The message consisted of a small HTML component and a gigantic JPEG image. Had it been smaller, it would have easily scored 15 points even before Bayes training (as spamsassassin -t demonstrated), but we don't run anything through SA larger than 256 KB (as is usually recommended).
I've blacklisted the IP, but it looks like a throwaway.
So I'm wondering - any ideas on dealing with giant-attachment spam?
I don't suppose there are enough efficiency gains in 3.0 to safely raise the size limit?
-- Kelson Vibber SpeedGate Communications <www.speed.net>