freedb

» see freedb.org as it lived, on the wayback machine

~ the eulogy ~

the first time i watched it work i did not understand that anything had happened. you put a cd in the drive, opened whatever ripper was on the machine, and the track names filled themselves in. song titles in order, with the album across the top. it read like the disc was telling the computer who it was.

it was not. a standard audio cd has no text anywhere on it. what the drive can see is how many tracks there are and where each one starts, and that is the entire fingerprint. the ripper turned those numbers into a disc id, sent the id over your connection, and asked a database in another country whether anyone had seen this exact arrangement before. on a turkish dial-up line in those years that lookup was the fastest thing the whole operation did, because it was a few hundred bytes against forty minutes of encoding.

the answers came from people. someone had put that same album in a drive earlier, typed the titles in by hand, and submitted them. the database was a pile of unpaid typing that had accumulated since the mid nineties, and by april 2006 it covered just under two million discs.

freedb held the free copy of that pile from 2001. magix, a german software company, bought it in 2006 and kept it running quietly for another fourteen years before putting up a notice that said freedb.org and its services will be shut down on march 31st of 2020.

~ the dossier ~

~ what it actually was ~

the original was cddb, built in the mid nineties around ti kan's xmcd player and steve scherf's database. it solved a problem the compact disc standard had simply not considered, since in 1982 nobody designing it expected the disc to end up inside a general-purpose computer with a network connection.

the design was small and it worked. the client computed an id from the table of contents and the server sent back a text record. if the server had never seen that disc before, you were invited to fill it in for whoever came next. almost every music program of the era spoke this protocol, which is why the behavior felt like a property of cds rather than a service that could be withdrawn.

~ the fork ~

it could be withdrawn. the company behind cddb went commercial and became gracenote, and between 2000 and 2001 it closed the door on the software that had been feeding it. in march 2001 unlicensed applications were cut off from the database, new licenses for the old protocol stopped being issued, and a paid replacement was offered instead.

the objection at the time was not really about money. the records had been typed by volunteers who were never asked whether their work could be enclosed. so a copy taken just before the lock went on became freedb, published under the gpl, and open source players repointed themselves at it within months. it is one of the cleanest examples of a community keeping its own contribution after the company holding it changed its mind.

~ where the names went ~

by 2020 the ripping era was over. streaming had made the question of what is on this disc largely academic, and magix said the legal and technical upkeep no longer paid for itself. the notice went up months ahead, the service answered its last lookups on 31 march 2020, and the servers were finally dark by that summer.

the data survived. gnudb.org took the collection and still serves it, and musicbrainz had already built the more ambitious replacement years earlier, having closed its own freedb gateway in march 2019. what is gone is the reflex. there was a stretch of about fifteen years where a machine could look at a silent disc and tell you what was on it, and the reason was that thousands of people had typed it in for nothing.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »