On Mon, Sep 22, 2003 at 07:42:10PM -0500, Rodney Richison wrote:
As a postfix user, setting up a box for another bussiness that plans to
handle 15 or so domains, I've begun implementation of the ole trusty
postfix+courier-imap+mysql+maildrop+sqWebmail solution. It dawned on me,
I have never even installed courier webmail. So, here I am. Blatantly
asking up front, what will the surprizes be? In particular, I noticed
this in the documentation.
**************************
Courier does not deliver mail to domains with broken MX records. Courier
also refuses to accept any mail with a return address in a domain with
broken MX records.
It is unacceptable to have a mail server which controls policy. Have I
misread? This seems to indicate if no mx record exists for a mailer,
Courier will not accept it. Is this true?
I think you've at least partly mis-interpreted. What it means is that
if an MX record is broken then courier won't try to guess at what its
supposed to mean. Two common examples are (in BIND/RFC format):
1) example.com. IN MX 10 10.0.0.1
This is bogus because an MX record has to point to a valid A record.
2) example.com. IN MX 10 some.cname.example.com
some.cname.example.com. IN CNAME some.other.example.com.
This is bogus for the same reason as in (1). A CNAME is not a host record.
So for these courier won't try to guess what the DNS admin was playing
at. While it seems obvious what both (1) and (2) are trying to do,
they're not RFC compliant and they make writing and maintaining an MTA
difficult (i.e. why do a special case for tracking down 5 layers of
CNAME's when its specifically not allowed?).
Most of couriers other policies are configurable at run-time or at
compile time (in fewer cases).
-Peter