"Marsha Petry" <[EMAIL PROTECTED]> wrote:
>1. Somebody mentioned a Japanese website that has performance stats
>(http://www.kyoto.wide.ad.jp/mta/eval1/).
That was me.
>That same writer mentioned that
>qmail did good on the comparisons, but honestly, I don't know what the
>graphs mean just looking at the axis values. Can anyone give me a
>description? I understand the values (DNS queries/answers,SMTP syn & fin),
>but to me these would be stats I would expect for an OS or maybe machine.
>What does it mean as far as MTA.. the more DNS queries/time an MTA sends out
>the better? etc...
Each MTA in those tests delivered the same set of test messages. What
you should look at is the shape of the curve and the time-to-
completion. Higher DNS and SMTP rates is generally associated with
higher performance, but a clever MTA can avoid extraneous DNS lookups
and minimize SMTP connections, so that doesn't tell the whole story.
>2. Is DSNs vs. VERPs something I should worry about? What I could gather
>from the thread is that some MTAs send back DSN-formatted bounces and others
>send VERP-formatted...is that right? Can qmail process both as easily?
>Will I have to write additional code to process DSNs? or is "a bounce is a
>bounce" to qmail no matter which format it comes in as?.
DSN is basically a set of standards for bounce messages. It's big and
clumsy, but that's not the real problem with it. The real problem is
that not all MTA's support it. If it was universal, it'd be OK. But
that's not going to happen any time soon.
VERP is a mechanism for processing bounces reliably, independent of
the format of the bounce. As long as an MTA sends bounces to the
envelope return path (ERP), as 99% of them do, VERP will correctly
identify the bouncing address.
VERP works with both DSN and non-DSN bounces.
DSN-based bounce handlers only work with DSN bounces.
-Dave
