ietf-smtp
[Top] [All Lists]

Re: Queued Mail or Unreturnable Mail?

2008-05-03 05:45:24

Robert A. Rosenberg <hal9001(_at_)panix(_dot_)com> wrote:
At 18:00 -0400 on 04/30/2008, John Leslie wrote about Re: Queued Mail 
or Unreturnable Mail?:

We certainly could design a mechanism _better_ than MX records to
document intent relative to receiving email, especially DSNs. But unless
and until we do, folks are likely to use MX records as part of the
balancing act of guessing the probability of NDNs reaching a responsible
reader.

IMO: The publication of a MX is a statement to "Send *(_at_)FQDN Addressed 
Email HERE". Once an SMTP Server does so and delivers a NDN message 
addressed to *(_at_)FQDN to the MX, it has done its job and turned over 
Delivery Responsibility to a MX designated Server. The fact that the 
MX Server (or one that it forwards the NDN message to) then turns 
around and sends is to /dev/null is not my SMTP Server's Problem 
since, as noted, it did its job by delivering the message to the 
Domain Designated injection point and turning over further delivery 
processing to the Domain's Mail System.

Thus the existence of the MX places the responsibility of "reaching a 
responsible reader" on the MX not the Sending SMTP Server with NO 
GUESSING involved.

   An excellent point. The MTA generating a DSN has no obligation to
guess the probability of that DSN reaching a responsible reader if
there is an MX record for the domain in question. If that MX directs
it to /dev/null, there's nothing the sending MTA could have done to
increase its chances.

   Conversely, if there _is_ no MX, it's a pure guessing game. Will
a particular IP address gleaned from a A)ddress record answer on
port 25 at all? If it doesn't answer now, might it answer an hour
from now? Might a different IP address gleaned from the same A)ddress
query give better odds? If we finally get a SMTP session, but get
"no such account", would it help to try a different IP address?

   My personal inclination is to declare "guessing" out of scope for
returning DSNs. We already know that not every domain wants them.
Declaring that one MUST send them even when the receiving domain has
not expressed an interest in receiving them leads to known problems
_today_. It would be good, IMHO, to have a clear way to declare an
interest in receiving them, or _not_ receiving them, and recognize
that in the absence of such a declaration, some MTA administrators
will choose to send them, while others won't.

--
John Leslie <john(_at_)jlc(_dot_)net>