Re: [DNA] Model: how to treat "link down" events

[Erik Nordmark <erik.nordmark@sun.com>]

Mohan.Parthasarathy@nokia.com wrote:
>  
> Erik,
> 
> Is the following scenario possible ?
> 
> 1) Host receives a link down event. Does not do anything.
> 2) Host attaches to the new link. Link UP event happens
>    slightly late or the event has not propagated completely
>    through the system. Meanwhile you are able to send and
>    receive packets. This is not good because you are not
>    in the optimistic state yet and you are on the new link.

We can't come up with any reliable DNA procedure in such a case. One 
would more or less end up ignoring the 'link up' event and just rely on 
the content of the RAs as suggested in 5 in CPL; but as stated that 
approach isn't robust; if a host moves too often it can conclude that 
the union of prefixes from multiple links are in fact the prefixes 
assigned to a single link.

And the two authors of CPL and Thomas agree that we can drop that section.

So for a reliably DNA we do have to assume that the link up event 
notifications are delivered in order relative to packets arriving from 
the network interface.

Note that draft-pentland-dna-protocol* can do slightly (as in epsilon) 
better than CPL, since the DNA procedure can tie the received RA to the 
RS one way or another. But this doesn't help tell whether the prefixes 
received in periodic, unsolicited RAs belong to one link or another when 
there is no 'link up' event notification.

    Erik