On Mar 14, 2012, at 1:07 PM, Peng Wu wrote: > Hi Alain, > > It's a little confusing now. Let me try to get things clear. > > So the sd-nat-02 is not quite similar to the earlier version, the > mechanism somehow changes.
One drawback of sdnat-01 was that the AFTR would still have to do ALG work, so it was not purely stateless. So, instead of asking each CPE to use ports 1024 to 1024+max and then slide this range into the correct one in the AFTR, sdnat-02 tells each CPE which actual port range min..max to use through the newly defined ICMP message. The net result is that the AFTR does not need to perform any ALG work anymore, making it truly stateless > In my understanding, now the principle of the mechanism is similar to > the lightweight 4over6 draft, but I may miss something here. The general idea is the same, however loghtweight 4over6 does not specify how to achieve it, it leaves the technique to assign address and port open. stateless DS-lite (aka sdnat-02) does specify exactly how you achieve this. > My question is, how is it stateless or deterministic, and how is IPv6 > anycast for multiple AFTRs working? See above, the AFTR is now entirely stateless. The port range is fixed, thus deterministic and no logs are necessary. > Seems the draft is suggesting that we put an identical profile on each > AFTR, the content of which is the mapping between IPv6 address and > IPv4 address+port range for all the CPEs (And of course, we can try to > find some protocol to synchronize the mapping automatically). Did I > get this correctly? Yes, this is correct. For a group of user sharing the same AFTR set, each AFTR in the set is configured with the same anycast IPv6 address and the same IPv4 address pool and per-subscriber mapping table. That IPv4 pool is announced as an anycast IPv4 route upstream by all AFTRs in the set. _______________________________________________ Softwires mailing list Softwires@ietf.org https://www.ietf.org/mailman/listinfo/softwires
