-of-the mill album? caller_6: Trying to build there. <ruaok> for a run of the mill album, there is a work, yes. <warp> how do we caputer order of sub-works under a work? <ruaok> and that is composed to sub-works. <jdlh> Good luck, y'all in pop music land, writing the Concept Album as Work style guideline. What is and isn't a Concept Album? <warp> capture <pbryan> warp: Good question... Answer may be that the NGS model doesn't really support that. <kepstin> I think that for most pop albums, people will not bother adding a super-work, because it doesn't add any additional useful information, and would be extra work. <caller_6> jdlh: if there needs to be a test, maybe it's that the work is often performed "in order" like Rush's 2112 <pbryan> Safe to say that if someone represents a pop album in works, it will/can be tolerated? <jdlh> caller_6: good observation <warp> pbryan: yes, any album is a work IMO. * nikki doesn't think any album is a work, we have releases for that <nikki> (and release groups) <warp> someone somewhere decided on a certain order for those tracks. <pbryan> nikki: Is there a way to make a distinction that we can use to guide? <jdlh> If "any album is a work" then no need for a Concept Album style guide. <caller_6> nikki: +1 <kepstin> basically, I only think a work is needed for a collection of works in the cases where either a) additional useful ARs can be added to the super-work, or b) the collection represents something different from the release. <nikki> pbryan: well, concept albums aren't something I know much about, I listen to stuff that's mostly "the artist's singles so far with some filler songs thrown in" :P but from what I've seen of simon's albums, they're often one "song" split into multiple parts and things like that <pbryan> Okay, so the consensus is concept albums, yes, typical pop albums, for the most part, no. If there's nothing of note in a pop album, it can probably be safely left out as a work. <jdlh> (Anybody have a URL which summarises the current and immediately planned NGS database schema, with technical detail? I'd like to look at it re: sub-work ordering.) <pbryan> So, works are multi-level. As warp brought-up, order of sub-works within a super-work is pertinent. And I don't think ARs as they stand can cope with this requirement. <ruaok> is it something that needs to be solved NOW? <jdlh> Works can't cut Gordian Knot of CSG without sequencing of sub-works. <pbryan> Can we live with sub-works being listed in non-sequential order? <nikki> pbryan: the only way to do ordering is like we did the part-of-set relationship <pbryan> nikki: Yeah, which presumes a work only appears in one and only one super-work. Is that a safe assumption? <nikki> but I think non-ordered works would be ok for now, ngs is never going to solve everything <kepstin> if we can't do ordering, then we're still going to need a CSG that encodes the order into the names of the sub-works :/ <pbryan> kepstin: I'm inclined to agree. <warp> pbryan: no <nikki> (and there'll be plenty of cleaning up to do in ngs anyway) <warp> pbryan: it is not uncommon for a track (a sub-work) to appear twice on the same album <pbryan> I think we're not going to fully solve the CSG problem with NGS as it stands. I think it's a step in the right direction. <jdlh> "Gordian Knot of CSG": refers to long post I sent to mb-style a couple of days ago. Basically, I think Works is our escape from the CSG mess, and if Works can't open the escape path, we should improve NGS until it does. <warp> pbryan: obviously it is common for a track (a sub-work) to appear on various albums <nikki> pbryan: yes, exactly. a step in the right direction :) <pbryan> jdlh: I agree in principle. I think it will likely happen in steps though. And the issues we encounter with works in the first step can inform future steps. <nikki> jdlh: there is always post-ngs though, just because it's not perfect for ngs doesn't mean it won't be improved later <pbryan> So, I think for the time being, we need to live without order-of-subworks. <kepstin> completely replacing a system with an entirely new one all at once really doesn't work as well as incremental steps, in my experience. <pbryan> But put it on the list of things we need post-NGS? <kepstin> for sure. <pbryan> Any major disagreement with that proposition? Going... going... ... (extra pause) ... okay. Sold. Okay, we've talked about hierarchy. <jdlh> Note that this NGS won't solve CSG mess, though. <pbryan> I totally agree. <ruaok> jdlh: its not intended to. its best to think of NGS as a new foundation with which we can solve CSG eventually. <pbryan> Okay, so we know we need a hierarchy. * jdlh resets his expectations. <pbryan> And we also want relationships to model derivative works, yes? <kepstin> yes, for sure. <caller_6> agreed <pbryan> So, there will be another hierarchy representing derivations. <kepstin> note that works may derive from recordings instead of other works directly - e.g. remixes/mashups/djmixes, etc. <pbryan> kepstin: