[Spambayes] Spam Prefiltering

[Tim Peters]

I advise training spambayes on the email you "actually get".  So, e.g., if
you're happy with the decisions your server blacklisting makes, toss those
msgs unlooked at by anything -- save the bandwidth and don't even fetch
them.

One of my ISPs (Comcast) does extremely conservative scanning, seemingly
catching only msgs that are exact matches to canned fingerprints.  Despite
that it's so unambitious, it does catch megabytes of viruses every day, and
after confirming that over the course of a month, I set the account to toss
those at the server (so I don't even learn about them any more).

It's partly a question of attitude:  I'm *not* training my classifier to
recognize spam!  I don't care about recognizing spam.  I only care about
weeding out messages I don't want to look at.  If a server can do part of
that for me, great, that's closer to the goal.  There's no practical benefit
in teaching my personal classifier to recognize crap a server already
recognizes.

OTOH, only Comcast's filtering is good enough (and seemingly because it does
such a timid job) that I trust it.  On the other end of that scale is MSN
9's junk email filtering, where virtually all msgs to jython-dev showed up
in my junk email folder there, despite "training" on dozens of them as ham
over the course of a month.  So I disabled all server filtering gimmicks on
that account, and let spambayes sort them out.