> I agree with Erik. Good :-)
> I see two alternatives: > > 1. ND proxy. Limited to single router, proxy from uplink to > downlink. No need for loop detection. > > 2. Multilink subnet routing. Handles arbitrary topologies. Must > handle loops. Agreed. But furthermore I actually don't see the utility of #1. First of all, based on RFC 3177 an ISP needs to be capable of handing of /48 prefixes to customers. The issue here is when the ISP doesn't do that. The unstated assumption for #1 seems to be that an ISP (that uses a form service differentiation I disagree with but that is beside the point) can not assign a single IPv6 address to one of their customers but is somehow forced to at least hand out a /64 because it needs to assign a /64 to the link between the ISP and the customer. I think this assumption is false. An router does not need to advertise an on-link prefix for the link between it and the client. The client machine does need to be able to allocate an IP address which is easy without an on-link prefix when using DHCPv6, and a bit more subtle using stateless address autoconfiguration. (The issue is how the host autoconfiguring an address effects the routing on the ISP side.) Thus I think the ISP that wants to charge differently for one IPv6 address per customer compared to a /64 or /48 prefix can do this as easily as it can in IPv4. This implies that ISPs that want to hand out /64 or /48 prefixes should use an explicit prefix delegation mechanism which explicitly delegates the prefix to the customer and not implicitly does this by assigning a /64 to the link between the ISP and the customer. Hence ndproxy at the router the customer has on the link to the ISP isn't necessary. Erik -------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------