Re: [mb-style] RFC: Improve Guidelines on Merging Recordings
having dozens of recordings that might or might not be unique (but are represented as unique due to separate recordings) in combination with the difficulties of not being able to share metadata between them is making me feel much more uncomfortable. +1 so much. Redundancy is bad, BAD. Especially when there is no good/efficient way to handle it. -- Sent from my Android phone with K-9 Mail. Please excuse my brevity. lorenz pressler l...@gmx.at wrote: 3. Versions of a release that are mastered in audibly different ways should use separate recordings.4. If audio restoration is used to create remastered audio from an earlier recording, a track containing the remastered audio should use a recording specific to that remaster. isn't 3) already covering 4)? or is that to emphasize that if a release is promoted as 'remastered' that we should create new recordings even if there is no audible difference? usually if i have reliable remaster information i create/not-merge recordings already (despite the fact i usally don't hear a difference). however i'm not sure what to do with recordings on generic samplers/compilations that don't have any information and where we most likely never get any more detailed information. i'll have to admit that i don't follow a consistent editing behaviour in these cases. in some cases i use label/publisher/year information to find the most likely (re)master (if a label/publisher had some recordings remastered all later released compilations will likely use that remaster); if compilations don't have any label/remaster/publisher information that correspond to any existing remasters i usually merge according to acoustIDs or i use the recording with the best metadata as a catch-all. i see the problem of invalidating MBIDs for recordings as a result of permissive merging not 'that' much of a problem as it would be for artists or release(-groups); having dozens of recordings that might or might not be unique (but are represented as unique due to separate recordings) in combination with the difficulties of not being able to share metadata between them is making me feel much more uncomfortable. best, lorenz. ___ MusicBrainz-style mailing list MusicBrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style ___ MusicBrainz-style mailing list MusicBrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Re: [mb-style] RFC: Split No linguistic content lyric language
So, I was reading some music metadata stuff, from the Association of Musical Libraries, and found this: zxx should not be confused or combined with code und (“undetermined”) which is used both when the language cannot be identified and when sung or spoken text is “vocalises, humming and other texts that are wordless or consist of nonsense syllables.” Which would suggest we just have to activate und, and use that for that stuff - that's easy enough, we can get it activated this weekend if people agree with the libraries. -- Nicolás Tamargo de Eguren ___ MusicBrainz-style mailing list MusicBrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Re: [mb-style] RFC: Split No linguistic content lyric language
Le 8 nov. 2012 21:19, Nicolás Tamargo de Eguren reosare...@gmail.com a écrit : So, I was reading some music metadata stuff, from the Association of Musical Libraries, and found this: zxx should not be confused or combined with code und (“undetermined”) which is used both when the language cannot be identified and when sung or spoken text is “vocalises, humming and other texts that are wordless or consist of nonsense syllables.” Which would suggest we just have to activate und, and use that for that stuff - that's easy enough, we can get it activated this weekend if people agree with the libraries. Yes but the documentation must be very clear or editors will start using it when they don't know which language was used. ___ MusicBrainz-style mailing list MusicBrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Re: [mb-style] RFC: Improve Guidelines on Merging Recordings
Le 8 nov. 2012 14:18, Profpatsch m...@profpatsch.de a écrit : having dozens of recordings that might or might not be unique (but are represented as unique due to separate recordings) in combination with the difficulties of not being able to share metadata between them is making me feel much more uncomfortable. +1 so much. Redundancy is bad, BAD. Especially when there is no good/efficient way to handle it. -- Sent from my Android phone with K-9 Mail. Please excuse my brevity. lorenz pressler l...@gmx.at wrote: 3. Versions of a release that are mastered in audibly different ways should use separate recordings.4. If audio restoration is used to create remastered audio from an earlier recording, a track containing the remastered audio should use a recording specific to that remaster. isn't 3) already covering 4)? or is that to emphasize that if a release is promoted as 'remastered' that we should create new recordings even if there is no audible difference? usually if i have reliable remaster information i create/not-merge recordings already (despite the fact i usally don't hear a difference). however i'm not sure what to do with recordings on generic samplers/compilations that don't have any information and where we most likely never get any more detailed information. i'll have to admit that i don't follow a consistent editing behaviour in these