The IP Over NBMA (ION) Archive

[Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>]

Juha;

> - does the scheme assume that all switches also to be able to route ip
>   packets, i.e., is it not enough that switches participate in an ip
>   routing protocol?
>   yes: - no: +

Can you distinguish a bridge from a router?

Just as Etherenet have bridges and routers, ATM has bridging switches
and routing switches, the CSR.

Let bridging switches speak Q.2931.

Don't try to let bridging switches speak IP nor participate in
IP routing.

It's routing switches, CSRs, which speaks IP for IP routing.

As you can see with Toshiba's implementation, CSRs can communicate
over bridging switches with Q.2931 signalling.

> - does the scheme work as such with existing ip multicast protocols in
>   the atm environment?
>   yes: + no: -

As multicast support, even best effort ones on packet routers, needs
certain amount storage to maintain state, it is perfectly OK to
establish a dedicated VC for routing switches for each multicast
address.

The computational complexity remains same.

Or, best effort multicast is a wrong thing to do and will be forbidden
in the near future, becasue it consumes dedicated storage on routers
without resource reservation and associated payment.

							Masataka Ohta