Chris Bartram wrote (privately):
>
> [log in to unmask] writes:
> >
> > This is simple mail congestion most likely; as I've explained
> > before, if an SMTP mail delivery transaction does not complete with
> > the final OK and clean socket shutdown, SMTP/sendmail/and friends
> >resend the message.
>
> I'd agree normally; I've gotten more than a dozen dups myself. And I
> know that with some of the changes we've made to our mailservers
> lately, they now can take much longer to verify a message (OK response
> after all DATA received). I've also seen alot of reports of timeouts
> from raven.utc.edu [...]

Our Internet feed has been *quite* congested lately, this is a
possibility.  We don't have a dedicated line, we're on the state "ring"
and thus often don't get the bandwidth we think we have.

> so I wouldn't be surprised...except...
> Most of the dups I've gotten are 3-5 days after the original was
> received.  With lots of other hp3000-l messages received in the
> interim. I would expect the dups to come much sooner if it was a
> timeout/retransmission problem. ???

Indeed.  I haven't quite nailed this one down yet, but perhaps some
network sleuths out there can help (if you receive and/or can view
full mail headers).

I explained the "Listserv bounce" problem caused by the Italian
subscriber's mailer bouncing to HP3000-L, and the proper recognition
of the post as a duplicate by Listserv.  but...

Listserv detects "duplicates" by checksumming the message body and
storing these checksums to compare future posts against.  However, it
can't keep this up forever, and I think the expiry period is 48 hours.
That would explain your observation of duplicates "3-5 days after the
original".

If you have in your posession a set of duplicates with headers, I'd
appreciate them being forwarded to me, or else check for yourself as
follows:

Check the Received: lines from top to bottom paying particular attention
to the "Message-id:" values.  At the uppermost level (your mailer) you
should obviously have different IDs.  But at some point down the line I
would expect either "extra" received lines (duplicates reached you from
some other source) or for the IDs to converge.  If so, the duplication
happened in transit between the "matching" and "differing" hosts.
However, if the IDs remain different, all the way to the original sender
then something is seriously amiss!

Jeff Kell <[log in to unmask]>