Please note I'm not using that way, nor I'm using spamd. That said.
From: Jeff [mailto:[EMAIL PROTECTED] > > I've seen an installation where postfix handed off mail directly to > spamd, treating it as a policy server in smtpd_recipient_restrictions. > I was thinking about doing this for a server with several thousand users > and heavy usage. > In that configuration, how would spamd behave under load? It would work more or less how it works in the more classical content filtering case, so I guess it wouldn't matter how you invoke it. > Is it likely to cause problems? I *guess* that the way you describe wouldn't allow the destinating user to have control of received spam. I mean, he/she wouldn't have a 'spam' folder in its mailbox. This may be a problem in case of false positives. BUT, also note that this way the senders gets informed of the undelivered message, since the postfix's connecting peer would get a 5xx error and, in case of a legitimate sender, he/she would easily get a bounce mail back from his/her smtp server. The 5xx error is not that bad in this, but also exposes a neck to spammers: since they have a feedback from your antispam engine, they may attempt multiple versions of their "payload" 'till they get into... > Should I be using amavis? I use it and I'm fine with it, but I don't see an easy way to integrate smtpd_recipient_restrictions with amavisd-new: amavis is designed to resubmit legitimate messages for final delivery to the MTA, so I don't see how could this cope with refusing the message at whole. However, if you forget the smtpd_recipient_restrictions way, you would get even antivirus handling thanks to amavis. > What's the best way to get mail to spamd when the volume is high? Actually, greylisting: you may decrease the inboung e-mail traffic a lot. giampaolo
