> What are you doing about spam that goes through being scored too low for
> spamassassin to get it? Do you create your own rules or try to block it
> some other way? Is there a way to get SA to train itself by telling it
> that certain message is a spam such as Thunderbird etc...?
Almost every time I get one I write another rule. That doesn't happen much
any more, my average spam score is around 50, with a 4.6 requirement, and it
is a rare spam that scores below 12. I generally scan anything below 20 or
so to look for new rule candidates when I have spare time.
Use the SARE rule sets that are appropriate to your circumstances, and the
others like backhair and weeds and tripwire that are also out there. Get
SURBL working if you run net tests (I don't).
And finally, turn on the bayesian filtering and get it trained. This will
take a little of your time daily for a week or two to get it well trained,
but then you should only have to pay attention to it occasionally to make
sure it is still working correctly.
Loren
