"Roger B.A. Klorese" <rogerk@QueerNet.ORG> wrote:
>
>This stuff was designed at a time when (a) all remote sites WERE nice
>guys, and (b) everyone was clever enough to know what their resources are.
Yes, point (a) is true, but they were also smart enough to realize
that (1) that wouldn't always be the case and that even nice guys
could run amuck, and (2) there had to be a way for a system to control
incoming connections.
As for (b), it's impossible for each site on the Internet to know
every other site's resources. That's why the remote sites *have* to be
responsible for controlling their own resource usage.
The fact is, the lower layers in the stack were designed for just such
a thing as queueing connections until the system was ready to handle
them.
>If I only have the CPU and memory resources to support ten simultaneous
>listeners, you can deliver mail to me much more efficiently by sending all
>of it to one listener. This has nothing to do with throttling bandwidth.
Efficiency is not the only criterion for determining what to do.
Performance, reliability, and capability are others. qmail is willing
to trade network bandwidth (efficiency) for delivery rate
(performance), simplicity of design (reliability), and VERP
(capability).
-Dave
Follow-Ups:
References:
|
|
