Thus said Robert Elz on Sat, 25 Apr 2015 06:30:10 +0700:
If gmail is predicting that Bob was going to receive a copy because of
the address in the To/Cc headers, then it is wildly broken - by the
time SMTP gets involved, those are no more than comments (as Ken said
in an earlier message, I have unusual levels of control over my mail
system - it is not without precedent for me to delete delivery to one
recipient before the message leaves my MTA - but they will still be
listed in the header fields).
This is a better observation, however, I still believe that Gmail does
duplicate removal, so I hereby revise my statement.
Both Gmail and the list server are ``removing duplicates'' and so the
only copy that Bob was allowed to receive was the delayed message
(because the MLM failed to deliver a message to him) and there was a 33
hour delay due to other factors in the direct delivery.
It's unlikely that Gmail would have ``waited'' for the directly sent
message knowing that it would eventually arrive based on past analysis
of email delivery patterns to Bob's mailbox.
It's highly likely that the MLM is deciding not to deliver a copy to Bob
because he's also subscribed. I think this behavior is wrong, because
just as Gmail cannot predict that the MLM address will eventually arrive
in Bob's mailbox, the MLM also cannot predict that the directly
addressed email will ever arrive in Bob's mailbox. Therefore, the most
sensible thing for an MLM to do is just to send the message.
If this is true, you will not receive two copies of this email I am
sending to you since the MLM will not deliver a copy of it to you. Also,
I did not receive a copy via MLM of this email that you sent to me to
which I am replying, which seems to also confirm that the MLM is
deciding not to send a copy where the To or Cc also includes the
targeted recipient.
His concern was the delay - he'd seen replies to a message that he
didn't receive for days (and if I'd noticed the holdup, and dropped
that copy of the message - expecting him to receive, or have already
received, a copy via the list - might never have seen).
I see. The concern was the delay (which at the time probably looked like
100% delivery failure), not the duplicate removal. Personally, I don't
mind delay---SMTP was designed for reliable delivery, not instantaneous
delivery. On the other hand, a mail system that ``removes duplicates''
seems to break the reliability of the system.
Thanks,
Andy
--
TAI64 timestamp: 40000000553ad870
_______________________________________________
Nmh-workers mailing list
Nmh-workers(_at_)nongnu(_dot_)org
https://lists.nongnu.org/mailman/listinfo/nmh-workers