Hi, I've been thinking about how IPv6 might be deployed in a DSL network with Ethernet rather than ATM backhaul. The IPv6 traffic rides directly over Ethernet - the so-called "IPoE" DSL model - rather than being encapsulated in PPPoE or PPPoA.
In this scenario, customers would be allocated a /48 (or longer) prefixes for their network, sitting behind the DSL CPE operating as a router, via DHCP Prefix Delegation. At the local exchange/central office, the bridged-ethernet-over-DSL traffic is aggregated from the DSLAMs using a layer 2 switch. This layer 2 switch aggregated traffic is then backhauled, also via ethernet, to a central aggregation point where the aggregation router resides, which provides access to the rest of the IPv6 Internet. Ideally, traffic between the prefixes goes between the CPE directly via the layer 2 switch in the exchange/co, rather than via the upstream aggregation router. However, as the CPE only has a default route to the upstream aggregation router this isn't going to happen. In the "convetional" routing scenario, the CPE router and the aggregation router would be participating in a routing protocol, and therefore the CPE would know about the prefixes residing behind the other peer CPE. In this DSL scenario this isn't scalable or manageble. I'm curious on what people's thoughts would be regarding having a DHCP option that enables the CPE router process ICMP redirects for these prefixes from the upstream aggregation router? This would encourage most inter-cpe prefix traffic to stay local to the exchange/c.o., without having to run a fully blown routing protocol on each of them. With increasing P2P style applications and therefore traffic, I think there would be real benefits in permitting ICMP redirects be processed by the CPE routers in this scenario. Regards, Mark. -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
