The IP Over NBMA (ION) Archive

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

Tatuya;

> Second, strictly speaking, the point is not that the Ipsilon's
> protocol removes the ATM signaling but that it's a LIS(Logical IP
> Subnet)-base protocol. Even if it supported the ATM signaling, it
> would not be able to provide the QoS you wanted, since the signaling
> had a significance only between two adjacent IP switches.

Ipsilon's CSR uses a point-to-point link as a physical subnet.

> As Peter said, our proposal -Cell Switch Routers- supports the ATM
> signaling(UNI 3.0 and 3.1), but since our protocol is also LIS-base,
> the signaling won't provide the QoS, either.

And, Toshiba's CSR use LIS as a subnet.

But, it's not essential to QoS support.

Both variants of CSRs can support (as an extension) QoS as long as they
use IP based signaling such as RSVP.

Of course, Toshiba's CSR suffers inefficient routing if LISes are
really logical.

But, it is a problem of LIS in general, which causes a lot of problems
EVEN with some sortcutting attempts such as NHRP (see draft-ohta-sun-00.txt).

And with LIS, which actually not logical, everything is fine even
with Toshiba's CSR.

						Masataka Ohta