Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!nntp.coast.net!howland.erols.net!news.mathworks.com!fu-berlin.de!cs.tu-berlin.de!news.informatik.uni-kiel.de!not-for-mail
From: ca@informatik.uni-kiel.de (Claus Assmann)
Newsgroups: comp.mail.sendmail,comp.mail.smail,comp.unix.bsd.freebsd.misc
Subject: Re: Multiple RCPTs taking time? (was: Sendmail vs. Smail...)
Followup-To: comp.mail.sendmail
Date: 17 Dec 1996 10:10:10 +0100
Organization: Dept. of Computer Science, University of Kiel, FRG
Lines: 41
Message-ID: <595o1i$a9t@mine.informatik.uni-kiel.de>
References: <57tf61$gq7@raven.eva.net> <58rvbf$r6t@news.fsu.edu> <1996Dec1510.41.00.4656@koobera.math.uic.edu> <1996Dec16.152941.1706@isac.hces.com> <594a58$g9a@nntp1.best.com>
NNTP-Posting-Host: mine.informatik.uni-kiel.de
Xref: euryale.cc.adfa.oz.au comp.mail.sendmail:35289 comp.mail.smail:2731 comp.unix.bsd.freebsd.misc:32727
[[Followup changed]]
>:>D. J. Bernstein (djb@koobera.math.uic.edu) wrote:
>:>: You start out not understanding something---e.g., why an
>:>: extra RCPT can cost so much time---and you refuse to check your mental
>:>: model against reality. So you never learn.
In <594a58$g9a@nntp1.best.com> dillon@flea.best.net (Matt Dillon) writes:
> He's talking about the DNS delay... sendmail wants to run DNS
> lookups for each RCPT. If you are mailing a destination with
> broken name servers or a broken network, this will create a delay.
Another question:
What are we talking about?
If I followed the thread correctly, the question was:
Is it better to pack messages together, when they have the same
destination (or at least same MX records),
or is it better to send each on its own (in parallel)?
Dan says the latter (that's why he implemented it in qmail),
and he says his benchmarks tell him that it is the best
in most situations (so far, so right?).
Now the claim is: "an extra RCPT can cost so much time",
and Matt gives as an explanation the time for the DNS lookup.
But: These messages are packed together since they have the
same (modulo MX records) destination system, right?
So: if we assume the "normal" case that we deliver directly
to the destination system (this is most often the case, right?),
then all the RCPTs are resolved by the local DNS anyway.
Is this correct? If it is, then there should be no additional DNS
lookup penalty. A broken DNS at the receiving side is hopefully not the
normal case.
So please, can you explain to us why "an extra RCPT can cost so much
time"? Is it just the timeout with a broken DNS?
Is there a chance to see the benchmarks (and the results) somehow?
Thanks in advance,
Claus Assmann
--
Common sense is the collection of prejudices acquired by age eighteen.
-- Albert Einstein
<URL: http://www.informatik.uni-kiel.de/%7Eca/ >