Loren Wilton wrote:
FWIW, I've been running that rule [checking for middle initial in
"From"] since before it was mentioned on the list, and it is still
moderately useful. It does hit ham, but at one point or however I
have it scored that isn't significant. On the other hand, that point
has more than once pushed a spam over the limit.
One has to remember that its ok for some tests to hit ham, as long as
they don't result in an FP.
The same would be true of a rule that checked to see whether the sender
had a ".com" address. It would hit a great deal of spam, and giving it
a small score would push a lot of spam messages over the threshold. It
would also hit a lot of ham, but the additional score wouldn't seem
significant.
The problem is that once you have a several of those less-discriminating
rules, the likelihood that more than one will hit on a ham message goes
up, and you get false positives.
--
Keith C. Ivey <[EMAIL PROTECTED]>
Washington, DC
