Adi Pircalabu said:
Michel Messerschmidt wrote:
Not always. Sometimes you want to reject messages as spam based on certain well-known (at least for a small timeframe) source IPs, FROM or RCPT-TO addresses
All due respect, you're wrong here. Bouncing is evil, it can fill up your bandwidth, and, somtimes, the bounce ends in the wrong place. You can also hit and annoy enough recipients (victims in this case) to get listed in various RBLs all over the net.
I fully agree with this. Once you use unverified addresses to activly send out automated mails, you've lost. But this only concerns bounces created by your own server in response to some (probably faked) addresses. I was talking about an SMTP reject that leaves the responsibility at the sending mail server - the one that really connected to your server (see the example at http://en.wikipedia.org/wiki/Bounce_message for the difference).
Well, how do you reduce the load of a mailserver by bouncing spam? Here is the scenario: the message hits your mailserver and is analyzed.
I know one case where the mail analysation (spam/virus) caused too much load and crashed the server :) The *temporary* reject of unknown origins exceeding a certain connection request rate may solve this problem. -- Michel Messerschmidt, lists@michel-messerschmidt.de $ rpm -q --whatrequires linux no package requires linux