On 9 feb 2011, at 12:40, Tore Anderson wrote: >> IPv6 BGP does in fact carry gloal and link local next hop addresses. > > Yes, but my point is that if you don't configure a global (at least > non-link-local) prefix on the router interconnections, there's no global > address that can be announced as the BGP next-hop. With no global > address, the immediate link-local next-hop cannot be resolved in the > IGP, and the route is discarded. Or am I missing something?
Right. Of course I was speaking about protocols like OSPF, which require that the routers to share a prefix for them to find each other with IPv4, but with IPv6 they don't because the routers communicate using link-locals. I'm not sure if BGP implementations reject link locals as the mail BGP next hop address, or if they are smart enough to know that it can work across one hop but not more than one. It's still possible to make everything work by rewriting the next hop address (such as configuring next-hop-self). One situation where this could be useful is if you want to avoid providing a DoS target by running BGP over unroutable address space. -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
