Arnold Kling     Essays | Short Book Reviews | Favorite Links | Internet Bubble Monitor | Home

A Fad Ap-peers

Arguing in My Spare Time, No. 4.05

by Arnold Kling

February 16, 2001

Not only am I without a Palm Pilot. I never used Napster.

This is not because I support the music industry position on the sanctity of copyright. Quite the contrary.

On the copyright issue, I agree with John Perry Barlow. As he puts it, you no longer can sell your knowledge in the form of atoms. You need to sell it in the form of relationships.

In this view, the music industry's attempts to force people to buy music in the form of CD's is untenable. And I believe that per-use pricing, even if technically feasible, is even worse. You need to charge people for relationships, not for information.

For example, I am writing a business book. Even modestly successful business books do not bring much money to the authors. In fact, most business books are written by consultants. The economic function of the book is to spread the consultant's reputation and to bring in more business. (I am not a consultant. There is no economic motivation for writing my book.)

In music, the analogy would be with concerts. In the long run, musicians will have better luck charging for attending concerts (or lessons) than charging people a license fee to listen to their music.

In one of my earlier essays, I sketched what I consider to be the most plausible solution for selling information. (It is one of several solutions given in Varian and Shapiro's "Information Rules.") I called the idea a "software club," because I was using software as an example.

One also could have a "music club." As a consumer, you pay a certain amount per year to belong. As a member, you get easy access to all of the club's music. You enjoy whatever additional benefits that the club provides, such as low-cost concert tickets, customized recommendations for music, and so forth. The club negotiates with musicians to pay them a salary for providing music to club members.

The "music club" solution will not be perfect. Some people will not join the club, and instead they will try to obtain music for free from club members. However, as long as the benefits of belonging to the club are sufficient to attract the most committed music fans, the club will work.

But this essay is not about selling music or other information. It is about technical architecture. In particular, the New New Thing sweeping the cognoscenti is "peer-to-peer" (P2P) architecture, which Napster exemplifies.

When it comes to technical architecture, will the music club adopt the P2P structure of Napster? Or would it instead work with a client-server architecture more like the current Internet?

In musical terms, will your personal computer be a tape player or a radio? If you use your personal computer to store and play the music, then it is a tape player. If you only use it to listen to the music, then it is a radio.

In the peer-to-peer model, everyone has a tape player, and the decisions about music storage are decentralized. In the client-server model, everyone has a radio (the client), and the music is stored on central servers (like radio stations, except that they connect to radios individually rather than via broadcast). Either way, as a listener, you choose the song you want when you want it.

Personal computers evolved before the Internet took hold. Because they were designed to be used in isolation, PC's were built on the tape player model. What you cared about in a PC was local processing power and storage capacity.

If your Internet connection is good enough, then you want to use your PC more like a radio. You do not care where anything is stored, and you need only enough processing power to do input and output.

For example, Don Britton, one of the entrepreneurs that I interviewed for my book, offers a complete outsourcing solution for small businesses (think of a law firm). Even if you are doing word processing, all of the processing and document storage are done on a server at Britton's firm, This gets you out of the businesses of network administration and software installation, which are overwhelming to small firms (and even many large ones).

Clay Shirky, an otherwise sensible chap, has positioned himself as a leading P2P evangelist. His argument is that the client-server model wastes all of the processing power and storage currently available on PC's. Most people do not use their PC's all the time. Also, most people do not fill their hard drives to capacity. P2P could allow for markets to be created whereby this excess capacity is rented.

But think about what it would take to enable the P2P scenario to work. It would require bandwidth that is high-capacity, reliable, and secure. If these conditions hold, then people are indifferent as to the physical location of computing resources. If not, then the only reliable computing power you have is what is on your own machine.

Shirky is correct that in a high-bandwidth environment, people will not want to hold excess processing power and capacity. However, the solution will be for them not to carry excess capacity in the first place.

If the bandwidth is sufficient, then I can sign up with Don Britton's outsourcing service, let him have my hard drive, and keep only enough processing power to operate a keyboard and a screen. Once I am indifferent to where the processing and storage are located, I will stop buying local processing power and storage, and the excess capacity in PC's will go away.

My bet is that the P2P concept is a mirage. The computing capacity to make it work only exists because we do not have the bandwidth to rely on client-server computing. However, that lack of bandwidth also implies that P2P is not ready for prime time as a tool for businesses to embrace for sharing files and distributing applications over multiple processors.

Ultimately, the scarce resource is not processing power or storage. The scarce resource is human talent. The cost of chips is falling. The cost of network administrators is not. Data centers and the client-server model can economize on network administrators and on software development to support them.

Someday, we will have enough bandwidth to facilitate powerful applications of P2P computing. At that point, precisely because we are indifferent about the physical location of computer resources, P2P will no longer have any marginal benefit. The conditions that are necessary to enable P2P to achieve its potential also will be sufficient to make it obsolete.