When the developer Erik Kemp designed the first metadata system for MP3s in 1996, he provided only three options for attaching text to the music. Every audio file could be labeled with only an artist, song name, and album title.
Kemp’s system has since been augmented and improved upon, but never replaced. Which makes sense: Like the web itself, his schema was shipped, good enough, and an improvement on the vacuum which preceded it. Those three big tags, as they’re called, work well with pop and rock written between 1960 and 1995. This didn’t prevent rampant mislabeling in the early days of the web, though, as anyone who remembers Napster can tell you. His system stumbles even more, though, when it needs to capture hip hop’s tradition of guest MCs or jazz’s vibrant culture of studio musicianship.
And they really, really fall apart when they need to classify classical music.
Digital music software has never been well adapted to classical music. iTunes only added a “Composer” tag in 2004. Two years later, the developer Stan Brown published a guide on his website to “taming” Apple’s software for classical music. Since then, other hacks and kludges have followed. But digital music was so endlessly efficient that classical fans, especially the younger ones, embraced it and adopted it.
Certainly that was the case for me, a music-school student in the late 2000s. In the waning hours of a camp or festival, you’d squat with the other kids on the carpet, set up Firewire cables between everyone’s laptops and hard drives, and move music files back and forth. (I still have a DRM-locked copy of Candide in my library that will forever require Will Carmichael’s iTunes password, which I do not know.)
But as streaming services have flooded out MP3s, the situation worsened. Apple, which long paid classical more mind than other big tech companies, debuted Apple Music with dismal classical options, as NPR’s Anastasia Tsioulcas and The New Yorker’s Alex Ross have documented.
And even beyond the streaming service, the new version of Apple’s signature music software seems especially broken. In the name of creating a “complete thought around music,” iTunes 12 has crammed a streaming service and a media library and a recommendation service and a file store and a device manager into one interface. The sum is that nothing “just works”—and MP3s especially don’t work well. I wondered: How were professional musicians handling the change? And how did they organize their music in the first place?
After all, their job requires them to have quick and fluent access to a library of working recordings.
I had another interest here. Earlier this month, Apple released the new and likely final versions of its iPod Nano and iPod Shuffle. They shipped with interfaces that looked more than three years out of date. The blogger John Gruber said that this was because “what remains of the iPod software team” had been absorbed by the Apple Watch division. Apple has already discontinued its iPod Classic, the last media player that could conceivably let you tote around your entire music library in one device. The company is floating to a streaming model.
If classical listeners are ill-served by streaming services, though, they will stick with music files; and that means they represent, as a bloc, the set of listeners who will continue to maintain personal libraries of owned music even as the larger public rents their digital music instead. How are they adapting?
* * *
As it turns out: poorly.
First, the bugs. The new version of iTunes disappears music. It confuses live tracks and studio versions. And its search bar cannot even find songs which it contains in its library.
“To give you a really specific situation, there are two settings of the Te Deum text by Benjamin Britten. And it would seem to me that if you type in ‘Britten’ and ‘Te Deum,’ you would see some of them,” the composer Nico Muhly told me. “But it says, ‘no results found.’”
I want to submit to the record here that Muhly’s hard drive contains seven different files that could be reasonably called the Britten Te Deum. In fact, it contains more than 2,000 files, or 11.9 gigabytes, of music by Benjamin Britten. It also contains 97 different settings of the Te Deum text.
“What’s extraordinary about it is that I tagged everything really, really well. It’s in Artist, Album Artist, all these things are organized,” he said.
But when “Britten Te Deum” is searched—and he sent me a screenshot of this—nothing comes up. “It’s not like, let me show you too many results. It just does not compute.”
(Even when the search function does locate a file, he says, pressing ‘return’ to play it does not start playing the highlighted file, but the first file listed alphabetically in iTunes. “Which of course is only Aaliyah.”)
