[Mimedefang] sanity Check: MD -> SpamAssassing checks ~/.spamassassin/user_prefs

[David F. Skoll]

On Thu, 02 Aug 2012 16:56:24 -0400
"Kevin A. McGrail" <KMcGrail at pccc.com> wrote:

> > I suppose you could enforce the former by telling sendmail to temp
> > fail RCPT commands after the first, to get messages in one
> > recipient at a time.

> That's brilliant.  I wonder if most MTAs will properly handle that!

s/brilliant/brilliant, but evil and certainly inadvisable/

Imagine a message for 50 local recipients (not unheard of for a school or
other large organization.)

Imagine a sending MTA with a queue interval of 30 minutes (not unheard of.)

The last recipient will wait more than a day to receive the mail.
The sender may very well receive a "Delivery Delayed" notification.
Not good.

Here's how CanIt handles this situation (in case anyone cares):

We divide recipients into streams.  Everyone in a given stream has exactly
the same rules, settings. etc.  If mail comes in and all the recipients
are in one stream, great!  We process it.

Otherwise, we mail out new copies (one per stream) and discard the original.
When the new copies come in (thank you, Sendmail, for the split queue!) we
process them.  Because we notice they're from the local host, we don't
bounce during the SMTP conversation and we also parse out the "real" relay
from the Received: header and do a bunch of other things slightly differently.

We lose the ability to reject during the SMTP session, but alas it's the only
practical solution, IMO.

All of this is driven by a SQL database.

Regards,

David.