Jari Arkko - Le 6/17/08 3:00 PM :
- No host change is "required". If there is none, the scenario that fits is NAT44+APBP in the CPE. - In the APBP-E2E scenario, a dual stack client connects to a v4-only server. (The "change" in the dual stack, i.e. the APBP client support, is the price to get E2E v4 connections from a site hat has no public v4 address.)Does APBP involve host changes?
The work put to SOFTWIRE is a very specific task: cross the v6-only cloud using tunneling, allow optional fairly standard NATting at the other side to handle overlapping RFC1918 space, no changes at all in the subscriber sites or hosts.
- "No change at all in the subsriber sites" seems strange to me.IMU something has at least to be said for tunnels to be established and released, and for v4 packets to be routed to the tunnels. - But I understand that, provided the softwires machinery is already present in the CPE, this can be a small change.
- Any WG is fine. The only concern is that respective scopes of Softwire and Behave aren't, at the end, both so restrictive that a solution like APBP, that uses tunnels and doesn't use NATs in all cases, has no place to be discussed.However, generally speaking other protocol work will go elsewhere, as soon as we know what that work is. The reason why SOFTWIRE is having a head start is that they are attacking a separable, reasonably well understood part of the overall space that fits the profile of an existing WG. We are looking at rechartering BEHAVE for the rest of the work.
Regards. Rémi