Re: [DNA-BOF] Using L2 to provide Instantaneous Movement Detection andNeighborhood Discovery

[Alper Yegin <alper@docomolabs-usa.com>]

>> The most important thing to take away from all this is that L2 "hints" are
>> generally fallible and that an implementation MUST be robust in the face
>> of misleading hints.
> 
> I agree. DNA scheme would be better not to rely too much on L2 hints.
> As far as I know, MD/DAN can't be done with only L2 information.
> It's just a hint not a confirmation.
> 
> But it's very useful for a node to receive  a L2 hint, for example 'Link up'
> hint, a notification that a new link has been set up.

Here is one way to look at this...

If we define it correctly, an indication from the link-layer that it
established a connection with an access network should be a "trigger". This
should come along with additional "hints" that will be used by the DNA
scheme. A trigger may or may not carry any hints. Depending on the type of
hint (weak, strong, definitive) DNA scheme will take different actions.

For example, when a PPP link comes "up" in cdma2000, an IP address is
already obtained in Simple IP service type. This is rather definitive, not
much for the DNA to do here. 802.11 examples are lacking this level of
strength in their hints.

I think we could say a trigger is used to engage the DNA scheme, and the
associated hints are used within DNA. Hints help enhance and optimize the
DNA scheme. Availability of hints depend on the link-layer technology and
sometimes deployment, and the effect of the hint on DNA depends on the
configuration mechanisms. E.g., DHCP-based networks would take better
advantage of such hints due to their stateful nature, whereas the usefulness
of hints for stateless address autoconfiguration is (much?) less.

Alper