On Fri, 23 Oct 1998, Georg v.Zezschwitz wrote:
> I consider your plans to be a brilliant idea that would perfectly
> help with some of the most recent problems we have:
Yes, those are exactly the kinds of problem I was trying to do something
about.
> E.g. they would not recognize 'Bye, John\0\n.\n' as a valid
> end of the DATA-section.
Well, strictly, those \n characters should really be \r\n.
> > (B) A message error is associated with a particular message when sent to a
> > particular host, but not with a particular recipient of the message. The
> > message errors are:
> >
> > . Any error response code to MAIL FROM, DATA, or "."
> > . Timeout after MAIL FROM
>
>
> . SIZE-feature in EHLO-response with a SIZE below the message size
> . Timeout after DATA
> (See above)
Yes, I agree with both of these and will modify the implementation.
> > This means that the address will not be processed
> > again, in any message, until its retry time arrives.
> ****************
>
> I think this policy is acceptable, but this might be possible:
>
> MAIL FROM: <mailing@???> SIZE=<the size is BIIIG>
> 250 You are welcome
> RCPT TO: <gvz@???>
> 250 O.k., Georg love's to get big mails
> RCPT TO: <poorboy@???>
> 420-Sorry, this user just has a 10 MB-POP3-box, your mail does not
> 420-fit in at the moment...
>
> poorboy will get no further messages from us, till his retry time
> is reached...
I realize this is a problem. Maybe I can do something about it.
Some time ago Exim was changed so that directing a local address and
doing local deliveries were not subject to retry delays except when
running the queue. Thus when a message first comes in, local addresses
are directed and local deliveries attempted once, before getting
delayed. This helps with a similar problem to the above with local
quotas and also with problems like broken filter files that get fixed.
Of course, local directing and delivery attempts are quite cheap.
The same change could be made to the retry delay on an address that had
previously had a 4xx response; an attempt would then be made to deliver
to such an address when the message arrives. If that failed, subsequent
queue runs would delay the address. This means that if many messages
arrive for that address, more delivery attempts will be made, but that
won't be any worse than the current release. Does this seem reasonable?
--
Philip Hazel University of Cambridge Computing Service,
ph10@??? Cambridge, England. Phone: +44 1223 334714.
--
*** Exim information can be found at
http://www.exim.org/ ***
