Re: [Exim] Stopping out-of-office auto-reply mail loops

Top Page

Reply to this message
Author: David Woodhouse
To: Exim Users' Mailing List
CC: Andrew Lewis
Subject: Re: [Exim] Stopping out-of-office auto-reply mail loops
On Fri, 2004-01-23 at 14:17 -0500, Greg A. Woods wrote:
> [ On Friday, January 23, 2004 at 10:03:14 (+0000), David Woodhouse wrote: ]
> > Subject: Re: [Exim] Stopping out-of-office auto-reply mail loops
> >
> > Autoreplies should only ever be sent to the SMTP reverse-path;
> That is very _VERY_ wrong.
> A proper out-of-office/vacation autoresponder is what's called a "mail
> agent". Such programs _should_ behave exactly as a user using a proper
> MUA would behave.


What we were looking at was clearly an example of implementing this in
the MTA, with no user present. An MUA gets to use those dangerous
addresses because, by definition, it has a _user_. The user isn't going
to _voluntarily_ participate in a mail loop.

Vacation messages aren't generated by an MUA, unless the user configures
it extremely strangely and leaves it running in her absence. They're
delivered by the MTA or more correctly I suppose the MDA. The whole
point is that the user is _absent_.

> The sender address is an unauthenticated (and unauthenticatable) address
> that _should_ only ever be used by an MTA as the destination for notices
> about the non-delivery of a message that MTA had taken responsibility
> for delivering.

It's also used for notices of _delays_ in receipt. The sender address is
not for replies, that's true -- it's for notification to the _sender_
(as distinct from the From address) about what has happened to the

What is a vacation message if not a form of delivery status notification
-- a notification that the message has not been received by its intended
recipient yet, just like the messages which inform the sender that it's
still on the queue after three days? Just like those and DSN (cf.
RFC1891) it should go only to the SMTP reverse-path; nowhere else.

Consider the example from page 42 (§A.1.1) of RFC2822:

      If John's secretary Michael actually sent the message, though John
      was the author and replies to this message should go back to him, the
      sender field would be used:

From: John Doe <jdoe@???>
Sender: Michael Jones <mjones@???>
To: Mary Smith <mary@???>
Subject: Saying Hello
Date: Fri, 21 Nov 1997 09:55:06 -0600
Message-ID: <1234@???>

This is a message just to say hello.
So, "Hello".

In this case, the SMTP reverse-path is (as usual) going to match the
Sender: header. Delivery failure reports go correctly to Michael the
secretary, as do reports of the mail being delayed and still on a queue
somewhere. Reports of the mail being delayed because Mary is on vacation
should _also_ be going to Michael -- not to John.