Ted,

>> the OSPF based prefix assignment handle all of these, "out of the starting 
>> blocks".
> 
> I'm under the impression that a number of issues you mentioned as solved by 
> OSPF and not solved by PD are actually not solved by OSPF.   To respond 
> individually:

which ones?

>> - in an arbitrary topology how do you decide which interfaces you are a 
>> client on and which interfaces you relay on
> 
> If you got a /64, you are not a delegating router for that prefix.

I can't grok your reply. how does a CPE determine which interface to relay on 
and which one to be a client on?
or does it do both?

>> - how do you handle the case where multiple routers try to assign a prefix 
>> to a link
> 
> Take them all.

it is a wanted property of the mechanism to have only one prefix per link.

>> - how do you discover the DHCPv6 servers
> 
> The same way you always do—by multicasting a solicit.

there are multiple DHCPv6 servers, aka sources of information in the network.
how does a CPE build the tree of relays to get to all the servers?
or are you suggesting that multicast routing is required in the home network?

>> - how do you handle the case with multiple sources of information 
>> (multi-homing)
> 
> As above, for each delegated prefix, either act as a delegating router for it 
> or not, depending on whether it is a /64 or larger.

can't grok this.
to ask this another way. how does DHCPv6 work with multiple servers having 
different information?
that's what a flooding solution solves elegantly, that a discover - 
request/reply protocol doesn't.

cheers,
Ole
_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet

Reply via email to