[SLL] collaborate on a dnsbl?

[Jules Agee]

Chuck Wolber wrote:
> I would recommend not re-creating the wheel then. Spamhaus already has 
> that and does an extraordinary job at it.

For what it's worth, I agree with that sentiment. There are already some
very good blacklists available. My favorites are spamhaus, spamcop, and
dsbl. I don't block outright on any of them, but if any IP is listed in
more than one of them, the message is blocked unless the bayes score is
really low.

I suspect adding another IP blacklist to the existing ones is unlikely
to improve your spam-blocking results by much because you'll never catch
the latest newly infected virus zombies until after you've received
their spam. A better approach is to diversify your spam-identification
tests.

When I look at why SpamAssassin works so well for me (better than 98%
detected, almost no false positives, and even the few FPs I've seen have
been "legit" bulk e-mail ads, real-opt-in), the huge collection of
static rules are only a small factor. By far, the most benefit comes
from the IP blacklists, the SURBL's and the Bayes filter. The AWL helps
quite a bit with accurately identifying legit senders who occasionally
send a spammy-looking message. If those four features were all I had, I
think I'd be pretty happy.

-Jules