RE: (ngtrans) Last call on draft-ietf-ngtrans-dns-ops-req-02.txt

[Christian Huitema <[e-mail]>]

> > Again, I'd say "no".  For the next very many years (unless something
> > unexpected happends) we'd be able to manage just fine with
dual-stack.
> 
> mandating a dual-stack on all v6-capable hosts might be an approach.
but
> i suspect it will not be acceptable in some application/vendor
domains.

If an IPv6 only host wants to resolve a name that is served only by IPv4
capable DNS servers, it can send the request through a protocol
translating NAT (NAT-PT), and it will receive the response. There is an
operational issue, the fact that the current NAT-PT specification
mandates an "application layer gateway" treatment of DNS packets,
automatically translating A records into AAAA records; I find that a
poor trade-off between convenience and robustness; we may need to
revisit the issue. Nevertheless, NAT-PT should provide a generic
solution for "v6 hosts discovering v4 hosts."

The opposite problem, v4 only hosts discovering services provided in v6
only domains, is somewhat harder to solve. Indeed, at some point it will
become impossible to solve simply by address translation, since 2^32 <
2^128.

-- Christian Huitema