MailScanner vs. SpamKiller

[Nick Nelson]

Martin Hepworth wrote:

> Without Bayes I'm getting about 95% hit rate, with bayes about 99.5%,
> yes bayes really does make that much difference! I'd really suggest you
> spend some time to give it the initial 200 instances of spam and ham...


If you are using the machine as only a gateway (mail only passes
through, never is stored on the server.)

What's the best way to train it? I saw a post on the list of a script so
that you could just forward spam to a certain address and it'll train
it, is that the best way? I'd definitely like to use bayes however not
sure of best way since no mail will stay on this server.

nick


--
Nick Nelson
www.easyservermanagement.com
We Make Server Management Easy!