On 27 Mar 2008, at 20:38 , Mark Andrews wrote:
OTOH, I think standardizing this convention makes all sorts of
not, of course, in 2821bis.
Why not in 2821bis? Is 2821bis really that time critical?
I would prefer to see the "empty field" intention implicit in "MX 0 ."
codified with more generality, so that (to give just two examples) it
can also be used in the MNAME field of SOA RDATA to indicate "I do not
accept dynamic updates", or in the RNAME to indicate "there is no
mailbox published in this RR".
UPDATES are supposed to be sent to the NS RRset. If the
SOA MNAME happens to match one of the nameservers listed
in the NS RRset then one should try that first. This allows
UPDATES to work when the primary nameserver is not directly
The client should stop trying to update on REFUSED. If a
nameserver doesn't implement forwarding of UPDATEs then
NOTIMP should be returned.
The DNS RFC's define the MX record format. They do not define
the semantics associated with the record. RFC 2821 does that.
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742 INTERNET: Mark_Andrews(_at_)isc(_dot_)org
IETF mailing list