[Bug-gnubg] multiple copies of list postings

[Jim Segrave]

On Fri 03 Oct 2003 (02:29 +0200), olivier croisille wrote:
> I've just no idea, Lavasoft ad-aware 6 just finished running on my PC, 
> nothing wrong here.
> Sorry for the inconvenience

The problem does not lie with you.

Albert, I think you need to send these headers (or a similar set from
a repeated posting) to whoever maintains the mail machines foxtrot and
salsa, which I assume is your ISP rather than your own self.

Every header is identical up until the last delivery step when
monty-python.gnu.org transfers a copy of a posting to either foxtrot
or salsa, which tells me that it was sent to the list once ny olivier,
passed through the list exploder once and a single copy was prepared
for sending to you. This identical copy was delivered at least 4 times
(the headers you sent) and I have no reason to believe it wasn't
delivered all 15 or whatever times. I can think of a couple of reasons
for this happening:

1) connectivity between gnu.org's mailservers and your ISP is very
   poor, so that it often happens that after sending the email to your
   ISP, the gnu.org mail server does not see the acknowledge of
   receipt from Brazil, so it decides to retry. This can happen, but
   it's rare (there's only one point during the entire process of mail
   delivery where this can happen, anywhere else causes the deliver to
   fail, but you won't get a copy of the message).

The more likely is this one:

2) Your ISP is running a virus filter on incoming mail. When an e-mail
   arrives for one of your ISP's users, after the message has been
   sent, but before the acknowledgement of receipt is sent back to
   gnu.org, the virus scanner checks for suspect contents. If a virus
   had been found, a bounce message would be sent telling gnu.org that
   the mail was not accepted and that no attempt should be made to
   resend it. If the scanner decides the mail is OK, then you ISPs
   mail server sends an acknowledgement saying the mail has been
   accepted and gnu.org can consider the job done. With the current
   Swen virus which is plaguing ISPs, there will be a *lot* of emails
   failing the scanner test, the average size of these is over 150Kb
   each. This virus likes to send to addresses culled from newsgroups, so
   it's quite possible that you, as an active poster, are the target of a
   large volume of these. (Demon does not run virus filters and I am
   getting between 30 and 100 of these every day). If your ISPs mail
   servers are spending too long on scanning the mails, by the time
   they are ready to send the acknowledgement of receipt, gnu.org will
   have timed out waiting and will assume that the delivery failed, so
   it will resend it again in a little while. I've seen this happen a
   lot with customers of Demon who run virus scanners which work
   during SMTP delivery time. I don't know what can be done - your ISP
   probably needs to increase the number of mail servers or change the
   virus scanning so that the mail is accepted first, then scanned and
   deleted afterwards.

   I doubt it's any consolation to know that Unix users can use the
   mailtool procmail (which does many nice things for sorting,
   filtering etc.) to delete any duplicate emails = ones whose
   Message-ID: header it has already seen. If there is such a tool for
   Windows, then you can use it to silently discard duplicates.



-- 
Jim Segrave           address@hidden