I have a corner case where I need to allow an emails generated at my site with certain off-site sender addresses to be routed through my MTA to the off-site smarthost officially responsible for the sender domain.
This can be easily done using sender-based-routing. However, it becomes an issues as soon as the recipient is on my side and the off-site MTA therefore routes the email back to my MTA. A loop will be detected, due to my MTA seeing its own Received: header from the previous pass. The smtpd the email generator delivers its mail to is configured with a prequeue proxy virusfilter. If I understand correctly, this precludes rewriting the Received: header on the first pass using postfix on-board equipment. QUESTION 1: Is this correct? I see these work-arounds: A) If I receive all email from said off-site MTA on a non-standard port, loop-detection will not happen. The drawback is that I will have to depend on the off-site MTA for loop-detection. QUESTION 2: Is there a definitive overview of all the ways postfix detects loops and at what stages these are employed? (I mean aside from the source code.) :-) B) If there was a way to make the sender-based-routing conditional (i.e. only use sender-based-routing if recipient domain is not in my relay_domains, otherwise use normal transports) I could shunt emails to my own users to my own back-end MTA without passing through the off-site MTA. QUESTION 3: Can one make sender-based-routing conditional in this way? QUESTION 4: Does this also work with an smtpd configured with a pre-queue proxy filter? C) I could set up a completely separate postfix MTA (not just the smtpd) solely for the first pass from the generator. Then I could simply use the standard transports and relayhost=offsiteMTA. This is not the preferred solution, as it will require either a second (non-standard) pfx installation on the existing system or an additional system with a standard pfx. Thanks for your insight! Michael
