I’ve been talking about these a lot on the SpamAssassin-talk list and other places, so forgive me for not blogging much about it.
Paul Graham talks about his naive Bayesian spam filter. We already use a very basic form of this kind of matching in SpamAssassin, in the SPAM_PHRASES matches; but it’s not proper Bayesian filtering. However, it looks like Matt is taking the bull by the horns and making it Work Right once 2.40 is released (any day now). (BTW it’s worth noting that Bayesian filtering doesn’t always seem to get the success rate that Paul talks about; we think this is down to what kind of mail you get.)
While we’re doing that, we’ll have to make sure we don’t hit this MS patent. grr.
Habeas Sender-Warranted E-mail has launched. It’s a very nice solution, allowing non-spam senders of all kinds to “sign” their mails with a “mark” indicating that it’s non-spam — and filters, like SpamAssassin, can then use that mark as a good compensation signal (SpamAssassin now has the HABEAS_SWE test in CVS).
The mark in question is a copyright- and trademark-protected haiku. Virtually every internet-connected country in the world honours copyrighted poetry with a high degree of legal protection, so unauthorised reproduction will be a big no-no, and result in a heavy battering in the courts.
As a result, they’re going to have to have some serious lawyers on their side. But it looks like they do. And to really press the advantage, they’ve teamed up with Dun and Bradstreet — who can seriously impact a scumbag’s ability to do business in the western world, never mind just Florida, if it comes to that.
However, there’s still money to worry about (as usual). It does cost a hell of a lot to pursue as many legal cases as they may have to. Let’s hope they can pull it off. Good luck folks!
Finally, cool — I’ve made Aaron’s see-also bar!