Chaining MX records illegal?
kcd at daimlerchrysler.com
Thu Nov 10 02:15:12 UTC 2005
Chris De Young wrote:
>Am I correct in my interpretation that chaining together MX records is illegal,
>despite the fact that it seems to mostly work?
>chud.net mx 10 mail.foobar.com.
>mail.foobar.com mx 10 smtp.yahoo.com.
>is a no-no?
As an Internet Standards issue, this is a fairly interesting question.
After defining the structure of the MX Resource Record, RFC 1035 punts
to RFC 974 as far as how MX records are *used* (i.e. semantics rather
than syntax), and the successor to RFC 974, RFC 2821, simply says "When
the lookup succeeds, the mapping can result in a list of alternative
delivery addresses" without specifying whether those "addresses" must be
exclusively A records or can be MX records or, presumably even something
else like SRV records. One must also take with a grain of salt the
singular number of "the lookup" in the quoted text above, since even in
a "normal" case there's no guarantee that all of the A records
associated with the MX records in the Answer Section of the MX-query
response will be present in the Additional Section of the response, so
the SMTP client might have to make one or more subsequent queries to
translate those MX targets into A records anyway.
As a practical issue, though, why would you want to tempt fate by making
your mail delivery dependent on a construct that was apparently not
intended or foreseen by the standards writers and/or implementors, and
at best only "mostly" works? I don't see any benefit to it.
More information about the bind-users