Won't this also be represented within works though? Or do you see merely providing composer credit to the new work? <kepstin> kind of; a work may have multiple performances/recordings tho; and a remix is typically of a particular recording, and adds some new creative input. should it be a recording→work AR, or would work→work be enough? <pbryan> I think if it's a remix, I'd be inclined to say it's a new work, with composition credit to composers of the reused recorded material. That's how I most often see it credited on albums. So someone takes a Bee Gees track, remixes it with their new vocal track, B. Gibb gets composer credit. <kepstin> for the cases of derived works with work→work ARs, would we have to specify the composer on each work, or would that be inherited somehow? <pbryan> I don't think there's any inheritance exposed in the UI today... <caller_6> not all attributes would be inherited, would they? i mean, not necessarily. <kepstin> right, for example a cover of a song with new lyrics would have a different lyricist AR, but the others would be the same <pbryan> I'm wondering if there should be any derivative work inheritance in works.... Recording ARs seem to do a better job. <kepstin> the other issue is cancelling out previous ARs. for example an instrumental arrangement of a previously vocal song doesn't have lyrics, but is still a derived work so derivative work inheritance is probably too hard to get right automatically. <pbryan> Derivative-work ARs would need to distinguish what part is derived. Sounds ugly. <kepstin> so, duplicating ARs onto derived works is probably our best option right now, I think. <pbryan> Yeah, I think so. Any objections to this approach for NGS 1? Going... going... <kepstin> i'm still curious about whether remix works should use a work→work or recording→work ar, anyone have opinions on that? <jdlh> Would be nice to have better UI for assigning many ARs to many entities. <pbryan> Indeed. Okay, 8 mins to gloss-over classical/opera. <kepstin> jdlh, that's on the post-ngs wishlist already, I think. nikki really wants that :) <pbryan> Main problem: opera especially. <jdlh> This is obstacle now for classical, where same 10 Artists may get assigned to each track of 3 Releases. <pbryan> Opera splits tracks at arbitrary places in the libretto. * nikki wants lots of things :P <pbryan> I'm inclined to limit works to musical numbers, not recitatives, not random dialogue. <jdlh> Disagree. The recits matter also. I sing opera, I need to study the recits too. <pbryan> This implies that a track may link to the larger work, not the more granular in some cases. Okay, numbers and recits? <jdlh> In any case, there's no reason that a recording's Tracks will always conform to Sub-Work boundarie. <kepstin> or a recording may link to multiple of the more granular works, possibly? <pbryan> jdlh: Yes. kepstin: I thin so. s/thin/think/ <jdlh> So in general I expect cases where multiple Tracks make up one Sub-Work, and one Track contains multiple Sub-Works. <pbryan> jdlh: I agree. Anyone have any other points on this? <kepstin> I don't think it would make sense to add new works for each possible splitting of subworks into tracks. <caller_6> kepstin:+1 <jdlh> For Opera and Musicals especially, agree that we shouldn't create Subworks according to Track boundaries. <pbryan> So, we're admitting in some cases an ugly-at-best mapping of tracks to works, for opera especially. I think classical suffers this to some extent—sometimes multiple movements appear in the same track. <jdlh> I see SubWorks as being governed by cultural tradition, not recording engineers choices. <pbryan> jdlh: For opera/classical, I agree. It's an easier decision w. popular/jazz music I think. <jdlh> pbryan: yes, definitely classical suffers this too. <pbryan> Any last thoughts before the top of the hour? <jdlh> Idea about sequencing subworks: Tracks can link to subworks. Tracks have sequence. This imposes a sequence on subworks from a tagger's perspective. That may be enough to make progress with simplifying CSG. <kepstin> right, so in some cases you might have "Track A is a recording of Work A, Movement A and Work A, Movement B", other cases "Work A, Movement A has recordings Track A, Track B" <pbryan> jdlh: Agreed, this is a good direction to be driving toward. My plan going forward: another IRC this evening PST, then begin drafting a guideline, posting to email, continuing dialogue in email. <kepstin> hmm. that second case, the track names may need some sort of "(part x)" style. <jdlh> Work A, Movement A will potentially have hundreds of Tracks in different recordings. <ruaok> pbryan: sounds good. thanks for holding this meeting! <jdlh> pbryan: thank you! <pbryan> Your input here was very helpful everyone. Thanks for the input. Feel free to ping me directly with ideas pre-and post-email.
_______________________________________________ MusicBrainz-style mailing list MusicBrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style