> >Hm.. we're required to have a postmaster alias for each of our virtual
> >hosts. Are you sure you want to have Majordomo take care of those? It
> >doesn't seem like a good idea to me.
>
> If we were to implement the above, we would need a number of forwarding
> aliases anyway (owner-<listname> for example). The postmaster one
> wouldn't be much different.
Under certain network conditions, a great flood of mail can arrive very
quickly for owner-<listname> addresses, so that it is not appropriate to
process it immediately with a perl program which is as slow to start up as
Majordomo. Hence anything which processes mail for owner-<listname>
addresses should have a C program (which doesn't need to be very
complicated) to spool bounces.
Hence the set-up which you're suggesting will require postmaster mail for
the virtual hosts to pass through three programs (namely the MTA, then that
C program which sees that it's addressed to a /postmaster/i address and
hence forwards it to mj_email immediately, and then mj_email which would
look up the postmaster's e-mail address in the configuration file) when it
would suffice to pass it through one program (the MTA). The reason why I
dislike this especially for the postmaster alias is that if Majordomo
doesn't work, then it is important for people to be able to report the
problem by means of the postmaster address.
For this reason I strongly believe that the current approach is much better.
The recommended set-up will be for Majordomo to use its own aliases file,
and eventually Majordomo will be taught to manage this file automatically.
I think this has all of the advantages of the approach which you've been
suggesting, and in addition the postmaster address is more robust (and the
other special addresses of RFC2142, too).
-- NB.
References:
|
|
