(ngtrans) Use of IP6.INT

[Erik Nordmark <[e-mail]>]

Please followup on the ipngwg list only.

There has been some discussion for a while in the IAB and IESG about the
use of .INT. Since this is assigned to ITU to do assignments for
international (treaty?) organizations it is clear that the IETF should
not use .INT for any future use.

The question on the table is whether or not we should switch IP6.INT
to be somewhere else. I'd like to understand the implications of
doing such a switch and in particular it it can be piggy-backed on
the transition from AAAA to A6, DNAME, and binary labels.

There seems to be consensus that .arpa would be a good place to put
things like the IPv6 reverse tree since it already contains
the IPv4 reverse tree. If we go this path we should presumably invent a new
meaning for .arpa. So far I've seen two good suggestions:
	Address Reversing Pointer Access
	Address and Routing Parameters Area

Thus ip6.arpa (or in6-addr.arpa if you want it to be closer to in-addr.arpa)
could be a reasonable domain to use.


Transition:
Looking at the draft-ietf-ipngwg-dns-lookups draft and the binary 
labels RFC (RFC 2673) there already seems to be a transition for 
reverse lookups, althought it isn't explicitly mentioned in dns-lookups draft.

The old AAAA spec says to look in IP6.INT by making a (text) label out of each
nibble and reversing the order.
In the new dns-lookups spec says to look in IP6.INT for a binary label 
consisting of the whole IPv6 address. This can be viewed as 128 one bit 
labels in reverse order. Note that RFC 2673 clearly says that a binary label 
never matches an ASCII label "0" and "1".
Thus for transition purposes it seems like the reverse lookup, just like 
the forward lookup trying A6 and AAAA in some order,  need to try two
different lookups in some order.

Thus it might be realatively painless to make the new way also look in
a different place; such as ip6.arpa instead of ip6.int.


Comments?
   Erik