From: Wietse Venema (wietseporcupine.org)
Date: Fri May 02 2008 - 12:04:16 CDT

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Victor Duchovni:
> On Fri, May 02, 2008 at 10:41:40AM -0400, Charles Marcus wrote:
>
> > On 5/2/2008, Victor Duchovni (Victor.Duchovnimorganstanley.com) wrote:
> > >It's the best I can do until I find the cycles to fully document and
> > >then release the tee proxy I am using. I don't think it is appropriate
> > >to release it in an undocumented state.
> >
> > Understandable, and I'm sure I and many others look forward to this...
> >
> > With the advancing requirements of burdensome regulations, this (some
> > kind of basic archiving capability) will become more and more important,
> > and in my opinion, it is not unreasonable for postfix to provide an
> > integrated/built-in method of performing this function, to ensure a
> > robust and safe - ie, don't bounce if there is a problem with the
> > mirror, but queue until it is back up, etc - functionality.
>
> I don't expect the proxy in question to ever be part of Postfix. Postfix
> supports at least 4 extension mechanisms:
>
> - Post-queue content filters
> - Pre-queue proxy filters
> - Milters
> - Policy servers
>
> It is up to the Postfix community and vendors to create add-on tools
> that make use of these features. So I don't see a "buit-in" archive
> feature any time soon.
>
> > It would also be nice if it could easily deliver to an appropriate
> > sub-folder - ie, one named after the local address part of the original
> > envelope recipient (the one(s) being tested for during recipient
> > validation stage when the primary server accepted the message for final
> > delivery)...
>
> The right mechanism leaves this choice to the administrator, the
> archive copy is created and queued, after that you can deliver it
> where-ever you want (configure the archive Postfix instance transport
> rules accoringly). It would be wrong to make delivery decisions in the
> archive module, they would never be sufficiently comprehensive.

If we can agree on a usable MIME encapsulation, then it should be
possible to spawn off a message in the cleanup server, after the
Milter processing has happened, and before the (SMTP) client is
notified that the mail transaction is complete.

However, Postfix is a general-purpose MTA, and you can already
configure dedicated delivery channels (with transport maps and
master.cf) that have soft-bounce turned on, so there is no need
for built-in special delivery modes that never bounce. Just set
the maximal queue time large enough.

        Wietse

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]