Henning Brauer wrote:
> QMQP is far more efficient for this kind of usage. Using SMTP makes no sense.
> I consider this "having some different mailservers accepting mail for the
> whole domain and one qmail-ldap host as dispatcher" a broken design.
Actually the design was "any qmail-ldap host can accept mail for the
whole
domain".
My next question is this: Is it safe to leave the QMQP port open to the
world, or is it a security risk (for unauthorised relay)?
> This is
> by far better implemented with different hostnames (e. g.
> sendmail.example.com, exchange.example.com, someotherbrokenmta.example.com)
> and mailForwardingAddress attributes.
No it isn't - this adds the added kludge of having to give people actual
userids tied to a specific box - which adds unnecessary complexity -
which means extra work and potential for misconfiguration - yuck.
The simplest way is to simply have an email address and a mailhost: If
the mailhost is defined, and it's not me, then deliver it there. Simple.
Rewriting email addresses using address forwarding is unnecessarily
complex and therefore bad design.
> > Using the mailHost attribute with SMTP is the most logical assumption.
>
> No. It's a misunderstanding.
Netscape Messaging server works this way, so there is precedant out
there for this kind of behavior. Yes, SMTP is heavy - but it works. My
primary design requirement is not performance, but the ability to handle
distributed operation in a *simple* way.
> > If I submitted a patch, would you consider it for inclusion?
>
> That's up to Andre, but I wouldn't.
>
> A quote comes to my mind:
> "If you want a bloated MTA whoms author happily includes every patch sent to
> him, use Exim."
> (Felix von Leitner on the qmail list)
Is this kind of approach to problem solving really necessary? Having
been involved in the Apache project for over two years now I have always
been amazed that despite all the arguing, debating, re-coding, vetoing
and carry on that I have watched, at no point has anyone put up with the
argument "I'm right, so stuff the rest of you".
There are many ways of achieving something. Where there is more than one
way of doing something, and where there is debate over which is the
right way, the software should allow both ways, it's simple.
Regards,
Graham
--
-----------------------------------------
[EMAIL PROTECTED] "There's a moon
over Bourbon Street
tonight..."
S/MIME Cryptographic Signature