May 13, 2003

Will spam doom e-mail?

A recent New York Times article notes, “This summer, unless something changes, the volume of spam messages will for the first time surpass that of e-mail from people who actually know one another.” I don’t know the source for that assertion, but the volume of spam does seem to have exploded over the past few months. I have to wonder how sustainable this is: Will we reach a point where the signal-to-noise ratio in e-mail become so high that people stop using it?

Whenever people start complaining about spam, someone, often a spammer, will note that individual spams are easily dealt with—how hard is it to hit the delete key? That misses the point, of course. Repeatedly deleting unwanted messages is annoying, but what about the time I spent downloading it? The storage space my ISP used to store it until I checked my mail? The network bandwidth between me and the spammer? If 50% of the messages being sent through e-mail are noise, that means that the network providers have paid for twice as much capacity as they need.

Unfortunately, there don’t seem to be any easy solutions. Filtering, even with clever Bayesian or latent-semantic algorithms, is an incomplete solution. It makes my life less irritating, but it doesn’t reduce the network’s burden. Now, if everyone used filtering, then spam might become less useful as a marketing technique, but as long as there are people who (1) read spam and (2) respond positively, there will be incentive. It won’t matter what laws you pass or what bounties you offer.

Pushing the filters further back in the network creates its own problems. If my filters accidently mark a legit e-mail as spam, that’s arguably my own fault. If my ISP blocks a legit e-mail, that’s a problem—especially since I might never find out about it.

Digital signatures might help identify valid e-mail, but nothing would stop a spammer from obtaining a key and using it to sign their spam. It’s hard to see how this could be useful in blocking spam without also blocking legitimate e-mail or requiring people to register with a central authority.

XNS theoretically allows people to block spam by requiring everyone who sent you mail to sign a contract promising not to spam you, but of course XNS has yet to result in anything like an implementation.

People like to blame the openness of the e-mail transport protocols for spam, but it’s not clear to me how much improvement could be made. Certainly, a system being developed today would have much sterner means to prevent spammers from hiding their origins, but even they won’t be perfect. The question is whether the improvements that could be made will be good enough to be worth all the effort in changing the mail protocols. That’s something I don’t know enough to speculate about. #