I said « there is no rush merging recordings and once it’s wrongly merged, a split means that we have to remove every then made ambiguity information like ISRC, PUID, durations, arranger, producer, instr. voc. perf. ARs and recheck everything for each new split MB recs. »
Merging recordings from simultaneous release variations is one thing (ok) but so is not the hey-saw-that-song-same-name-same-diuration-guys☞MERGETHEMCRAZEYAY !
Because same name same duration doesn’t mean anything to start with (examples I got).
OK be warned pipeule, I’ve given up trying to block this kind of shit from happening down here.
I only tried one week and it’s impossible as the flood is massive guys.
Tell me if there are still some no votes lingering from today so I cancel them right away (and unsubscrube maybe).
Do them merges, do them if you don’t care information loss. If only what counts is recs-list-size-little--near-work-list-size. If you just think “long list = mess” and not “corrupt recordings = mess”.
Here is my crappy stuttering http://chatlogs.musicbrainz.org/musicbrainz/2011/2011-09/2011-09-07.html#T08-47-18-608489
Now what I do is BLINDELEY add MY AR (like master, produce etc.) on MY TRACKS whatever the merging mess that has, is or will be going on behind them.
Because don’t count on me splitting shit now.
If pipeule choose no worry, I will too.
That relieves me so much.
Fortunately I haven't seen many bad merges yet (and most of the ones I saw were wrong auto-merges during the conversion to NGS), but the current situation is certainly not optimal.
(random semi-related musings follow)
The general consensus seems to be that "album" or otherwise original versions should have no recording comment, but especially for popular songs which appear on many obscure compilations (e.g. http://musicbrainz.org/artist/34cf95c7-4be9-4efd-a48a-c2ea4a0bb114/recordings), this makes it harder to distinguish known-album-version recordings from unknown-version ones and increases their risk of being wrongly merged. Perhaps it would be better to add such a comment in these cases.
Another idea I remember seeing somewhere (this list? chat?) was to have a single "unknown version" recording to use when the version is not known; we could then merge the recordings we don't have enough information about, and when in the future we have data showing which version is on a particular release, we can change that one individually.
When MB or related services store ISRCs and audio fingerprints for songs, it would be helpful to split wrongly-merged tracks if for each submission we could see which release it came from, not only the associated recording. Then if a recording has two codes or very different fingerprints, we could see which releases contain each version.
I do think we still need to specify more clearly what constitutes a unique recording. Different edits should certainly be separate ones, but if two recordings are the same version but have different mixes or fades, when should they be merged (http://musicbrainz.org/edit/15028033)? A vinyl album and its CD version usually share the same recordings, but if the album is later remastered or redigitized, should it be a separate recording? etc.
2011/9/7 jesus2099 <[hidden email]>
I said « there is no rush merging recordings and once it’s wrongly merged, a
MusicBrainz-style mailing list
|Free forum by Nabble||Edit this page|