I have long believed that DNS should either encode within itself or have
available to it (via another protocol) information about CIDR delegations.
Once we have a description of multilevel delegations, we can use it to
locate the NS data for address-to-name lookups, and we can also use it for
core aspath access lists as Bill and Vadim are now discovering.
Two proposals for CIDR-style IN-ADDR.ARPA delegation were presented at the
DNSIND WG meeting of San Jose's IETF. Both were thrown out, one due to its
complexity and the other because it had bad failure characteristics (and I
mean Really Bad) during a network partition.
Address->Name translation is suffering more from this than routing is, so
I'm not sure I agree with Bill or Vadim that this really has to be solved.
I'd like to point out, while I've got everybody(?)'s attention, that Vadim
said "whois is bad" whereas Bill said "rwhois is good", thus talking right
past each other. Rwhois is probably the right answer to this problem.