Saturday, September 14, 2002


Oren Etzioni discusses anti-spam measures in today's NYT. While I agree with the sentiment, I disagree with several of his points. He proposes that we all bounce spam back to sender. That won't work, for several more-or-less well-known reasons:
  • many spammers manipulate mail headers to disguise the source of the message, and the reply will simply fail to go through, creating yet another message, a failure report
  • many spammers change headers to make the message appear to come from an innocent bystander, who would then be the victim of the replies
  • header tampering is hard to defeat because someone's email client has no access to the true IP address of the sending SMTP server, which only the receiving SMTP server at one's ISP knew; and in any case the sending server is often an open relay, or a dynamic IP address assigned to a "hit-and-run" ISP account that will soon be inactive or assigned to some innocent party
  • some spammers use email address-guessing schemes, and they actually *want* replies because they validate the addresses they have guessed

Oren also dismisses charging for email on the grounds of unfairness. But we all pay more for our credit cards and phone calls to cover the losses from fraud. It may not be pleasant, but it's the only way for the service providers to recover those costs, which arise from exploitation of the very convenience we appreciate in credit cards and phone service. In other words, someone has to pay eventually to the added traffic and wasted time caused by spam. Distributing the cost by charging for SMTP packets and requiring peering equity would pretty much stop spam and help ensure that a critical infrastructure is well funded.


Oren notes by e-mail that the NYT edited from his piece material that addresses in part my two points above. However, I still think that "bounce to sender" is unlikely to work well because any method for guessing the original sender automatically will soon be circumvented in the same way as rule-based spam filters are. We need an incentive for ISPs to put a lid on spammers, which is what my charging proposal does.
10:42:09 AM