The difference between a system with prices and one without is visible, alas, in my own behavior. When I dial into PaperChase, the $18-an-hour link to the medical database MEDLINE, I plan my incursion with the care of Stormin' Norman attacking the Iraqis. I write down my goals beforehand, blast in, grab my data, and rush out -- all in a few minutes. Because users like me are in and out quickly, it always seems possible to make the connection. If the lines get jammed, PaperChase will put in more equipment, knowing that it can recoup its investment.
By contrast, when I go onto the World Wide Web I may clog up the network at peak time by downloading a video of former President Bush bonking somebody on the head with an errant golf ball. After I glance through this for a few moments, I may then skip on to the collection of recorded burps that some aficionado of eructation has put on-line. If the lines get jammed, the Bush hater and the burp man who uploaded the files I retrieved will do nothing. Nor will I. "There's no incentive to do otherwise," says Varian. "Why should people restrain themselves when the next guy is free to abuse the system?"
Varian and others have proposed methods to make users sensitive to costs, but all methods face grave difficulties. Some of the difficulties are technical. For instance, the World Wide Web and many other Internet applications are "client-server" based. The client sends in a request for information to a more powerful computer, the server, which then pulls in the data and sends them back to the client. To the network as a whole, it appears that the server is generating most of the traffic. The client, who should be charged, is almost invisible.
Overcoming the problem would require labeling and tracking each of the separate packets of information in the TCP/IP protocols, which would be an accounting nightmare. Telephone calls can be billed with relative ease because they are continuous. To itemize them, companies need to generate only a single accounting record -- a connection between one number and another that lasted, say, 10 minutes. Because of TCP/IP, the Internet is different. What looks to the user like a single stream of information is really many small chunks of information that travel through the network by different routes. As a result, following every Internet "call" would involve many accounting records, as each packet traversed its separate route through the system. Indeed, Varian has estimated that billing a 10-minute Internet session telephone-style would require 25,000 accounting records. (A newer protocol, asynchronous transmission mode, simplifies matters somewhat by sending all packets on the same route. But it has never been fully tested or implemented.)
That means, according to Jeffrey MacKie-Mason, another Net economist at the University of Michigan, that billing for Internet sessions as if they were telephone calls is "flat-out impossible." Instead, he argues, use of the Internet could be priced in somewhat the same way as use of the postal service -- by stamps. Consumers would buy Internet "stamps" from their providers and spend them every time they made a connection. Charges would depend on the type of service and the demand on the system at the time. Still, he says, "the accounting costs are a big question. They could well be impossibly high."
But even if the technical problems can be overcome, MacKie-Mason notes, a successful pricing system would create another problem -- one with special import for the thousands of businesses now clamoring to get onto the Net. If the Internet is no longer free, people would be forced to decide how much they value it. In many cases, the answer seems to be "not much."
* * *
In my constant quest for bargains, I have briefly signed up with a number of Internet providers, only to become exasperated by the maddening randomness of the Net. I keep searching for something that is cheap enough to make slogging through the complicated searches not feel like a waste of time. Not long ago I was halfheartedly registering with yet another service while reading the New York Times. It happened to be the day after the FBI arrested Kevin Mitnick, the hacker who broke into the computers of computer-security experts. As I dialed up the on-line registration form, I read how Mitnick had also slipped into the WELL, the network in San Francisco. There he had stolen the credit-card numbers of 20,000 members. My modem made that strange electronic flushing noise as it connected to the on-line registration service.
Then the service asked me for my credit-card number. I reacted emotionally. I clicked the "cancel" button and ended the procedure. A few days later, I joined, realizing that my earlier ventures into the Internet already had sprayed my credit-card information all over cyberspace.
But my hesitation is emblematic of another fundamental problem with the Internet -- lack of trust. As everyone in business knows, transactions cannot take place unless there is some trust between buyer and seller. Buyers need to believe that merchants are selling goods that are what they appear to be; sellers need to believe that consumers will actually pay. In a store the necessary trust is easy to establish. The shop, with its cash registers and displays of merchandise, is a statement of the vendor's permanence; it will be there if the buyer needs to return the purchase. The advantage for the seller is that the store allows buyers to pay in cash or by check in person; the merchant can verify the consumer's identity on the spot.
The Internet is different. Generally speaking, businesses locate on a part of the Internet called the World Wide Web, which allows them to set up "pages." Pages are identified by strings of letters and numbers called Uniform Resource Locaters (URLs). URLs can be unwieldy. For example, the one for Noteworthy Music, a compact-disc seller in a service called the Netmarket, is http://www.netmarket.com/note-worthy/bin/main/:st=yln4pbm227|4. When users enter URLs into their Mosaic software, they are conveyed to the appropriate Web page (if the Internet isn't too crowded to prevent the signal from going through). Pages typically display company logos and descriptions -- and little pictures called icons that guide consumers to further information on the products and to order forms. The format is dictated by a computer language called Hypertext Markup Language (HTML). As a result, pages look very much like one another, just as one display ad in the yellow pages looks like its fellows.
