TL,DR: Looking for a recipient dependent option, similar to sender_dependent_default_transport_maps.
I've got a working solution sending messages through 2 distinct custom transports. This solution solved a problem I had which was *not* recipient dependent. I implemented a unix socket that responds to sender_dependent_default_transport_maps queries with a pre-defined transport name and that worked with minimal performance decrease (lookup time is pretty fast). Fast forward and the dependency has changed: sender became irrelevant and recipient became relevant. Ruling out sender_dependent_default_transport_maps. The problem: Based on the recipient, send that message down a custom transport (defined in master.cf) and this should be applied to outbound messages only. By outbound messages I mean messages sent to domains other than $mydomain. So I tried transport_maps. And it works, but ... When the socket is queried for "*" or $mydomain, it replies with ":" (null transport and null nexthop therefore _do not change: use the delivery transport and nexthop information that would be used when the entire transport table did not exist._) I have two extra queues (anti-virus and secret-biz-sauce) and that means transport_maps is queried a several times for every message and as result the time spent on lookups became a problem. What would be the best (or most efficient) place to plug a recipient dependent transport map? Thanks Marcus
