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

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

Sathya Narayanan wrote:

> It is the minimum requirement - similarly the same linkID can
> conceivably be used in more than one link (as long as they are
> further apart), I don't remember global uniqueness of LinkID being
> agreed upon. But, the linkID MUST ATLEAST BE the same within the link
> in one case (link scoped) and its enough if the host picks its own in
> the other (host scoped).

How unique a LinkID has to be is a matter for debate, but I think the
requirements are greater than link scope.  Having the same LinkID on all
routers on link provides the positive acknowledgement that "yes, this is
the same link that I last received an RA from".  This is not really any
better than sending out an RS and getting an RA with a prefix that we
have already seen.

A LinkID needs to be unique within the set of links that a host might
potentially move to (possibly within some limited timeframe) so that a
change in the LinkID received from a previous one gives us the negative
acknowledgement that "no, this is definitely not the same link as one
we last received an RA from" (rather than having to wait for a timer to
expire).

Brett.