cases. in some cases i use label/publisher/year information to find the most likely (re)master (if a label/publisher had some recordings remastered all later released compilations will likely use that remaster); if compilations don't have any label/remaster/publisher information that correspond to any existing remasters i usually merge according to acoustIDs or i use the recording with the best metadata as a catch-all. i see the problem of invalidating MBIDs for recordings as a result of permissive merging not 'that' much of a problem as it would be for artists or release(-groups); having dozens of recordings that might or might not be unique (but are represented as unique due to separate recordings) in combination with the difficulties of not being able to share metadata between them is making me feel much more uncomfortable. best, lorenz. But eliminating redundancy would be at the price of loss of data. So which is worse? Either we keep too many or we merge too often. Our problem here is we are unable to find the limit, not even approximately. Another issue is that by merging recordings we are doing irreversible damage. AFAIK once recordings have been merged, there is no way to restore the AcoustIds again to their original releases (or more exactly to their tracks). Maybe again Kuno's suggestion is good. I know, I already changed my mind once before about this :-) ___ MusicBrainz-style mailing list MusicBrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Re: [mb-style] RFC: Improve Guidelines on Merging Recordings
On 11/08/2012 02:39 PM, Frederic Da Vitoria wrote: Another issue is that by merging recordings we are doing irreversible damage. AFAIK once recordings have been merged, there is no way to restore the AcoustIds again to their original releases (or more exactly to their tracks). Maybe again Kuno's suggestion is good. I know, I already changed my mind once before about this :-) This was brought up in IRC the other day.[1] Luks basically said that it wouldn’t actually be that big a deal to move acoustIDs from recordings to tracks, because “most recordings/fingerprints have only one release so those are not affected by these changes”. “tracks” in this context means “releaseID+disc#+track#” — this would cause problems when tracks get reordered or renamed because they were incorrect, but discussion indicated that that was probably the best way to do it — see IRC log for details. 1. http://chatlogs.musicbrainz.org/musicbrainz/2012/2012-11/2012-11-05.html#T16-27-14-253884 signature.asc Description: OpenPGP digital signature ___ MusicBrainz-style mailing list MusicBrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Re: [mb-style] RFC: Improve Guidelines on Merging Recordings
Made another update, hopefully the final one. Expected expiration for RFC: Sunday, 11th November 2012 Changes: - Removed the specific text about remasters, and included it in case 3. - Added a case back in about conflicting relationships, which mysteriously got lost when I was writing the proposal. - Rewrote all the specific cases to remove all references to tracks, and releases. Only recordings are mentioned now. - Removed as much technical terminology as I could. The most technical parts now are audio channel and mastered. Extending the RFC until Sunday for now, just in case there are any further suggestions. If discussions are ongoing then, we'll keep going until everyone's happy. If not, I hope to move to RFV by Monday, if there's a +1 by then. -- View this message in context: http://musicbrainz.1054305.n4.nabble.com/RFC-Improve-Guidelines-on-Merging-Recordings-tp4644561p4644894.html Sent from the MusicBrainz - Style mailing list archive at Nabble.com. ___ MusicBrainz-style mailing list MusicBrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Re: [mb-style] RFC: Improve Guidelines on Merging Recordings
Just as an overview, the proposal now makes these changes: - Extends the first paragraph to include a clearer summary of what a recording should actually represent. - Completely rephrases all the existing examples to make them clearer and give fuller explanations. - Adds cases concerning silence at the start and end of recordings and different numbers of channels in recordings. - Explains the flaws of ISRCs and AcoustIDs in more detail, and stresses that they shouldn't be used as the only reason for a merge. Please let me know if there's anything else that should be changed, or else feel free to show your support for the proposal as it is by giving a +1. Thanks for all your suggestions so far! -- View this message in context: http://musicbrainz.1054305.n4.nabble.com/RFC-Improve-Guidelines-on-Merging-Recordings-tp4644561p4644895.html Sent from the MusicBrainz - Style mailing list archive at Nabble.com. ___ MusicBrainz-style mailing list MusicBrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style
Re: [mb-style] RFC: Improve Guidelines on Merging Recordings
