[dns-operations] named actually asks for ANY

[Paul Vixie]

> Here I (and possibly one or two others) had the belief that named
> would "never ask for ANY".

no.  the RFC isn't clear on this, so named will forward an ANY query to
the closest enclosing servers ("iterate") if two conditions are true:
first, RD=1, and second, the local cache has no records under that name.

in an early EDNS1 proposal which was never accepted by IETF, i proposed
to make this behaviour controllable with new flag bits, and to add the
possibility of requiring that "what's in the cache is the result of an
ANY query".  the consensus was "that's too many moving parts for too
little benefit" so i dropped the matter.

a recursive nameserver who ignored RD and always forwarded, or ignored
RD and the cache and always answered based only on what its cache had,
would be as RFC-compliant as what BIND currently does.