The Routing Over Large Clouds Mailing List Archive by date

[jhalpern@us.Newbridge.com (Joel Halpern)]

One of the useful properties of routing protocols like PNNI (which is
what many ATM deployment both public and private are likely to use) is
that while the topology is advertised heirarchically (as is necessary for
scale) the VC path does not actually follow the heirarchy.  Part of the
reason this is possible is that one can put a lot of information in 
the call setup, without burdening the data, wince once the VC is setup
the path is fully established.

Also, there is no reason to expect Routers at all topologically significant
spots.  If there are not routers (CSR or otherwise) at all such locations,
the routing path will having significant excursions compared with the
ATM path.

All of this assumes rational pricing based on resource usage, and somewhat
rational policies.  It is also dependent on certain models of relative
cost and bandwidth management efficiencies of ATM Switches (including CSR)
as compared with large conventional routers.  Many of these assumptions
are subject to change.  But refraining from standardizing something because
it might become overtaken by events is a very bad reason.  We should do
the work and develop the proposed standard.  If no-one uses it, then it 
will die on the standards track.  There is no shame in that.  Let us not
prejudge the market on this matter.

Thank you,
Joel M. Halpern				jhalpern@newbridge.com

FUJIKAWA, Kenji wrote:
"
In my opinion,
if ATM switches are connected to switches that is physically
near to themselves,
NHRP's shortcuts may be effective.

However, in an environment where ATM switches are connected 
hierarchically and routers are located properly,
paths created by NHRP is almost identical to those by 
classical IP routing except via routers.
If you want to use virtual LANs,
you can use IP mobility.
"