Re: BCP for handling DNS SERVFAIL results

Valdis Klētnieks writes:
> I think Arnt's meaning was "If you get a SERVFAIL trying to get the 
> MX, then you *don't* know there's no MX, so dropping back to an A is 
> improper. The message should be queued and retried later".
Exactly.

I'd say that SERVFAIL is best covered by RFC 2821 section 6.1, which has 
this to say about error handling: «It [the MTA] must take this 
responsibility seriously.  It MUST NOT lose the message for frivolous 
reasons, such as because the host later crashes or because of a 
predictable resource shortage.» Although the DNS is not mentioned, the 
passage doesn't seem to condone speculatively falling back to A in case 
of an MX query timeout, SERVFAIL, or other uncertain response.
Arnt