HI Satoru, I am saying that hub & spoke model is quite simple in case of MAP-E. In the draft, draft-ietf-softwire-map-02, as you have done it in draft-mdt-softwire-map-encapsulation-00, specify hub&spoke case first.
I know that mesh model is important for MAP-E as BR may be deployed deep in the network. You are explaining in detail in Section 5. I suggest adding one clarifying statement saying that it applies to the mesh model. As for the details of encapsulation in reference to draft-mdt-softwire-map-encapsulation-00, I suggest that you either make draft-mdt-softwire-map-encapsulation-00 a normative reference and make it up to date, or add some details to draft-ietf-softwire-map-02 on how the encapsulation in MAP-E works in reference to RFC 2473. As for the source address of IPv6 packets sent by BR, I suggest removing the requirement to make it an anycast address. I read in a vendor web site (Cisco) it is stated that anycast addresses must not be used as source addresses. RFC 4291 does not require this though. The reason? As I explained, at times MAP-E is stateful, how would CE find the right BR if it does not know BR's unicast address? I hope the above is clear. I am removing the conversation we had before, but I am curious, what is MRT? I could not find it anywhere? > It looks that there're two cases for hub-and-spoke mode. One is that a CE > forwards packets with MRT as same as mesh mode, but it has only DMR. The > other is that a CE forwards all packets with following DMR even it has > several FMRs. Do we need to distinguish each of them? then write up both > cases for hub-and-spoke? > Regards, Behcet _______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
