Re: ocean: do not boil

[Steve Deering <deering@cisco.com>]

I agree with Alain: the domain of applicability of NAT-PT (or other
translation approach) is not so much networks/devices that are today
running IPv4, but rather new kinds of networks/devices for which
(for whatever reason) it is undesirable to run both IPv4 and IPv6
in parallel, while still needing some applications to work between
IPv6-only devices and the existing IPv4 Internet.  Examples *might*
be low-cost wireless devices, where dollar- or power-cost of memory
to hold two IP implementations might be the dominant concern, or
the military's "wired warfighter" with his or her body-area network
of IPv6-enabled sensors and weapons, linked into its environment
with IPv6-specific autoconfiguration and ad-hoc routing protocols
(but still needing access to the IPv4 world for some apps, like
email to home).

The specific scenarios where I've found myself suggesting NAT-PT
recently are:

	- for customers who are planning 3G wireless networks,
	  where it is still unclear whether the handset vendors will
	  be producing IPv6-only handsets, or just dual-stack ones.
	  The NAT-PT comes into play only the case of v6-only
	  handsets.

	- for customers who are IPv6 True Believers, whose motivations
	  for deploying IPv6-only hosts and/or networks go beyond just
	  finding the easiest way to add IPv6 to an existing IPv4
	  environment.

And as Alain said, at some point we may reach the stage that some
regular ol' residential and corporate networks are be easier to run
IPv6-only, whether because of strategic choices made by OS-vendors
or specific technical benefits of IPv6 (e.g., being able to get rid
of DHCP servers or whatever).  Translators may become the preferred
way for such networks to maintain access to the v4 world (access that
is no better or worse than what they would get through an IPv4 NAT).

Steve