What we have here is a conflict of interest between two different groups on what unique, distinct audio (or whatever we might call it ) is. Both groups, I guess, would probably include each about 5% of the editor and users – while the remaining 90% wouldn't care about the subtleties of each of them… Forgive me if I'm (only slightly) caricaturing them for better visibility ;-) The first group, let's call them the audiophiles are those that agree with this RFC: They listen to their music only on high-end machinery, tracking the slightest difference between different masters and remasters, pressings and re-pressings, differences that may eventually, if at all, be catched by ear on such high-priced audio equipment or only distinguished by the use of technical devices, the comparison of AcousticID or fingerprinting. They are vaguely aware that something as deprecated as vinyl LPs, cassette tapes or even shellac did exist in the ancient times, or if they still listen to vinyl it would be the 180g collector's pressings. Their interest is to keep apart all these versions which they consider different. The other group, what I'd call the historians, listen to quite different categories of music, where other things matter. They still cherish their shellac and vinyl collections, adding these to MB. They know that a given track has – in the thirties and forties – been recorded several times by the same artist for different labels, with different arrangements and different orchestras. They will want to keep track of the similarities of these divergent audio masters over the years. They are aware that technical progress and the introduction of better methods and supports will make this original master from 1928 sound differently whether you listen to the original shellac from the late twenties, the re-release on vinyl in the fifties, the compilation on tape in the seventies and the various transfers to digital data from the mid-eighties onwards. But they don't care (or at least not that much) about subtle sound differences because they are used to poor sound quality anyway, be it for the pretty worn out shellac they found on the flea market or the early transfers from equally poor sources on CD. For them all these are emanations of the same unique audio-master, which belong together and have to be kept apart a) from the previously unreleased second take from the same session and b) from the recordings of the same song made by the same artist in 1930, 1935 and 1939. In the opinion of these historians two different digital transfers of the original version of e.g. Comme un moineau by Fréhel, recorded in November 1933 and released as Ultraphone AP 1320, matrice P 76929 (1) have to be considered unique audio recordings that should be merged, even if the first transfer was has been made in the early nineties and the other one 10 years later, with improved technology, and even if the source may not even be taken from be the same (lost) master, but from two distinct copies of this original shellac release from early 1934. I see the legitimacy of both approaches, depending also on the musical genres one works on. But I find a guideline that completely ignores the historical view unacceptable. I acknowledge the massive improvements made since the first version of the guideline. But the new version still says Do not merge anything at all, unless you have in hands every single version you plan to merge, have listened carefully to all of them to detect eventual differences and have personally established that the acoustic ID of all of them is perfectly identical. And what the guideline has to say about AcousticIDs and ISRC eventually being wrong would translate, if taken seriously, into: Do not make any edits based on any sources at all, because these sources might be wrong ;–) It forbids, if only implicitly, the merge of vinyl and digital data (this should, IMO, been explicitly allowed, since no acousticID or similar is included with data for analog releases), it leaves open the problem of the many un-sourceable VA-sampler/compilation tracks and it still relies (in point 2 and 3) on subjective, unsourced criteria. Seen the additional work this implies for editors when there are dozens of unmerged (and, based on this guideline, un-mergeable) recordings I don't think this should pass. (1) http://musicbrainz.org/work/2a163e44-e9b1-4288-8dff-18d8679c1700 Chris/chabreyflint On Thu, Nov 8, 2012 at 9:39 PM, Frederic Da Vitoria davito...@gmail.comwrote: Le 8 nov. 2012 14:18, Profpatsch m...@profpatsch.de a écrit : having dozens of recordings that might or might not be unique (but are represented as unique due to separate recordings) in combination with the difficulties of not being able to share metadata between them is making me feel much more uncomfortable. +1 so much. Redundancy is bad, BAD. Especially when there is no good/efficient way to handle it. -- Sent from my Android phone with K-9
[mb-style] RFC: Improve Guidelines on Merging Recordings
2012/11/9 SwissChris swissch...@gmail.com What we have here is a conflict of interest between two different groups on what unique, distinct audio (or whatever we might call it ) is. Both groups, I guess, would probably include each about 5% of the editor and users – while the remaining 90% wouldn't care about the subtleties of each of them… Forgive me if I'm (only slightly) caricaturing them for better visibility ;-) The first group, let's call them the audiophiles are those that agree with this RFC: They listen to their music only on high-end machinery, tracking the slightest difference between different masters and remasters, pressings and re-pressings, differences that may eventually, if at all, be catched by ear on such high-priced audio equipment or only distinguished by the use of technical devices, the comparison of AcousticID or fingerprinting. They are vaguely aware that something as deprecated as vinyl