DevHeads.net

sender_dependent_default_transport_maps

Hello list -

We have some fairly involved routing requirements, and have been using a
script that creates a transport table from a number of source files.
It's been working well for some years, but now we have a need for
sender-dependent transport rules. We periodically creates the
sender_dependent_default_transport_maps, which appeared to work
perfectly, but then we discovered that the transport table overrides
sender-dependent transport - exactly as documented.

We have a requirement for sender-dependent transport rules that override
everything else. I thought of setting up another postfix instance just
to handle the sender-dependent transport before handing it off to either
the current smtp server or one of the designated transports, but it
seems like overkill. Is there any other way to make
sender_dependent_default_transport_maps take priority over the transport
table?

Thanks for your input -

Joe

Comments

Re: sender_dependent_default_transport_maps

By Noel Jones at 05/07/2012 - 22:04

On 5/7/2012 8:58 PM, Joe wrote:
The transport priority order is not configurable.

I suppose you could use a check_sender_access map that returns
FILTER transport:nexthop for the target senders. Note this solution
is only useful with mail submitted via SMTP and is incompatible with
an after-queue content_filter (unless you do some master.cf gyrations).

-- Noel Jones

Re: sender_dependent_default_transport_maps

By Viktor Dukhovni at 05/07/2012 - 22:15

It also breaks mail routing on any real MTA that needs to route
different recipients to different destinations. The only real
use-case for sender-dependent routing is on shared laptops and
home machines where all of a user's initially submitted mail
is relayed via that user's ISP, but then one just empties
out the transport table and voila, the default transport wins.

My advice to the OP would be to separate the sender-dependent
traffic onto a separate MTA that does no (normal) recipient
dependent routing.

Re: sender_dependent_default_transport_maps

By Joe Sloan at 05/09/2012 - 13:41

On 05/07/2012 08:15 PM, Viktor Dukhovni wrote:
Thanks for the sage advice - After some consideration and a bit of
testing I think we finally have a reason for going to a multi-instance
postfix configuration. It's either that, or spin up a new server to
handle sender-dependent transport, and our support organization charges
by the server, so postmulti seems to be the best option.

Joe