The world of digital music flowing over the Internet is a confusing jumble of technical standards, mathematical constructions, file formats with strange names and a good amount of legal wrangling. The good news is that while the legal storm continues, the technical aspects of online digital music are becoming easier to understand and utilize. All the interest in using the Internet to share music is leading companies to create better software packages that are gradually turning the personal computer into an easy-to-use jukebox with a personalized selection.
The technical area can be divided into two different challenges: storing the music as a computer-readable file, and delivering that file over the Internet. The first task involves choosing among a number of different file formats competing to offer the best sound in the smallest amount of space. The MP3 format is by far the leader, but there are newer formats jostling for position, including Advanced Audio Coding, Nippon Telegraph and Telephones TwinVQ, Microsoft Windows Media Audio and RealNetworks RealAudio G2. Of course, new formats and versions are appearing constantly as researchers try to squeeze out extraneous data.
The other challenging part of the proposition — moving music files across the Internet — is the one that raises legal questions. Many of the earliest solutions, with names such as Napster and Gnutella, have shocked music content companies, which see them as tools for piracy. Of course, the same tools are also quite useful for kids who want to send their grandmother a song or musicians who want to give their music away. The next generation of these systems will try to balance the needs of consumers with the demands of those who hold the rights to the music. In many ways, the legal minefields are the real challenge facing the companies developing delivery mechanisms.
Understanding the various formats on the Internet is not difficult. While there are hundreds of possible ways to express music in a computer file, most of them use the same basic solution and tuning it to handle each style efficiently. Spoken word recordings, for instance, can be stored differently than classical music because people generally use only a narrow range of frequencies when they speak, while an orchestra uses a large range.
All digital music begins by converting the audio information into a stream of numbers that represents the intensity of the sound at a particular instance. Most of the sophisticated, modern mechanisms sample the intensity about 44,000 times per second, a number that is determined by the limits of human hearing.
This raw data stream can be quite large, so many researchers have developed compression algorithms that can make them smaller, often one-tenth of the original size. The MP3 format is the most famous of the compression algorithms.
MP3 and most other compression schemes work by analyzing the data stream to identify the frequency of the sounds being played at each particular moment. Then, the most important frequencies are kept and the others are tossed away.
Deciding which frequencies to keep and which to discard is very much an art. This is why many performers and music aficionados can talk for hours about the differences among music compressors. Each programmer and manufacturer can take a different approach to choosing the most significant frequencies, so each software package will compress the same music in a different way. The MP3 format and most other popular ones encourage this kind of competition by simply playing the frequencies they are told to play without specifying much about how the software developers should choose the frequencies.
The human ear may detect a wide range of frequencies, but the brain only picks out a few significant ones. Strong sounds will often mask weaker similar ones.
Right now, developers are building better, more sophisticated perceptual models of human hearing. Each has its own devotees who believe intently that their format has made the right artistic decisions.
The clash between the two cultures can be quite hilarious, and many engineers can tell humorous off-the-record anecdotes about spoiled rock bands with very specific ideas about the timbre of feedback: One artists signature sound is another compression engineers garden-variety noise.
In other cases, a clash emerges between the educated musicians and the untrained ear of the engineers. Classical musicians spend years learning to pick out the subtle ways in which dozens of instruments blend together in an orchestra. They want to hear every note. Yet many casual classical listeners pay attention to only the main melodic line and dont really notice if minor harmonies disappear. Finding the best compression algorithm often depends upon the audience for the music file.
Spreading the Word
Creating the music file is the first half of the game, but delivering the file is the most controversial. Digital music is just a computer file like any other, and any mechanism for sharing files can be used to ship the files from person to person. The old favorites, floppy disks and e-mail attachments, are just as useful as the newer, more controversial tools such as Napster.
The most basic delivery service is the Web site, and there may be hundreds of thousands of sites that offer free MP3 tracks for download. In many cases, these recordings are placed there by artists who are giving the copies away in the hope of building a following. Sometimes the advertising revenue generated at these free sites for popular artists can be surprisingly good.
Occasionally, new technology services cause friction with older, established music publishers. MP3.com, which took its name from the music format, is now notorious for developing a system that would allow people to listen to CDs on their computer system without going through the time-consuming process of creating an MP3 file. The company stored copies on its servers and allowed anyone with a legitimate copy of the CD to stream the files from MP3.coms servers. This practice ran afoul of the music companies, which sued MP3.com and eventually settled with the service. Now, users of MP3.com must pay extra to use the service and still demonstrate that they own legitimate copies of the songs.
The notion of using a store-bought CD as proof of ownership is being embraced by a number of companies. Gracenote.com, formerly CDDB, is building a CDKey system that would allow record companies to distribute extra content over the Web that would be unlocked whenever someone inserted a legitimate CD into his or her computer. This system may include extra tracks, liner notes, lyrics or even special events such as online chats with the band.
Of course, the most controversial services are the online trading pits such as Napster or Gnutella that allow users to exchange MP3 files. The services keep a running database of the files located on users hard disks. If someone wanted a copy of a song by the Beatles, the Napster service would be able to point the person toward a set of users who have it ready for download.
Napster defends this service by claiming that the sharing is not much different from when a person makes a tape of an album for a friend. Record companies, of course, see it differently, and suggest that the central databases are contributing to massive amounts of copyright violations between people who have never met.
Gnutella, one of the more sophisticated cousins of Napster, removes the need for the central database by letting users query each other. This makes the searching process much less efficient and limits the effectiveness of the Gnutella network. It does, however, make it harder for the record companies lawyers to find someone to sue.
In the most abstract sense, Gnutella is a version of Napster empowered to take on the job of the central server for itself. While the Napster clients rely on the central Napster server for directions to the right songs, each computer with Gnutella can contact any other computer running Gnutella.
One of the more ambitious service models involves sending a stream of music to each user. A number of Internet "radio stations" do this already by broadcasting copies of their music in packets using formats such as Apple Computers QuickTime, RealAudio or Windows Media.
These solutions are often inefficient because they require sending a complete copy of a song to each user each time it is broadcast. While the copyright holders seem to like this approach because it does not leave copies around for people to play again and perhaps pirate, the costs are substantially higher than other solutions that trust the computers of listeners to hold a copy.
These debates attack the very nature of the technology. Most computer users and manufacturers feel that the personal computer should let the owner access all of the information on the computer. They believe that the machines are most useful when people can record sounds, edit them and distribute them. The most ardent copyright holders, however, see the devices as potential tools for piracy. They believe that the computers and handheld players should detect illegal copies and enforce restrictions.
One potential compromise is a digital watermark, a signal mixed into a music file that identifies it as copyrighted material. The method shows much promise, but there is considerable debate about whether the watermarks will be strong enough to resist tampering. If they are, then the tools would be able to distinguish between original works of art.
All of these technical challenges can be so daunting that some engineers feel the most outrageous and radical solution is to use an old technology — the radio or television — to deliver the music files. Radio and TV stations can broadcast copies simultaneously to millions of users, without requiring a separate connection to each one. The right to rebroadcast music is already well-understood and negotiated.
It all looks relatively simple. Sometimes in order to move forward, you may need to look backward.
Peter Wayner is the author of Compression Algorithms for Real Programmers and Digital Copyright Protection (Morgan Kaufmann Publishers).