Hi all, In the last Softwire meeting, there were three multicast drafts were presented. They are:
1. draft-brockners-softwire-mcast-gi-ds-lite-00<http://datatracker.ietf.org/doc/draft-brockners-softwire-mcast-gi-ds-lite/> 2. draft-qin-softwire-dslite-multicast-02<http://datatracker.ietf.org/doc/draft-qin-softwire-dslite-multicast/> 3. draft-xu-softwire-mesh-multicast-00<http://datatracker.ietf.org/doc/draft-xu-softwire-mesh-multicast/> They define how multicast works in different softwire technologies. In the meeting, the chair asked the authors to get together to analyze any way to merge them into a single multicast solution. This is our analysis: 1. GI-DS-lite's GW is normally the next-hop of the MH (e.g. GGSM), so the mcast-gi-ds-lite suggests to create an IPv4-in-IPv6 tunnel to the upstream IPv4 mcast router and the GW will join the IPv4 mcast tree. This solution is simple and doesn't require any new specification. 2. The basic DS-lite defines a solution to deliver unicast IPv4 packets over an IPv6 access network. Multicast is left out-of-scope. The basic DS-lite defines a single AFTR to serve hundreds to thousands B4 elements and the AFTR can be deployed deep inside the network. If mcast was delivered over the IPv4-in-IPv6 tunnels by the AFTR, this would overload the IPv6 access network. The proposed solution is to map the encapsulated the IPv4 mcast group into IPv6 mcast group. This will use the IPv6 mcast distribution tree to deliver the encapsulated IPv4 mcast packets. 3. Mesh-Softwire Multicast shares some similarities of Multicast DS-lite, but it defines a solution of how to use BGP to advertise the next-hop between E-IPs. This is not required for either GI-DS-lite or Multicast DS-lite. In short, the authors believe the drafts should stay separately because they address different problems in very different use cases. Comments and thoughts? Yiu
_______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
