Hi Randy, On 04.07.2011 00:01, Randy Bush wrote: > IPv6 subnet anycast is not used operationally, complicates > implementations, and complicates protocol specifications. The form > of anycast actually used in the Internet is routing-based, and is > essentilly the same as that of IPv4 anycast. Therefore, this > document deprecates IPv6 subnet anycast.
I was more referring to host anycast, I don't care about the special anycast addresses of routers as defined in sec. 2.6.1. of RFC 4291 http://tools.ietf.org/html/rfc4291#section-2.6.1. or also RFC2526. There are definitely cases where it can be useful to have a group of anycast hosts inside a subnet, e.g., in order to increase robustness of services etc. For instance, we had a nice use case for bootstrapping into P2P networks by using subnet anycast addresses. There is no need to inject anycast routes into the routing system for this. I don't see the point why the currently proposed mechanisms (as I quoted in my earlier mail) complicates protocol specifications. All that a host must do is to join the solicited node multicast address for the anycast address, and then to omit DAD. Setting the overwrite flag to zero for an NA is also not such a big thing. I don't see a need to implement any special functionality in routers and I don't see how anycast complicates the host stacks as this boils down to: support a local anycast flag for addresses; if flag is set then omit DAD; when sending a neighbor advertisement: the overwrite flag to zero if flag is set. Seems to be simple enough... Regards, Roland -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
