On 2007-10-18 10:35, Alain Durand wrote:
On 10/17/07 2:49 PM, "Brian E Carpenter" <brian.e.carpenter@gmail.com>
wrote:
In my opinion we have to be very systematic about this. The only way
that we can avoid the intrinsic problems of stateless translation
seems to be to have awareness at the IPv6 end that translation is
happening. (The underlying problem is how to reach legacy IPv4
hosts, who by definition can have no such awareness). That seems to
mean getting rid of the "less" in "Stateless IP/ICMP Translation",
so I think we need to go back to basics. I don't have any specific
proposals in mind, but I'm pretty sure that signalling between
the translator and the IPv6 host will be needed.
We have been studying network-triggered translation vs
end-device-triggered
translation. One thing that came out of that study is what happen when
the
translator in the middle fails. If you have a model where mappings are
triggered by the end-devices, you are going to create a storm of mapping
requests to the fall-back translator.
Yes, that seems correct. I don't see any reason why translation
should be initiated by the IPv6 host, but on the other hand, do
we want the translator to be a single point of failure?