LPs, cassette tapes or even shellac did exist in the ancient times, or if they still listen to vinyl it would be the 180g collector's pressings. Their interest is to keep apart all these versions which they consider different. The other group, what I'd call the historians, listen to quite different categories of music, where other things matter. They still cherish their shellac and vinyl collections, adding these to MB. They know that a given track has – in the thirties and forties – been recorded several times by the same artist for different labels, with different arrangements and different orchestras. They will want to keep track of the similarities of these divergent audio masters over the years. They are aware that technical progress and the introduction of better methods and supports will make this original master from 1928 sound differently whether you listen to the original shellac from the late twenties, the re-release on vinyl in the fifties, the compilation on tape in the seventies and the various transfers to digital data from the mid-eighties onwards. But they don't care (or at least not that much) about subtle sound differences because they are used to poor sound quality anyway, be it for the pretty worn out shellac they found on the flea market or the early transfers from equally poor sources on CD. For them all these are emanations of the same unique audio-master, which belong together and have to be kept apart a) from the previously unreleased second take from the same session and b) from the recordings of the same song made by the same artist in 1930, 1935 and 1939. In the opinion of these historians two different digital transfers of the original version of e.g. Comme un moineau by Fréhel, recorded in November 1933 and released as Ultraphone AP 1320, matrice P 76929 (1) have to be considered unique audio recordings that should be merged, even if the first transfer was has been made in the early nineties and the other one 10 years later, with improved technology, and even if the source may not even be taken from be the same (lost) master, but from two distinct copies of this original shellac release from early 1934. I see the legitimacy of both approaches, depending also on the musical genres one works on. But I find a guideline that completely ignores the historical view unacceptable. I acknowledge the massive improvements made since the first version of the guideline. But the new version still says Do not merge anything at all, unless you have in hands every single version you plan to merge, have listened carefully to all of them to detect eventual differences and have personally established that the acoustic ID of all of them is perfectly identical. And what the guideline has to say about AcousticIDs and ISRC eventually being wrong would translate, if taken seriously, into: Do not make any edits based on any sources at all, because these sources might be wrong ;–) It forbids, if only implicitly, the merge of vinyl and digital data (this should, IMO, been explicitly allowed, since no acousticID or similar is included with data for analog releases), it leaves open the problem of the many un-sourceable VA-sampler/compilation tracks and it still relies (in point 2 and 3) on subjective, unsourced criteria. Seen the additional work this implies for editors when there are dozens of unmerged (and, based on this guideline, un-mergeable) recordings I don't think this should pass. (1) http://musicbrainz.org/work/2a163e44-e9b1-4288-8dff-18d8679c1700 Chris/chabreyflint Yes, I believe you accurately described the issue. It means the 2 groups understand something different in the word 'recording'. I believe this means we miss a concept or level. This could the recording group. Restoring the tracks to something closer to pre-NGS, as Kuno suggested would also work until the audiophiles find a definition of the limit between what they think should be separated and what should be merged. -- Frederic Da Vitoria (davitof) Membre de l'April - « promouvoir et défendre le logiciel libre
Re: [mb-style] RFV: Officially deprecate the cover art relationship
It's been 48 hours without a veto, so this has passed. Nikki On 8 November 2012 02:37, Nikki aei...@gmail.com wrote: On 7 November 2012 09:29, Staffan Vilcans lift...@interface1.net wrote: Perhaps depend on http://tickets.musicbrainz.org/browse/MBS-4641 as well to make it easier to add cover art. I definitely agree that that would be nice to have (I already voted for it :)), but I don't think it should be a dependency of this proposal unless we have a reason to believe a lot of people are preferring to add covers using the cover art relationship rather than the CAA because they only have to paste a URL for the cover art relationship. That's not the conclusion I would come to looking at the numbers[1] though, since the number of CAA uploads completely dwarfs the number of new cover art relationships. Nikki [1] In the last eight weeks, according to the edit search: - 21 open/accepted edits to add cover art relationships (2 or 3 per week) - 27,972 open/accepted edits to add images to the CAA (~4,000 per week) - 4,897 open/accepted edits to add Amazon links (~610 per week) i.e. for every single new cover art relationship, ~1,330 CAA images and ~230 ASINs are added. ___ MusicBrainz-style mailing list MusicBrainz-style@lists.musicbrainz.org http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-style