[sfc] How to progress our control plane requirements document

["Joel M. Halpern" <[e-mail]>]

Jim and I have been talking with Alia about this, and talking with each 
other.  Along the way we realized that we have some significant concerns 
about what we have done with this document.

We expect to discuss this next week at the interim, looking for ideas. 
And to discuss it further on the list.

The existing control plane document when viewed from the perspective of 
an implementer is hard to decompose into actionable parts. Given that 
the IETF tends to work by solving pieces of problems, rather than an 
entire control solution for all aspects, it is important that 
requriements we define be usable for such piece-wise work. Thus, we need 
to describe the requirements in ways that provide components which may 
be combined into an overall control plane solution for SFC.

While the document provides a set of control plane requirements it does 
not provide guidance on how those requirements can be successfully 
consumed into a standards compliant control plane. In particular, the 
interfaces C1 - C4 are each made up of behaviors likely to be provided 
by different solution components.  Thus, while there is an attempt to 
document a set of interfaces (C1...C4) it is not clear how an 
implementer can provide a solution component that addresses a 
well-defined set of requirements.

Given this, we need to go back and rethink the bundling of functionality 
so as to end up with bundles of requirements that are can be usefully 
addressed by IETF work while avoiding (as we have done well in the 
current version) prescribing the control plane architecture.

In addition, functional objectives are not fine grained enough to 
implement; for example, text such as "Rationalize the management of 
classification rules" is not something one can implement.

Yours,

Jim & Joel