Some confusion here, probably because I have used the wrong terminology. When I said I have a content filter that needs to influence how a message is routed, I meant a content filter that is implemented as an SMTP proxy sitting between two Postfix instances.
That said, Wietse's response led me to a solution I think: The SMTP proxy can add an X-header indicating where the message should be routed. In the upstream Postfix instance I can configure header_checks to check for the X-header and return a FILTER response based on the header value using a regex. I'll need to take steps to ensure this can't be exploited, e.g. by stripping the X-header if it exists in the downstream Postfix instance, or in the SMTP proxy. Would that work? TIA, Rob On 2 June 2016 at 18:29, Wietse Venema <wie...@porcupine.org> wrote: > Viktor Dukhovni: >> >> > On Jun 2, 2016, at 12:04 PM, Rob Maidment <rmaidm...@gmail.com> wrote: >> > >> > I have a content filter that needs to ensure certain messages are >> > relayed to a specific nexthop MTA, based on the message content. How >> > can the content filter influence how the message is routed? >> >> That's what content filters do. They preempt the routing of messages >> by recipient domain, and route all the recipients to the selected >> transport:nexthop. >> >> FILTER smtp-or-clone:[nexthop-host.example.net] >> FILTER smtp-or-clone:nexthop-domain.example > > As currently implemented, FILTER is a message property, not a > recipient property. That means, there is no per-recipient FILTER. > > Wietse
