[Top] [All Lists]

Re: What is the history of 2821 and implict MX?

2008-04-09 15:31:51

At 14:29 -0500 on 04/08/2008, Pete Resnick wrote about Re: What is the history of 2821 and implict MX?:

The MX rule of 974 is perfectly reasonable to apply to IPv4 and IPv6.

RFC 974 also requires the use of a WKS record to verify that the IPN address has a MTA running on it. Since we do not use WKS records for this purpose, I'd suggest that for IPv4 only MTAs (handling a FQDN) the RFC 974 rule of the MX being optional be observed BUT once FQDN maps to an AAAA record (or both AAAA and A) that the MX be required in the case where all the IPNs that are mapped to FQDN are NOT running a MTA so that the RHS can supply a FQDN1 other than FQDN (IOW: Use the list of RHS FQDNs to supply the function that WKS would have supplied). Thus the MX is still optional if all IPNs mapped to the FQDN are running a MTA.

IMO: The implicit MX (ie: Generating a FQDN MX 0 FQDN) per RFC 974 is ONLY valid if every host that is mapped via the FQDN IN is a MTA. So long ANY are not running a MTA then the lack of a MX that points ONLY at MTAs is a definition problem (since non-MTA running IPNs in the FQDN IN list can be erroneously contacted to try to pass them relayed email).

Note that this is a change in my prior "MUST have MX for AAAA records" stance. I now would like to see EITHER a MUST have an MX if there is a IPv6 Server requirement OR an EXPLICIT statement that the MX is optional only when all IPNs are running a MTA.