dave wrote: I've used the same approach for about a couple of years, and it works well. I get about 100 spam attempts per hour. 99% are rejected by Postfix using two checks: (1) Does the FQDN in the EHLO resolve to the client IP, and does the client IP reverse resolve to the same FQDN? (2) zen.spamhaus.org RBL. Of the 1% that gets through, over 90% is nailed as spam by the Bayesian filter in Thunderbird. So I get about 1 or 2 spams a day in my inbox, and about 20-ish in my junk folder which I can peruse at leisure to spot the very occasional false positive (about one a month). As far as Bayesian filtering is concerned, I've never bothered with spamassassin. The filter in Thunderbird works well enough, needs no configuration, and offers an ideal user interface for picking up false positives. I also never bother with virus checking, as I have no Windows machines and there's no risk of me sending viruses to anyone who does. It also puts quite a heavy load on a mail server and slows things down. (My server is an old 450 MHz beige box, which I like because it uses only 29 Watts of juice. Guess who pays the electricity bill here.) And Clamav, good though it is, does seem to have a regular supply of security holes. I'm now looking at other objective tests such as SPF and DKIM. I don't expect these to do much about spam, but they may be of more help against phishing. -- Richard Lamont http://www.lamont.me.uk/ <richard(a)lamont.me.uk&gt; OpenPGP Key ID: 0xBD89BE41 Fingerprint: CE78 C285 1F97 0BDA 886D BA78 26D8 6C34 BD89 BE41