Excerpts from transarc.system.ietf-822: 6-Oct-01 Re: Loop control James
M Galvin(_at_)acm(_dot_)org (1136*)
>(2) Store, in the MTA, the Message-IDs of messages passing.
here, to, the problem is that the message can legitimately transit
the same site more than one time.
Right. My old MTA stores the Message-ID and the (canonicalized)
destination address for each delivered message, for a few days.
Including the destination address means that the message will be
forwarded as often as necessary, as long as it's to different recipients
The only intrusiveness this method requires is that if the MTA forwards
a message to some other MTA--that is, not doing final delivery
itself--it must ensure that there is some Message-ID on the message
rather than none.
It would be a little more bulletproof if it also recorded and matched
something resulting about the message body, like perhaps its size. I've
seen this method reject duplicates that had severely-truncated bodies.