From: mouss (no email)
Date: Mon Apr 16 2007 - 10:38:23 EDT
Черкендов Александр Армаисович wrote:
> ----- Original Message -----
> From: "mouss" <>
> Cc: <>
> Sent: Monday, April 16, 2007 3:58 PM
> Subject: Re: Empty MAIL FROM
>> Mark Martinec wrote:
>>>> you must accept <>, but you can use reject_multi_recipient_bounce in
>>>> restrictions, for example in smtpd_client_restrictions
>>> You may, but the use of a null return path is legitimate
>>> for multi-recipient messages too.
>>> RFC 3461:
>>> When attempting to relay a message to an SMTP server that does
>>> not support this extension, and if NOTIFY=NEVER was specified for
>>> some recipients of that message, a conforming SMTP client MAY
>>> relay the message for those recipients in a separate SMTP
>>> transaction, using an empty reverse-path in the MAIL command.
>>> This will prevent DSNs from being issued for those recipients by
>>> MTAs that conform to .
>> I remember you (Mark) sent some "false positives" a long time ago.
>> another case is a bounce that is subject to alias expansion. rare, but
>> could happen.
>> anyway, if the check is done in smtpd_client_restrictions, then that
>> should only result in the message being sent multiple times, once to
>> each recipient. this is bad, but still acceptable. aa real problem
>> occurs if the check is done in data stage.
> is it wrong to check reject_multi_recipient_bounce in smtpd_client_restrictions?
the first thing is to check your logs and see if it blocks spam. I doubt
it. most spam comes with a forged sender, not an empty one.
so, I've never seen it block spam, but it has been reported to block
legitmate mail. and from a standard perspective, nothing prohibits a
multi-recipient mail with a null sender.