rules have associated "weights"
such as -5 (for a user-agent such as Pine)
or +2.5 (body incorporates a tracking ID number)
rules applied to a message add up to the message's "score"
user can set the score that indicates "spam" (default: 5)
if the score goes above the trigger-point, it gets handled differently:
"Spam-Flag = YES" header
verbose description of the rules it matched
wrapped in MIME so can't set off javascript or web-bug images