>I just got IMGate running on a test box using Red Hat thanks to the easy to >follow instructions in the read me :-)
good, RTFM works again :) >I was wondering if message weighting/scoring could be done with postfix / >IMGate? not with IMGate basic or advanced, which is oriented to rejection based on envelope information in the SMTP dialog, before the DATA (content) is received, so maximum bandwidth is conserverd. postfix has non-weighted, regex content-scanning for the msgs that aren't rejected. weighting and heurisitics are mostly content-scanning systems, they use the envelope information + the entire message body to evaluate. So you must receive the entire message for them to be effective. >I havn't found alot of good information on this and the stuff I did >find, I don't quite grasp. see vipul's razor, spamassassin, Message Sniffer (commercial and very good), etc, etc. Running these as postfix content-filters AFTER IMGate's other filtering should work fine, since 90++ % of msgs passed to the content-filter will be legit. So the content-filter will be trying to catch the few messages that IMGate doesn't. I have asked in the IMail forum twice and got no response: I would like people who are running Imail+junkmakil AFTER IMGate to tell us how many rejects/day by IMgate, and by junkmail. and even better, the Junks rejects/day before IMGAte was installed, and the Junkmail rejects/day after IMGate was installed. Nobody has answered yet. I would also like to work up some scripts for junkmail logs that harvest junkmail rejects and email the ip's, PTR's, email addresses to IMGate for incorporation in IMGate filtering. I want to do the same with MadScientist's Message Sniffer. The objective is to convert the expensive content-scan rejects to the more efficient envelope-stage rejects. That is, "feed forward" from the back-end content-scanner to the IMGate front-end envelope-filtering. >I dont' trust some of the RBL's completely They aren't that effective anyway, they help, but don't predominate. there is no single silver bullet. Len
