From: mouss (usebsdfree.fr)
Date: Tue May 02 2006 - 13:42:45 CDT

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Robert Felber wrote:
>
> Well, I have it in the recipient restrictions and (the default)
> smtpd_delay_reject = yes.
>
> Allthough this one stops me when I test out something, but I use
> HELO then, not EHLO. I am not sure whether Wietse means "and" as
> logical "AND" or rather as or.
>
reject_unauth_pipelining is useless at RCPT stage in the following
situation ("logical AND")

- ESMTP is used (Large majority of mail?)
- PIPELINING is advertized in the EHLO response (The recommended default)

This is because you can then send RCPT without waiting for the server
reply to the "previous" MAIL or RCPT command.

If in addition, smtpd_delay_reject = yes, then client/helo/sender
restrictions are performed at RCPT stage, in which case, the above applies.

I have once asked whether it was a good idea to set smtpd_delay_reject =
no and use "sleep 1, reject_unauth_pipelining" in client restrictions
("greetpause"). See the corresponding thread. Also one shouldn't disable
PIPELINING as it saves bandwidth (See a post by Cami).
> Nevertheless, I've seen no problems with ESMTP clients so far.
>
> That's why I was asking why some suggest it to be in the data
> restrictions, where it is "too late".
>
It is not too late. Clients should wait for the response to the DATA
command before sending their message.

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]