Re: [DNA-BOF] Using L2 to provide Instantaneous MD & ND

[Bernard Aboba <aboba@internaut.com>]

> Personally, I think Optimistic DAD is good enough.
> If this is the case, there's no need to expend so
> much effort on L2 solutions.

I agree. In the case where the host remains on the same subnet,
reachability can be confirmed in a single roundtrip to the gateway.  As
we've discussed, this typically cannot be avoided anyway, but in many
cases it is not costly -- latencies on the order of 1 ms are achievable on
802.11 networks.

The expensive case is where the host has changed subnets but does not
know this, fails the reachability test, and then has to obtain an
address, potentially on a network on which it already has a valid IP
address.  Latency arises from having to retry the reachability test
in case of packet loss, as well as from having to obtain an address
unnecessarily.  The latter cost is substantial for stateful address
configuration where the host may have to communicate with a DHCP server.

If "hints" can help avoid this case then the contribution is substantial.