Re: Application Statement

[Bruce Cole <[e-mail]>]

> >Imagine an NBMA network with a single CIDR block allocated to it. Ie all
> >stations on the NBMA net have IP addresses within the CIDR block, and all
> >IP stations off the net have IP addresses outside the CIDR block.
> 
> To begin with, this may *or may not* be a realistic assumption. I would
> especially question this assumption if an NBMA network has routers and
> hosts that are under control of different organizations.

Granted.  If you have multiple administrative controls within the same NBMA
network, you wind up with a network more closely resembling the example one
for which routing loops were shown.  My point is that there are simpler
networks, for which NHRP can live in loop freedom.  Which kinds of NBMA
networks are the realistic ones?  My other point was that I don't think we
can conclude whether or not NHRP is useful for more than just address 
resolution until we decide what kinds of NBMA networks are really to be
deployed.

> But even if there is such an environment, what are the mechanism(s) to
> guarantee that this address assignment will be preserved -- that is
> no host or router off the NBMA network wouldn't get an IP address out
> of the CIDR block ?

Enforced by routing.  As long as the station has a misassigned IP address,
it is unreachable, as routing directs IP traffic for the CIDR block to the
NBMA network.

> >Extending NHRP to be aware of topology changes, or perhaps even to generate
> >changes of its own would seem to be a possible solution for a class
> >of more complex networks.
> 
> Would you explain how this would help to solve the problem that Joel
> described in his message posted to this list in February ?

In the above, I was trying to describe a network more complex than the single
CIDR block example, but less complicated than the multi-administrative control
network from Joel's example.  So I was not trying to claim that NHRP could
be aware of the topology changes in Joel's example.  We already know that
is not feasible.

Now, in Joel's example, one could limit the NHRP speaking routers to only use
NHRP to resolve IP destinations within their respective administrative control.
Either through ACLs, or by building multiple logical NBMA networks on top
of the physical NBMA network.  I believe this would solve (or at least help
solve) the example looping problem.