> > Consider a scenario where dual-stack hosts coexist with IPv6 only hosts > > (IPv6 only stack or dual-stack with only IPv6 address), should they be > > configured with different DNS servers? > yes, that's how they should be configured. Please note that this does > not require two dns servers, you can run them in the same box as two > separate process, and use two different ip addresses. That would be the > best configuration (especially cause it would work with legacy nodes, > either dual stack or v6 only). (please note that other mechanisms > recognizing some bits, either the prefix or some bits in the iid would > require updating the node, and that is why this method is superior)
Can the DNS server be automatically configured according to whether the client is dual-stack or not? If not, I don't think this is the best choice. > > That is to say, should dual-stack > > hosts be configured with a normal DNS server, while IPv6 only hosts > > configured with a DNS64 capable server? > > > > Or do you believe the above scenario itself is rare? > > > > > >> Second, you don't solve this problem by adding some bits in the middle > >> for recognizing the synthetic addresses. The point is that i am making > >> is that having them in the rpefix or in the iid achieve the same > >> capability (and have the same limitations) > >> > > > > There seems no bit in a NSP to tell whether an IPv6 address is a synthetic > > address or not in the current address-format draft. > > the NSP itself is what needs to be learned by the hosts that want to > recognize a synthetic address! > > I mean, in particular with the rfc3484 poilicy table, you need to > configure the NSP in the policy table, so that native connectivity is > preferred. If I understood it correctly, the basic precondition of this approach is the NSP is known to the hosts in IPv6 realm. Hence, I wonder whether this approach is still suitable for the IPv6 Internet to an IPv4 network scenario where NSPs used by different IPv4 sites are much various. > the main difference between this iand having a fix set of bits is that > these need to be learnt by the hosts in order to recognize the NSP. Note > that there is ongoing work to dynamically configure the rfc3484 policy > table, which would deal with this issue. Another option is to use the > WKP which can be hardcoded. > > On the contrary, we > > could use some bits in the IID to achieve that goal. > > > > Another possible way is as follows: the NSP used for synthesizing IPv6 > > address should be a specific prefix, e.g. starting with the binary value > > 00010101. > we have that, and that is the WKP There are three choice for the WKP in the address-format draft: o reuse the IPv4-mapped prefix, ::FFFF:0:0/96, as specified in RFC 2765 Section 2.1; o request IANA to allocate a /32 prefix, o or request allocation of a new /96 prefix. Could you tell me which one is equivalent to the specific NSP I mentioned, which starts with a special binary value? Best wishes, Xiaohu > > In this way, the constraint on IID format mentioned before is > > removed due to starting with the binary value 000. In addition, this address > > is also provider aggregatable because the remaining part between the above > > binary value and the last four octets (i.e., IPv4 address to be embedded) > is > > network specific. > > > > Comments? > > > > Best wishes, > > Xiaohu > > > > > >>> > >>>>>> Regards, marcelo > >>>>>> > >>>>>> > >>>>>> > >>>>>> > >>>>>> > >>>>>>> -- Christian Huitema > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> _______________________________________________ > >>>>>>> Behave mailing list > >>>>>>> beh...@ietf.org > >>>>>>> https://www.ietf.org/mailman/listinfo/behave > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>> _______________________________________________ > >>>>>> Behave mailing list > >>>>>> beh...@ietf.org > >>>>>> https://www.ietf.org/mailman/listinfo/behave > >>>>>> > >>>>>> > >>>>>> > >>>>> > >>> > >>> > > > > > > > > -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------