From: Michael Alan Dorman on
> The transport map can reject a recipient at SMTP RCPT TO time,
> by resolving the recipient to the error(8) or retry(8) transport.
>
> The transport map must therefore be searched BEFORE the filter.

I had not considered that. Ah, well, with 2.6, multi-instance isn't
such a huge burden.

Mike.

From: Victor Duchovni on
On Thu, Mar 11, 2010 at 03:12:04PM -0500, Michael Alan Dorman wrote:

> > The transport map can reject a recipient at SMTP RCPT TO time,
> > by resolving the recipient to the error(8) or retry(8) transport.
> >
> > The transport map must therefore be searched BEFORE the filter.
>
> I had not considered that. Ah, well, with 2.6, multi-instance isn't
> such a huge burden.

Indeed. I don't use content_filters very much. I just have pre-filter
instances with:

mydestination =
local_transport = error:5.1.2 Mailbox unavailable
default_transport = scan:[127.0.0.1]:someport
relay_transport = $default_transport
virtual_transport = $default_transport

--
Viktor.

P.S. Morgan Stanley is looking for a New York City based, Senior Unix
system/email administrator to architect and sustain our perimeter email
environment. If you are interested, please drop me a note.

From: Victor Duchovni on
On Thu, Mar 11, 2010 at 03:31:21PM -0500, Michael Alan Dorman wrote:

> > And do use "proxy:ldap:" rather than "ldap:" for virtual_alias_maps,
> > and other tables that are used by smtpd and cleanup. Maintain a
> > simple (indexed file) transport table that routes domains, not users.
>
> Fortunately, the transport map is the only thing for which we use
> LDAP.

Sadly, this is the one place where introducing LDAP latency and potential
unavailability has the maximum negative impact. The queue-manager calls
trivial rewrite when bringing mail into the active queue, this is a critical
function that is not parallelizable.

> Am I right in assuming that since there's only ever one trivial-rewrite
> process, using proxy:ldap is just adding an extra layer to no avail, or
> are there other benefits that would still suggest using it for this
> purpose?

Yes, using "proxy:" for transport is counter-productive.

--
Viktor.

P.S. Morgan Stanley is looking for a New York City based, Senior Unix
system/email administrator to architect and sustain our perimeter email
environment. If you are interested, please drop me a note.