Re: [DNA] Confirming today's face-to-face meeting decisions

[Brett Pentland <brett.pentland@eng.monash.edu.au>]

Erik Nordmark wrote:
> Brett Pentland wrote:
> 
>> There may be extra cost in making all the routers on a link agree on
>> a "link scoped identifier", but that is an out of band management cost.
>> Using "does this landmark still exist" has a cost during movement in
>> that there is no negative acknowledgement and we have to wait for a
>> timer to expire if the link is new and the landmark does not still
>> exist.
> 
> 
> While the timeout is the case in the current rrd draft, it isn't a 
> fundamental property of landmark-based schemes.
> 
> For instance, in a landmark scheme the routers on the link can share the 
> landmarks they have so that any router can respond. If you do this, a RS 
> containing a landmark can be responded by any router, whether on the 
> same link or not. If on the same link the router would respond "yes, 
> this landmark is on this link" and if its it a different link the router 
> would respond "no".
> 
>    Erik

Ok, fair enough.  It does require a little work from the routers to make
sure that they know that landmarks they have seen on the link are still
current, but that's certainly doable.

Brett.