Continuation of MusicBrainz Discovery (I).

One notable thing about MusicBrainz is that Rob Kaye and a small number of core developers and supporters have pursued a consistent vision for roughly six years with very little funding or even understanding outside this small group. It isn’t easy to really “get” MusicBrainz (I think it took me two years), though I think that at some point in the next few years everyone will “get” MusicBrainz more or less all at once.

If you’re a geek it’s hard not to get hung up on MusicBrainz use of acoustic fingerprint-based technology. Acoustic fingerprinting is fragile in three ways — it is subject to false positives and false negatives, there is no open source implementation of the concept, and the technology MusicBrainz uses, Relatable TRM, is proprietary and requires a centralized server. Indeed, many of the technology questions at Tuesday’s music metadata panel concerned acoustic fingerprints.

It is important to understand that while MusicBrainz uses acoustic fingerprints, it does not rely on them. TRM matching is just one mechanism for track identification. File metadata included in (e.g., ID3 tags) or with (filename) the file can and I believe are used to match existing records, as could track duration and file hashes (see if Bitzi or a P2P network has any metadata for the file in question). Additionally, file identification is only one component of MusicBrainz.

If you’re not a geek, you won’t notice acoustic fingerprints, because you wouldn’t, and because you’re not likely to get that far. So what the heck does MusicBrainz do? Here’s an attempt:

• MusicBrainz can organize your music collection. Download the tagger.
• MusicBrainz uniquely identifies artists, albums, and songs, facilitating rich and precise music applications, all on a level playing field.
  • Not at all speculative potential: include a MusicBrainz song identifier in a blog post, cover art (with your Amazon referrer of course) automagically appears in blog post, blog aggregator publishes top n lists and personalized recommendations.
  • Another: publish a playlist of MusicBrainz identifiers and others can recreate the experience so defined with no file transfer involved.
  • There are several others, some that could be offered by MusicBrainz itself, outlined in MusicBrainz tomorrow. I have to quote one because it’s fun:
    

    Music Genealogy: MusicBrainz may keep track of which artists/performers/engineers contributed to a piece of music, and when these contributions took place. Combining this contribution data with data on how artists influenced each other will create a genealogy of modern music. Imagine being able to track Britney Spears back to Beethoven!

• The MusicBrainz database, created by the community, will remain free, unlike others.

Having been around for awhile, MusicBrainz has run into many of the technical and social problems inherent in music metadata and an evolving community website, and produced much good documentation on solutions, realized and potential. Here’s a sampling:

• Data License Issues and the solution.
• Style Guidelines — you wouldn’t imagine how hard good capitalization rules are, for starters. Ugh. Don’t even attempt to think about Classical metadata!
• Cross-platform desktop app, using wxPython, same as Creative Commons.
• How to identify a compact disc.
• Tracking relationships among tracks, albums, artists, and much more.
• Survival of the fittest and steps towards survival of the fittest — how do you take some wiki lessons and apply to a large and highly structured database and featureset?
• MusicBrainz Metadata Initiative — web services and RDF, hand rolled.
• MusicBrainz Non-Profit White Paper, already linked above this list is very interesting on several subjects.
• How the server works.
• Everyone loves stats.

By the way, as of Wednesday MusicBrainz has a blog.