Re: OSPFv3 fwd'ing address

[Naiming Shen <[e-mail]>]

Alex Zinin said the following on 08/04/2005 10:38 AM:
>>Ok - let's say R3 is in a different routing domain and R1's OSPFv3 does
>>go through some measure (possibly extraordinary) to determine a valid
>>global address for R3's associated interface and assure the address is
>>associated with a global prefix advertised internal to OSPFv3. R2 would
>>also have to determine the prefix is directly connected (a route lookup)
>>and then go to neighbor discovery to map the global address to the
>>link-local address. I can thnk of other solutions but they would require
>>more information advertised with the forwarding address.
> 
> 
> Right. From what I remember, the ND-based option was a bit complicated,
> but if you went in that direction, it would have to be spelled out.
> Another option was to have a link-local LSA that would map global to
> link-local addresses, and then install a GA1/128->LA1 route in the RT,
> which would automatically resolve the lookup for the external.
> 

Assume on the shared media there is no OSPF neighbor with the nexthop,
there will be no link-local LSAs. ND interaction would be the most
reliable way to solve this, and remember link-local address can
also change at anytime(thus the relation needs to be callback
notification or constant polling).

But if we think of this problem in another way, this 'nexthop' is only
an internal state within the router, what does anyone care if
I use a true 'link-local' address as a nexthop, or use the global
address as the nexthop which passed by the OSPF forwarding address?
I would not implement the IPv6 forwarding in a way to only
accept the link-local as nexthops.

thanks.
- Naiming


> Alex