The IP Over NBMA (ION) Archive

[Albert Manfredi <manfredi@arl.bna.boeing.com>]

Joel Halpern wrote:

[ ... ]

> > If one specifies only anycast or static configuration, for
> > registration, is one not guaranteed that the NHS will be in the same
> > LIS as the NHC?
>
> NO.  For a simple example if one had a private corporate ATM network, one
> might well use a single anycast address for all NHS.  Hence when one
> connected using anycast, one might get any NHS, not necessarily one that
> is serving you home LAG.
> If we require one anycast per LAG, we might as well not bother.

Okay, point taken, but that wasn't the crucial part. If you have a
corporate network with very few egress points, then it's clear that
having an NHS per LIS is not important. The ultimate goal of that
NHS to LIS ratio is, I think, to ensure that the ATM shortcut be
established based on a sensible route. Since ATM knows nothing about
the L3 topology, having the IP frame take the "correct" IP route
before establishing the best ATM shortcut makes a lot of sense.

In a large corporate ATM net with few egress points, each LIS would
not need an NHS to ensure an optimum (or just plain workable)
shortcut. The large corporate net is not a crucial part of the
Internet's infrastructure.

Still, the fundamental point I think remains: static configuration
and anycast alone ought to be specified for NHRP registration. I
don't disagree with you that the behavior of multicast shouldn't be
investigated. I merely question its relevance.

Bert
manfredi@arl.bna.boeing.com