RE: spc connections

["Brungard, Deborah A, ALABS" <dbrungard@att.com>]

Lyndon,

It's been a long thread and I am not sure anymore what we are discussing:

- use of the ERO for SPC? Can you be more specific on your concerns as this is where OIF/ITU chose for a different implementation. We need to understand if it resulted from an unclear description or technical concern (as both Yakov and Kireeti requested).
- use of G7713.2 (3474) as a solution? This is also a needed discussion item to progress the ASON signaling work.

On the use of G7713.2: we all agree it provides a *solution*. The question is - is a G7713.2 network overlay (not to be confused with the GMPLS overlay model) the only solution needed? Or do we need a 3473 solution?

Your 3473/3474 interworking draft and the new appendix added to the draft on GMPLS signalling for ASON both describe the 3473/3474 differences. The key driver for defining new 3474 objects was the implementation decision to support multiple addresses (IPv4, IPv6, NSAP) in the protocol objects (vs. address mapping at ingress/egress (e.g. G7713.1 ASON PNNI's implementation)). Multiple family address resolution/management and support of the new objects are part of 3474 node processing (regardless if one is tunneling a new transparent object as an overlay/terminating or if one is using IP-only addressing). 

Operators are well familiar with the tradeoffs of using interworking functions, address mapping, and the operations to support multiple address families. And the use of interworking/overlays vs. a GMPLS backward compatible solution have been viewed as two different solutions to very different applications (for both operators and vendors depending on their network evolution/product support). Are you saying (below) that you view G7713.2 overlays as *one solution* or you view it as the *only one* solution needed?

Deborah

-----Original Message-----
From: Kireeti Kompella [mailto:kireeti@juniper.net]
Sent: Monday, November 17, 2003 9:48 PM
To: Ong, Lyndon
Cc: ccamp@ops.ietf.org
Subject: RE: spc connections 


On Mon, 17 Nov 2003, Ong, Lyndon wrote:

> I respectfully disagree with the suggestion, as
> 7713.2/3474 provides a valid solution that does not
> require changes to the text to begin with.

Here's how I look at it: explicit label control as described in 3473
appears not to be clearly understood.  Therefore, a clearer
explanation is needed in any case.  Given that, the choice is between:

	3473+explanation: works without a new object, has label
		control per hop, is trivially backward compatible
and	7713.2: needs a new (to 3473) object, has only egress label
		control, needs interworking spec with interoperate
		with 3473 (assuming interworking issues are fixed)

Kireeti.
-------