Remi Respres wrote:
RFC5382 is, by no means, a deployed standard.
Not even a standard-track document (but nevertheless useful to
> comply with).
A BCP is a standard document of IETF, which has little influence
over NAT.
Beside, assuming that ISPS assign 10/8 IPv4 addresses, the only
> customers to be concerned seem to be those:
> - assign 10/8 addresses internally (not typical for default
> behaviors of unmanaged CPEs)
Hmmmmm,
You, perhaps, are worrying about deployment of NAT444 by Softbank
BB, which is, perhaps, the largest ISP in the world, using not
ISP independent addresses nor ISP shared addresses of NAT444 but
10/8 for its internal network.
But as its CPEs are fully under the control of it, it can arrange
so that the private network behind CPEs does not use 10/8.
In my understanding, although NAT444 + 6rd is far from being the
only model to offer IPv6 service while dealing with the IPv4
address shortage, IT IS a deployable approach.
For some ISPs, it has a very good performance/cost ratio.
That's an argument similar to ones heard for these 15 years.
Which, of course, isn't sufficient for it to be wrong.
Well, NAT444 with LSN is not a good idea, because ISP internal
network can not be multihomed to external networks. It's a lot
more straight forward to use plain routers between ISP internal
and external networks. Or, if an LSN box fails, all the connections
will be lost, whereas if one of a router between the internal
and external networks fails, alternative routes will be used
without breaking external connections.
That is,
External --- LSN --- ISP internal --- CPE --- customer private
is a lot less robust (prohibitively less robust, if you are a
serious IPv4 ISP) than
External \
\
\
\
External --- ISP internal --- CPE --- customer private
/
/
/
External /
6rd?
Stateless 6 to 4 should be able to have multiple gateways with
the latter configuration.
Masataka Ohta
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf