Leslie Mikesell <les@Mcs.Net> wrote:
>
>The protocol authors probably never dreamed anyone would intentionally
>abuse other site's resources in this way. Or (considering the time
>frame) that anyone could afford to waste their own outbound bandwidth.
It'll be hard for us to come to terms until you stop using loaded
terms like "abuse" and "waste". I know you're upset, but that doesn't
help.
The protocol authors undoubtedly realized that a site could control
their level of SMTP traffic simply by refusing connections at a
preconfigured threshold and reallowing them at another, lower
threshold. Relying on the remote site to Do The Right Thing is folly
because (a) not all remote sites are nice guys, and (b) even the nice
guys have no idea what your resources are--nor, in my opinion, should
they have to try to deduce them or assume they're minimal.
>And given that qmail is the *only* major MTA that doesn't make any effort
>to be efficient,
It doesn't make any effort to minimize bandwidth, if that's what you
mean. People who use qmail are more interested in speeding up message
delivery than saving SMTP bandwidth. See Dan Bernstein's Modest
Proposal for an entertaining treatment of the bandwidth topic:
ftp://koobera.math.uic.edu/www/sarcasm/modest-proposal.txt
>I have to think that it is qmail's fault. The argument reminds me of
>the guy who took the slow-start out of his tcp stack and speeded up
>the retry timers and claimed that since his tests showed it was
>faster that way the Internet would work better if everyone did it.
I'm not familiar with the details of slow-start and retry
timers. Sounds like "Death of the Net predicted. Details at eleven."
>If you are filling my inbound bandwidth, how do you expect me to
>deal with it by replacing software?
Replace it with software that, for the umpteenth time, doesn't accept
connections it can't handle. If ten SMTP sessions are all you can
handle with your bandwidth, don't accept more than ten simultaneous
connections. Even sendmail can approximate this with
maxdaemonchildren.
>... If qmail is faster than this as everyone claims, how can
>it avoid overwhelming everything else at a site with many recipients
>getting copies on separate connections?
Don't accept...well, you've heard it before enough times by now that
it's either gotten through, or it's not going to.
-Dave
Follow-Ups:
References:
|
|
