Hi Hui,
Adding to what Alain already said. When you projected the slide that compared various transition approaches, for the UE to UE bullet point, for the Gateway Initiated Dual-stack lite solution, you have marked it as "Unsupported or No". So, I asked the question, is it for over-lapping or a non-overlapping case and from our conversations offline and online both at Shanghai and at the IETF, you said it was about non-overlapping case. Now, for this case, in the proposal we have, any IPv4 traffic from the UE, always hits the GGSN over the mobility tunnel. At the GGSN, its a routing decision to forward the traffic to the internet via CGN, over the CGN tunnel, or local foward. If its a local forward, the packet is routed to the target UE/SGSN over a different mobility tunnel. The GGSN is the anchor and all traffic goes through that point and it can local forward. That's how we can support this scenario. For the case of overlapping IP, we talked to many operators and no one confirmed that they have a legacy application that are supported in this mode. So, the general thought was to have a sensible approach of using IPv6 transport for such applications. Now, if we argue that we need to support over IPv4 transport, we need to ask many questions, as how the application resolution happens, how the application server delivers that information and how it learns the NAT'd address of a UE ..etc. These are unresolved questions both as a requirement and even for PNAT if you pitch a solution based on host translation. As I see it, its a new requirement and as Alain/Dan Wing pointed out, there are potential ways to solve it. In simple terms, what ever you do on the host, we can infact solve it on the network. Now, moving to your other pointer and the motivation for the PNAT work, you have listed: a.) a need to support IPv6-only transport network b.) support IPv4 legacy applications, over a IPv6 transport c.) avoid airlink over-ahead d.) enable UE to UE comminication. For #a and #b, mobility architectures and the solution we have allow the case where the access network and the core network is either IPv4 or IPv6. There is no issue on this aspect. For #c, our solution does not introduce any air link over-ahead, as there is no tunnel from the UE. We dont have the host architecture, or touch the UE. For #d, the solution can support UE to UE over IPv4 or IPv6 transport, for the non-overlapping case. For overlapping case, we will plug in the feature as needed. Also, I'd like to remind that your slide missed number of other bullet points, for example, it should have had a bullet point, "Changes to host architecture", with a marking "Yes" to PNAT solution, and "No" to Gateway Initiated Dual-stack lite solution. Since, its not a trivial design point that you can ignore. Regards Sri On Mon, 23 Nov 2009, Hui Deng wrote:
Dear Alain and Sri, I guess that you two have already read below draft: http://tools.ietf.org/id/draft-cao-behave-hbt-req-00.txt This draft has been presented in detail during the last IETF meeting. During the meeting, you two have standed up and said that DS-Lite and GW-Init-DS-Lite could also meet the requirement of host to host direct communication. I really cann't understand it, could you help to elbaroate more here about details? Many thanks for your kind explanation. Best regards, -Hui Deng
_______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
