Hi Jerry,
On Mar 15, 2007, at 10:55 PM, ASH, GERALD R, ATTLABS wrote:
Hi JL,
IMO an important missing PCECP requirement is identified in
http://www.ietf.org/internet-drafts/draft-ietf-pce-pcep-07.txt
Appendix A (Compliance with the PCECP Requirement Document):
" Here is the list of currently unsatisfied requirements:
o Allow to select/prefer from advertised list of standard
objective
functions/options
o Allow to customize objective function/options
o Support "unsynchronized" & "synchronized" objective functions"
They all deal with the missing 6 objective functions required
in RFC 4657, and are an important omission in PCEP, so far.
Let's try to step back for a minute.
1) PCEP has been designed so as to support the functionalities listed
above.
RFC 4657 does not mandate to define the objective functions but to
allow to
select objective functions, ... and that is clearly feasible with the
PCEP design.
As such I don't see any particular issues.
2) Yes, defining such functions is a good idea, thus the proposed
draft, which is
not at this point a WG ID indeed. Thanks for helping out on the
content of this draft.
3) Whether to add those functions and other features in the base spec
has indeed
been debated during the last two WG meetings and the agreement was to
freeze the current spec in order to stabilize the specification, wait
to see implementations,
interop reports, ... As you know, we have to draw the line somewhere
and if we keep
adding feature, the base protocol spec will not stabilize. The PCE
being a new path
computation concept, it is crucial to first stabilize a set of core
features, acquire some
experience based on deployments rather than keep adding features.
Does that make sense ?
Cheers.
JP.
In montreal we indicated that objective functions would not
be covered in the base PCEP spec.
There was no objection. This was recalled in San Diego, again
no objection...
Well that doesn't cast it in concrete, I'll register my objection.
Without objective functions, PCE would not be able to do much to
optimize path computation. I think it's important for PCEP to include
the objective functions required in the PCECP generic requirements in
the base PCEP spec.
You say that PCEP is 'stable' so you don't want to add them
there?
What I say is that this would create dependency between PCEP
and OF drafts which is not good. The base PCEP draft should
not have any dependency with other drafts.
Does that mean no further changes to PCEP before
WGLC?
Definitely no!
But, again, note that it was agreed that objective functions
would not be covered in the base spec.
See above comments.
I don't think other new drafts (e.g.,
draft-lee-pce-global-concurrent-optimization-02.txt) that
specify perhaps *new* PCEP requirements, should be satisfying
the original RFC 4657 PCECP requirements on PCEP.
Why?
Draft-lee specifies two objective functions listed in 4657,
where is the problem?
Several. No rationale to put generic PCECP requirements in some other
draft besides the base PCEP draft. The draft isn't a WG draft. There
isn't any compelling reason to not specify the objective functions in
the base PCEP draft.
You also
say 'We can add other synchronized objective functions in the
next revision of the gco draft.' I don't think that's the
right place to satisfy the PCECP requirements.
Why?
See above comments.
Thanks,
Regards,
Jerry
_______________________________________________
Pce mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/pce
_______________________________________________
Pce mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/pce