Timo Andres, a composer and pianist, reported fewer problems with the new version of iTunes, though he echoed that it had some odd behaviors. With his music files, he had moved the composer to the “Artist” field. The performer, conductor, and other information went to the file’s general comments field. “I started to do that because there is no performer field,” he said.
Andres has also started to use Apple Music, but has stumbled again with labeling. Apple Music only searches for songs in its library by Kemp’s original three tags—artist, album, and song title—which, combined with often-flawed metadata, makes searching for a specific recording of a specific piece arduous, if not impossible.
So many of the problems seemed to come down to metadata. According to Jeremy Morris, a University of Wisconsin professor, Kemp’s tagging system took off when it was adopted by the Compact Disc Database (CDDB). The CDDB is the database which iTunes used to detect what was on a CD while ripping it.
Classical is not the only genre that works poorly with this tagging system, said Jonathan Sterne, author of MP3: The Meaning of a Format, a history and meditation on the technology. Audiobooks, lectures, and sound art don’t really adapt well either.
If anything, said Sterne, the long-playing record and the compact disc—the two great audio formats of the late 20th century—might have been special cases.
With both the CD and the LP, he said, “it just so happened that things that were of interest to the broader world of people who made recorded media, and people who were in the music industry, lined up with those of classical performers and audiences.”
Audio formats before the LP also failed to capture classical music, at least by our measures. Sterne recently bought a 1928 Victrola, he said, and it came with a 78 marked “classical music.” The record boasts five minutes of Debussy. That’s not, to be clear, a five-minute work by Debussy, nor in fact any labeled work at all: It’s just five minutes of some Debussy piece. For more than a century, classical music has been marketed as prestigious, even though what’s being sold may be distant from what the composer first wrote.
“There’s lots of issues” with MP3 tagging beyond classical, said Sterne. “Engineers, producers get left off records. Studio musicians are often left off in ways that were often much more detailed in liner notes.” A week ago, he was trying to find out the engineer who recorded Jay Z’s “Death of Autotune.” The MP3 ID tags didn’t say, nor could he find the answer in five minutes of googling. That’s exactly the kind of information, he said, that would have been printed in liner notes.
* * *
If classical’s messy software is ever fixed, it will require, first, better metadata.
“I get really obsessed with specific vectors of classical music performance—I’ll spend weeks listening to just one performer, or one work, or both,” said David Yee, a software developer who trained as a classical musician, in an email:
Being able to explore a particular movement of one of Mahler’s symphonies as performed by various orchestras in different halls would push all my buttons. The reason you can’t do that now has everything to do with titles—some recordings will refer to the movement by number, some by tempo marking; some put the catalog number in the track title, some in the album title.
New software that could provide more granular and specific access to songs in a huge classical library in some ways has to come after better metadata (both the fields to provide that metadata and community/label effort to provide that metadata). Once that’s in place, you can imagine putting together a day’s worth of listening that focuses on just one movement, or being able to add a complete symphony from a box set of Beethoven’s complete works with a single click.
I have two thoughts here. The CDDB, the industry’s leading database of MP3 metadata, is now privately owned and controlled, but it began as a crowd-sourced project with volunteer contributions. There is no reason this now-private database couldn’t be supplemented by a more robust, more complete database of audio file information maintained on a wiki-like basis.
“I want to take every other part of iTunes and propel it into the sun—everything from suggestions to equalizers,” said Yee. Andres and Muhly both expressed a similar desire for simplicity; some independent iOS and Mac developers have wondered this week why iTunes, the file management software, couldn’t be a separate application from Apple Music, the streaming service. To use iTunes now, it seems like, requires understanding not only Apple’s current music strategy but also its previous several.
But if Apple is committed to a cruft-ridden iTunes, other developers could step in the void. It’s not just classical music libraries: Many users with their own sizable libraries want software that lets them listen to MP3s and AACs. Plenty of minimalist text editors for Macs and PCs persist in the world. If iTunes is beyond repair, it might now be time for some minimalist media management software.
We want to hear what you think. Submit a letter to the editor or write to firstname.lastname@example.org.