Re: Optical Link Interface

["Adrian Farrel" <afarrel@movaz.com>]

Just a brief $0.02 (perhaps only $0.015).

From an implementation point of view it seems preferable to extend a protocol
that one might already have implemented thereby re-using code and reaching
stability more quickly.

From a standardization point of view there seems no sense in having two
protocols that can provide the same function.  We should be looking towards ease
of interoperation - why shouldn't the OLI be an interface between components
manufactured by different OEMs?

To say that LMP-WDM as an extension of LMP is of no help because you don't
necessarily have LMP only puts LMP-WDM on the same footing as NTIP, not behind
it.

It seems to me that issues of function level in the two protocols are a
diversion since it is well within the capabilities of all concerned to enhance
either protocol to provide the function required.  So it should be possible to
make LMP-WDM or NTIP satisfy the OLI requirements.

This leaves us with Kireeti's list which I paraphrase as

1. One protocol or two?
1a. IETF should (will!) only allow one of LMP-WDM and NTIP to go forward, as
Kireeti has said.
1b. People implementing LMP will surely prefer LMP-WDM.
2. Peer or master-slave relationship. Of some interest, but it is always
possible to make a peer-based protocol behave in a master-slave way and vice
versa.
3. Underlying transport.  Does a protocol's dependency on certain
characteristics of the underlying transport constrain the implementation and
introduce issues?

My vote is for LMP-WDM (but I am an author).
I sympathize with the inventors of NTIP and urge them to bring what they have
learned to the LMP-WDM draft to make the finished protocol even more successful.

Adrian
--
Adrian Farrel
Movaz Networks Inc.
afarrel@movaz.com