alastair
commented
9 years ago A change in musicbrainz policy means that remasters should have the same mbid. This means that we can have two recordings with different audio and the same mbid. We can't store them both in the docserver.
Options:
- Store just 1
- disambiguate with the release id (we could have a remaster on the same release though)
- disambiguate with track id (guaranteed unique, but isn't really exposed in musicbrainz or dunya)
- Use track id in the docserver without using recording id
My current preference is to use track id, probably in the docserver
I have just stumbled upon two historical recordings http://musicbrainz.org/recording/809a61fc-fc80-4bb3-9f27-01133c2341ad and http://musicbrainz.org/recording/3464258b-bcfe-4fcc-ac1b-5b4b5fd6f0ba . They are actually the same but have different mastering. MusicBrainz has changed its policy to merge such recordings (https://musicbrainz.org/doc/Style/Recording#Recordings_with_different_mastering), so I merged them. Nevertheless from a research point of view it might be interesting to have access to them (e.g. does audio-score alignment perform similarly?).
When we query to get an mp3 (docserver.get_mp3), we ask it with it's MBIDs. Can we have an optional input to download all the recordings having that MBID?