RE: Optical Link Interface

[Bala Rajagopalan <BRaja@tellium.com>]

Kireeti:

Thanks for distilling the issues into the 3 bullet
items. I would also like to add my two cents on these
points as an OXC vendor:


> a) LMP-WDM as an extension of LMP vs. NTIP as a new protocol.
> 
  I think both choices have pitfalls.

  LMP-WDM: First, LMP and LMP-WDM serve two different purposes. One
may exist without the other. Indeed, for a WDM vendor, LMP-WDM
will be a new protocol to implement. If they balk, I can't unilaterally
choose LMP-WDM on the OXC side. Also, my experience so far
has been that neighbor discovery based on LMP is not simple
to realize in transparent switches, the main candidate for OLI.
So, the question of LMP-WDM as an extension of LMP (as it
currently stands) may be moot. 

  NTIP: The present definition does not provide for neighbor
  discovery, a function required under OLI. Is another protocol
  required for this? If so, the value of NTIP is diluted.

So, I think a case can be made for NTIP only if the neighbor
disc. function is taken out of OLI. Similarly, a rather strong
case for LMP-WDM can be made if the LMP neighbor disc. functionality
could be simplified. I would personally prefer the second
option since it provides a complete package
(if there is any all-optical vendor out there who
has got LMP neighbor discovery running, send me email :-) ).


> 
> b) Two protocols vs. one.
> 
    One. But one that is useful in practice.
> 
> c) The model: master-slave vs. peer.
> 
 Both LMP-WDM and NTIP can operate in the request-response mode.
I don't see this as an issue.

> 
> d) The transport protocol: TCP or raw over IP.
> 
  No big deal either way.