iTunes in depth: working with MP3 ID3 tags

While housecleaning in our MP3 collection recently, we stumbled upon a disturbing behavior in iTunes 1.1: modifying song information without asking us beforehand. We realize that iTunes uses CDDB to look up track information for CDs, but we were surprised when it updated track information for MP3 files, in several cases overwriting information we had added ourselves. We do not know if the 'updated' data came from the song file itself or from a network source such as CDDB.

First, a little background. MP3 files can store information about the artist, track name, and other descriptive information in an ID3 tag. (See Apple KB article 25293. Also, this utility can edit ID3 tags under Mac OS X.) When you drag a folder of MP3 files into iTunes main window, iTunes scans the folder for MP3 files and imports the contents of each file's ID3 tag in order to populate the music database. Sometimes these tags become messed up, or were never created in the first place, in which case iTunes only has the MP3's Mac filename to use as an identifier.

We wanted to use our desktop Mac's MP3 collection on a laptop, so we enabled file sharing for the desktop drive and dropped a remote folder of songs on the laptop's copy of iTunes. It turned out that many of the songs did not have ID3 tags, so instead of looking like the desktop Mac's iTunes list, a large number of songs had descriptive names such as "track 23 - JB.mp3," which is why we needed to do a little housecleaning.

In iTunes on the desktop Mac, we manually fixed spelling errors, filled in missing data, and generally made cleaned up the music list. Next, we selected all the songs in the list, and using the Convert ID3 Tags item in the Advanced menu, had iTunes create v2.3 ID3 tags for all the songs in our collection, a step we had never bothered with previously. After this process, when we dragged the shared folder of songs onto iTunes on the laptop, all the music information was recreated properly, mirroring our desktop Mac. That should have been the end of this narrative.

However, on the desktop Mac, we noticed something we have never seen before, which we suspect was triggered by creating the ID3 tags. When a song began playing, occasionally text would appear in the previously blank genre and/or album fields. After a little experimenting, we were able to determine that the song information was being updated the first time the song was played. Sometimes it filled in blank fields, and other times it overwrote whatever we had typed in manually.

The updated information seemed limited to more recent songs, and was never incorrect, although it often did not match what we had previously entered. Also, if we edited a track's info after an update, the change would stick. We are not aware of any documentation that details iTunes song recognition, and even if there is and we just missed it, the fact that iTunes overwrites user-entered data is a no-no. Hopefully this will be addressed in the forthcoming release of iTunes 2.

Update: James Huston writes: "iTunes, like SoundJam before it, encodes songs with the unique CD ID data in a special ID3 tag in case it is not able to perform a lookup prior to encoding. This data can be used at a later date to look up information from the CDDB to tag your files. Only ID3 versions 2.2 and 2.3 support this feature and iTunes defaults to 2.2. So I wouldn't be surprised if iTunes looked up this information automatically during the editing."