Re: [DNA] Considerations for DNA Schemes with multiple Interface andLayer 2 Technologies

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

Bernard Aboba wrote:
<snip>
> It is also possible for a single radio to be used to construct multiple
> virtual interfaces, each connected to the same network, as long as
> distinct MAC addresses are used on each virtual interface.  Again, in this
> case, DNA would be run on each virtual interface with no interactions
> between them.
> 
> DNA is fundamentally about optimizing the problem of address assignment.
> This is quite distinct from the problem of determining the outgoing or
> incoming interface for packets, which is a routing problem.
</snip>

I agree, assuming I'm understanding the point correctly.  DNA is about
detecting, for a given link, which link that is and obtaining an address
that can be used on that link.  It's another process to decide whether
or not to use that link for, say, a default router.

Say a host is connected both a fixed Etherenet (eth0) and a wireless LAN
(wlan0).  DNA might tell the host that wlan0 has just connected to a new
network, and that it can use address X on that network, but it's active
connection on eth0, which is where it's default router currently is, is
unaffected and so it shouldn't change anything.

I think that the decision to not change default router (or to change to
one on wlan0 if DNA on eth0 indicates that it has just gone down), is
not really part of DNA.  Perhaps it should be?  I'm not sure. Any
thoughts?

Brett.

PS.  I realise that I've diverted from the discussion of
make-before-break interfaces, but if the virtual interface abstraction
can be made then I think this all still applies.  This would be a very
useful abstraction too, because if an interface can be connected to two
links simultaneously, then it difficult to say what DNA for that
interface should pass to the routing decision making component.