Why technology won't kill spam. A remarkable thing happened when I downloaded my e-mail a few moments ago. Forty-six messages flooded into Microsoft Entourage. Three - all of them legitimate - stayed in my in box. The rest were transferred to a folder labeled "Spam" before I could even look at them.

I began paging through the spam folder and found the usual foolishness. Come-ons for Viagra substitutes. A get-rich-quick scheme from Nigeria. Pornography. And lest you think that's a fairly light load of garbage, be advised that this was only since 10 p.m. yesterday.

Yet I also found one message that shouldn't have been there. It was an e-mail I had sent out last night that for some reason bounced back. If I hadn't inspected my spam message-by-message, I never would have seen it. And that's why - despite an impressive error rate of just under 2.2 percent - I'm going to remove the spam-filtering software I've been playing with for the past week.

That's the problem with trying to eliminate spam. Losing one good message is worse than having to sort through scores of bad ones. The program I installed a little more than a week ago - SpamSieve - is highly rated, and it does seem to do an excellent job. But no program is perfect, as the SpamSieve manual itself acknowledges: it promises "to catch nearly every spam message yet produce very few false positives."

Well, if there are any false positives, or even the possibility of one, then I have to go through my spam folder with exactly the same attentiveness as I did with my inbox before I installed SpamSieve, don't I?

This isn't the software designer's fault, of course. (And, in fact, it would work a little better if I were more diligent: I keep getting warning messages that I've programmed SpamSieve to be oversensitive by showing it too many bad messages and not enough good ones.)

But the false-positive problem shows the limits of technology, and demonstrates further why computer users are dependent on Congress to deal with spam in an intelligent way. Will a new law called CAN-SPAM - whose implementation is described in today's Boston Globe by Chris Gaither - make a difference?

I hope so, but I'm skeptical. As this recent piece at Wired.com makes clear, CAN-SPAM may make so small a difference as to be nearly worthless.

One of the best overviews is this article by Christopher Caldwell that was published in the Weekly Standard last June. Since spammers depend on sending out millions upon millions of e-mails - a practice that now costs them virtually nothing - Caldwell proposed taxing e-mails - a very un-Standard-like approach, but one that might actually work. He wrote:

A penny-per-e-mail charge would drive most spammers out of business, subject them to jail time for tax evasion if they hid their operations, and cost the average three-letter-a-day Internet user just ten bucks a year. If even that seems too hard on the small user, then an exemption could be made for up to 5,000 e-mails per annum.

Sounds good to me. In the meantime, CAN-SPAM takes effect on New Year's Day. Perhaps when people see how ineffective it is, they'll demand something more toothsome. Caldwell's article would be a good place to start.