Matthias Andree <[EMAIL PROTECTED]> wrote:
>Ok, here goes the story:
>
>0. stop qmail-smtpd
OK
>1. list your local domain as smarthost in SMTP routes
OK
>2. kill the postmaster aliases, we're sure that the smarthost has a
> postmaster and proper virtual mapping and will catch the mail.
Not OK. Postmaster is not optional.
>3. run qmqpd.
>4. inject a mail with incomplete last line through QMQP.
>
>What happens now:
>
>1. qmail-remote fails to deliver the mail via SMTP, "partial final line"
>2. qmail-send creates a bounce, qmail-remote again fails
>3. qmail-send tries to contact the postmaster, qmail-remote again fails
Because you removed the postmaster alias. Duh.
>4. qmail-send discards the bounce. Mail and bounce are lost.
>
>Reliable workaround: stop running qmail-qmqpd, inject mail through
>qmail-smtpd and qmail-inject instead.
Or recreate the postmaster alias.
>I know that SMTP cannot transfer mail with incomplete last line,
>qmail-remote mentions that.
>
>But how come qmail-qmqpd accepts mail that cannot be delivered across
>all transport and that cannot be bounced?
>
>On a related issue, qmqp is not faster than smtp. SMTP must be
>implemented anyhow, two competing protocols just make the testing less
>thorough.
QMQP requires fewer round trips. For high-latency connections, this
could be important.
>"Keep it simple, stupid." qmail-send was a tad too stupid when creating
>the bounce. It must be able to ultimately send its bounce across SMTP,
>lest it be accused of trashing innocent mail.
Non-RFC-compliant mail is not innocent.
>And that's what you get for ignoring RFC-1894 and 2045..2049 for your
>bounces, they also fail.
Preachin' that ol' time DSN religion, Matthias?
-Dave
