Nate Davis wrote:
Howdy,
We are running SpamAssassin 3.0.2 called through QmailScanner. We are not using Bayes, and are using systemwide settings. This Server, receives the e-mail, and then scans it with SA, and then F-Prot, and then using qmail (smtproutes) forwards the e-mail on to our internal mail server.
Now that you understand the setup, more and more Spam seems to be getting through these days. When we first set this box up about 6 months ago, it seemed to be tagging tons of spam. Almost everything that was spam was being tagged as such. Now, more and more spam is not getting tagged. Below are the settings we have always used. Because the e-mail is stored on a totally different server, bayes is not a reasonable option.
Why isn't Bayes a good option? Like many people, I have an internal messaging server which is the final destination of all incoming email. However, my SpamAssassin box uses global Bayes with autolearning and its performance is fantastic. I strongly suggest you reconsider this.
