Charging a little on the Internet is even harder than charging a lot.

Transactions involving money--say, the purchase of something at a store--cost money. The cashier must be paid, the cash register bought, the bookkeeper hired, etc. Honest mistakes and dishonesty also take a toll. These expenses can be expressed as a cost per transaction. The higher that cost, the larger the transaction has to be in order for it to make financial sense.

Computers are very good at counting money and keeping records, so they have been used for decades to manage transactions. Today, when your credit card is run through the little machine at the checkout counter, computers handle the entire transaction. Nobody is going to look at that little slip you sign unless a problem develops.

What is the cost of a transaction done purely by computer? A recent estimate put the cost at less than one-millionth of one cent per transaction--i.e., 100,000,000 transactions can be done for $1. Digital cash--where the value is stored on the card you carry and there's no need to check your credit against a faraway database--may be even cheaper.

Regardless of how it's done, transaction costs will continue to plummet as computers get more powerful. Low transaction costs are a wonderful thing if you're in the transaction business. They're wonderful for consumers too, making it cheaper and easier to buy things and creating new things to buy. All this is especially true now that the Internet provides a direct connection between sellers and millions of consumers.

In a normal situation, you could end there. But this isn't a normal situation. This is the Internet, where any legitimate idea is immediately taken to ridiculous limits. That's because there are plenty of would-be Internet visionaries competing to outdo each other, both in philosophizing and in trying to sell shares in their start-up companies to the public. The magic words "on the Internet," if inserted into nearly any sentence, seem to protect it from normal critical scrutiny.

So it is with cheap transactions, which tend to go by the sexier name of "micropayment systems." Press reports about micropayment schemes extol their virtues in breathless wonder. Micropayments, it is said, will revolutionize the economics of journalism on the Net (a topic I've opined on before). People will pay some small fee--like 10 cents--to read an article, rather than buy a whole magazine for $2.95. Or, what the hell, they can pay by the paragraph, sentence, word, or--why not push it to the extreme--the bit. Wouldn't that be wonderful?

N o, actually, it wouldn't. Micropayments might be important in some limited areas, but most of the hoopla is very poorly thought out. To see why, consider Slate, which recently decided to remain free to users for now. But imagine that, several years hence, Slate is trying to decide among micropayments, subscriptions, and remaining free. Let's consider what effect these choices will have on a couple of typical users (or "readers," as they used to be called--tellingly, the computer industry shares with the drug industry the habit of referring to its customers as "users").

Joe Avid is what we in the computer industry would call a "heavy user." Joe reads every bit of every issue of Slate. Jeff Accidental rarely reads Slate on purpose--but every now and then, he follows a link from another site into Slate. Somewhere in the middle is Tom Average, who reads Slate just as much as the average user does.

What price should we set for the micropayment? If we set the price so that it will yield the same total revenue as a subscription, then Tom Average will pay the same amount either way. Jeff Accidental will save a bundle by buying occasional articles instead of having to buy a subscription. But it's tough luck for Joe Avid. He--and everybody else who reads more of Slate than its average reader--will pay more than a normal subscription would cost. In customer-service circles, this is known as "screwing your best customer." That's OK in the sex industry, but not so great in others, because it tends to alienate the loyal supporters who form the core of the business. (It is, though, standard practice in the print-magazine industry, which gives new subscribers a bargain rate and rewards steady renewal by raising the price.)

The problem, for Slate and other Internet sites, comes from having to charge for usage, when what they're selling is intellectual property with a flat production cost. Slate doesn't get "used up" by being used. It costs virtually the same amount to produce, no matter how many people use it, and no matter how heavy the use.