-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On 10 Mar 00, at 10:56, Pavel Kankovsky wrote:
> Current implementation does not avoid the execution of "one zillion
> system calls", it postpones it until qmail-send decides to deliver the
> message later. Yes, there are benefits: reduced latency of
> qmail-smtpd, ability to absorb briefs peaks of load but sooner or
> later, you have to do the validation anyway.
Agreed.
> Let us assume the implementation of "early validation" is simpleminded
> and does the validation twice (both in qmail-smtpd, and in qmail-local
> or whatever does the final delivery), and let us assume its
> installation X collapses when fed with at least N deliveries per
> second. Then the same installation with vanilla qmail collapses at no
> more than 2N deliveries per second because it needs no less than 50 %
> of X's CPU time to do its work.
I beg to differ. qmail-smtpd's are running in parallel, and need the
validation reasonably fast (otherwise SMTP connection times out,
and is retried; with suboptimal retry schedules of some MTA, that's
a disaster, quite likely). On the other hand, the validation in
qmail-local does not have to be that fast and can be run in serial
(the only problem being the queue growing up).
-----BEGIN PGP SIGNATURE-----
Version: PGP 6.0.2 -- QDPGP 2.60
Comment: http://community.wow.net/grt/qdpgp.html
iQA/AwUBOMi6RVMwP8g7qbw/EQINuACfX42qCoZaae2XPyCJynCxL1a+IU4An0jF
ZuMe20DyUskrjTdPvkULAqI2
=JmMn
-----END PGP SIGNATURE-----
--
Petr Novotny, ANTEK CS
[EMAIL PROTECTED]
http://www.antek.cz
PGP key ID: 0x3BA9BC3F
-- Don't you know there ain't no devil there's just God when he's drunk.
[Tom Waits]
