On Thursday 01 February 2007 14:05, Mike Cardwell wrote:
> * on the Thu, Feb 01, 2007 at 01:54:21PM +0100, Jakob Hirsch wrote:
> > We could introduce modifiers for the delay keyword (like "delay =
> > 10s/noflush"), if somebody wants the old behaviour, but it's probably
> > not worth it.
>
> If you're only using very short delays and autoflushing, it half defeats
> the point of pipelining come to think of it. Perhaps, there should be a
> control to enable/disable it?
The shortest delay is one second. That is in the vast majority of cases well
above the round-trip time. PIPELINING is meant to save time by saving
round-trips. That time is gone anyway when you deliberately delay the SMTP
dialog. Fewer round-trips also means much less overhead to the MAIL, RCPT and
DATA commands, and slightly less overhead overall. But does that matter? In
any case, "delay" is not something you use in normal cases, is it? It's
something you typically use to slow down [suspected] spammers. (Always
inserting a short pause before the initial banner is a common trick to trap
ratware that doesn't follow the sync rules, but that is before PIPELINING
comes into action.)
--
Magnus Holmgren holmgren@???
(No Cc of list mail needed, thanks)
"Exim is better at being younger, whereas sendmail is better for
Scrabble (50 point bonus for clearing your rack)" -- Dave Evans
