On Nov 9, 2009, Scott Brim wrote:

>>> An argument has been made, and I don't intend to endorse it, that
>>> stateless NAT66 would be a fine solution to the problems of
>>> multihoming, BGP scaling, and renumbering hassles, all in
>>> one simple wrapper.
>>
>> This argument requires PI space used for outer addresses for session
>> survivability for provider failure events.  And in that case, well,  
>> you
>> have the entire route-scaling problem in a nutshell.
>
> That assumes no changes to hosts and no pinning of routing.


Right, fail-over support in hosts would enable session continuity.
Thus, NAT would be used in combination with Shim6, Six/One, the
name-oriented stack, or something similar.  (Don't know what you mean by
"pinning of routing", though.)

An attractive property of this combination of solutions, deployment-
wise, is the alignment of benefits and costs:  The network operator pays
for provider-independent addressing, by setting up a NAT.  The host pays
for improved session robustness, by upgrading to fail-over support.  And
importantly, there is no dependency between the new network support and
the new host support.  Either one would work without the other.  (Though
fail-overs still depend on support at the remote peer.)

Without intending to sales-pitch this particular combination of
solutions, I think its alignment of benefits and costs is key.

- Christian


_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to