Re: [DNA] Issue about multiply connected to the same linkandDNAbe ingper-interface or not (was Re: WG LCcommentsondraft-ietf-dna-goals-01. t x t)

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

Brijesh Kumar wrote:

> In the above scenario, you are right.
> 
> But, I am not sure when a device moves if the inter-relationship 
> between two interfaces continues based on past knowledge of prefixes. 
> May be the device movement caused the first interface to loose the 
> connectivity, and the second interface to attach to a new attachment 
> point (no RA received yet to determine any thing about the network).

In general you need to receive at least RA to verify that the interface 
is connected to a particular link.

There are exceptions when things can be determined at L2.
The "non movement" L2 exceptions are things like PPP multilink and IEEE 
802.3ad link aggregation where L2 knows that two or more L2 interfaces 
are "bonded" together to form one. Typically this is completely hidden 
from the IP software.

But in the case of DNA there might be interesting hints from L2
that do not require L3 verification. For instance, if the L2 detects 
that it has changed 802.11 APs, and that the MAC address of the AP is 
the same as the AP I used 5 minutes ago, then it would probably be 
reasonable to use whatever L3 configuration I had 5 minutes ago (asuming 
I retained it) (of course, subject to the prefix and default router 
lifetimes for that retained information).

    Erik