[EMAIL PROTECTED] wrote:
One comment I neglected to make in my previous response was, that
encouraging the use of PI, even when used on the far side of PA NAT,
means that the identity of (and NOC contact info for) anyone who
accidentally leaks packets with a source for which there is no route,
into the DFZ, is more likely to be discoverable, by mere virtue of the
assignments and registration that go along with PI space.
Wouldn't that be equally true of ULA-C space?
And in response to your earlier post encouraging the use of 1:1 NATv6,
wouldn't it be just as easy/beneficial to use ULA-C or ULA-L space for
the private side of the NAT? The only difference would be in the
multihomed case, where allocation of globally-routable PI space could be
delayed until the recipient applied for their ASN, and the recipient
network could transition the public side of their NAT to use the new PI
space, leaving the inside of the network as ULA or ULA-C.
Now all of this ignores the fact that, architecturally, NATs have been
rejected as bad for end-to-end connectivity. To minimize problems
caused by NAT, it would probably be better to use autoconfiguration
(stateless or DHCP) to assign PA addresses to hosts (in addition to any
non-globally-routed space used).
-Scott
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
