From MusicBrainz Wiki
Jump to navigationJump to search

FRBR is short for Functional Requirements for Bibliographic Records, the title of a document released in 1998 by IFLA, the International Federation of Library Associations. FRBR specifies a data model for organizing bibliographic databases (who-wrote-what), as opposed to library catalogs (inventories of books). It should be possible to apply these principles to discography databases as well, including MusicBrainz. For reading up on FRBR, a good start could be Wikipedia's introduction, or What is FRBR?. You can also read the full document from IFLA's website. If you google for FRBR there are plenty of easier-to-read presentations of the core ideas.

What it says

FRBR specifies a data model consisting of three groups of entities:

  • Products, i.e. books, films and records. These follow a hierarchy of four entities:
    • Works, e.g. the novel Robinson Crusoe by Defoe
    • Expressions, e.g. the written novel, a translation, a film script based on the novel, an illustrated abridged version for children
    • Manifestations, e.g. a print edition of this translation of that novel
    • Items or physical copies, worn and torn, one of three held by your local library, each having a unique inventory number
  • People or corporate bodies (i.e. artists, groups, record companies) responsible for each entity of a product, e.g. the author of a story, the translator, the printer, the publisher, the library that owns a copy
  • Subjects, e.g. events and places where the story takes place, the date when the book was printed, etc.


Despite this spec being almost a decade old, nobody in the library world has really implemented it yet. It is on the verge of becoming the "flying car" of library catalogs, a utopian dream never fulfilled. The Library of Congress catalog is still structured as a bunch of MARC records, which is little more than a 1960s digital equivalent of a 19th century card catalog. One reason for this delay is that FRBR doesn't specify who is going to do the job, it only says how it ought to be done. (And then the Y2K bug and dotcom crisis got in between.) Traditionally every library has a catalog of the books it owns. The idea that information about books (bibliography) could be separated from the local inventory was all new to libraries in the 1960s, when cooperative institutions like OCLC were born. With time these co-ops have started to act more and more like private monopoly suppliers that libraries depend on but really hate. Here, have some bibliographic data, the first samples are free, then you have to pay your soul, and we will supply your institution for the rest of its life. Throw in a lifetime subscription to the Encyclopædia Britannica as well.

Enter the 21st century and Wikipedia. The common knowledge needs not be owned by a monopoly supplier. Every private music collector and library can keep their local inventory in the shape of links to existing, shared information at MusicBrainz. Find an error, fix it in the shared pool, not in your local inventory. This is how OCLC and Gracenote work. Except that no single entity owns Wikipedia or MusicBrainz, because the contents can be copied freely by anybody. This answers the who question.

Fortunately, Wikipedia and MusicBrainz were created not by the tired old people who wrote the FRBR specification, but by fresh minds from the filesharing generation. Even if this means some shortcomings in the initial data model (just albums, tracks, and artists), leaving plenty of room for improvement, the important difference is that they are getting the job done, as opposed to just theorizing about it.

Applicability to MusicBrainz

FRBR specifies a data model consisting of three groups of entities:

  • Group 1
    • Works, These would be compositions e.g. "One" (by Bono)
    • Expressions, These would be recordings e.g. "Original studio recording of One", "2006 recording with Mary J Blige" (this?), "Live recording at ..."
    • Manifestations, These would be the albums that the recordings appear on e.g. "Achtung Baby", "U2, Best of 1990-2000". A manifestation can contain multiple expression, which can in turn be of different works (i.e. these are the tracks of the album).
    • Items Not relevant to musicbrainz, but potentially relevant in tagger software - these would be individual copies of the albums - RjMunro's copy of "Best of 1990-2000". Argubly, downloaded or ripped music files (MP3s etc.) jump over the manifestation category. Perhaps iTunes is a manifestation of all tracks downloadable from it.
  • Group 2
    • People i.e. individual artists
    • Corporate bodies i.e. bands and record companies
  • Group 3 Subjects, This kind of provides all the AR type stuff not covered already :-)