anti-spam measures Hunter, 2/22/04 [more]

The Artifex now has a very competent spam filter available for general use. It uses a Bayesian algorithm, which means that you train it to recognize your spam and non-spam. Training typically takes only a few days, is very easy, and results in a very low number of false diagnoses.

To use spam filtering on your account, read using bogofilter.

Errata:
We're here in the enterprise of permanent e-mail. It was only a matter of time before our addresses became known to every one of the major spammers out there.

In December John [jsaboe] wisely suggested using Bayesian filtering. There were a number of programs out there to do this sort of thing. Our recent move to the qmail mailserver finally moved us toward a fine program called bogofilter; that's what we're using here. It works very nicely with qmail, and was written in Python, too. How can you beat that?

The work so far has been to implement user-level, personal spam filtering. Further work to do on spam will focus on a server-wide policy, using a Real Time Blacklist, probably from Spamhaus, one of the high profile anti-spam databases. Qmail is an excellent mail transfer agent. Along with simplifying our effors against spam, it should also help us with other great things like sensible mailing lists and better support for custom user configurations. Coming to a server near you.

If you want to get involved with any of this or have questions, write to me and Mike Hugo [hblanks & mikeh].