RE: Outsourcing and Provisioning common model

["Moisand, Jerome" <jmoisand@unispherenetworks.com>]

Yes, agreed, Shai speech sounded right to the point to me, and it would be
real nice to hear the COPS experts discuss this "common" model.

1. I'm afraid that COPS-PR will always stay controversial if it's viewed as
a pure provisioning play. (please don't flame me! have mercy! ;-)

2. The outsourcing model tends to generate quite a load of traffic by having
to manage each handle (and related policy objects) fully independently.
Would be nice to be able to *reference* (and often share) pre-defined
profiles.

3. It always made me uncomfortable to see outsourcing discussions being
centered on RSVP sessions. There are many other types of sessions where
outsourced dynamic decision (at session establishment, during its "life", or
at deletion) could be very useful. Think RSVP but also PPP, EAP, DHCP, IGMP,
IPSec associations, all sorts of tunnels, LSPs, etc. Also possibly routing
protocols adjacencies, etc. Anything which is signaling-oriented. As well as
various objects configured in a more static way in nature (e.g. SNMP or the
dreaded CLI ;-), but still requiring dynamic policing in a
transaction-oriented way. 

Now when you think to a merged model, where various things intended to be
shared can be provisioned in advance or 'just in time' via COPS-PR via a
"coarse-grain" handle, and referenced by decisions made on more dynamic and
fine-grain objects (handles in the outsourced model), then the whole picture
is becoming WAY more attractive. At least for me...

I suspect that technically speaking, we are actually close to such 'merged'
behavior, notably with the new framework PIB, simply it seems to never be
discussed or illustrated in tutorial/educational/examples material. 

So putting all these thoughts together, yes, would be nice to see this
thread of discussion more active... and possibly captured in the updated
policy management framework/umbrella document that the WG recently
initiated.

Jerome


-----Original Message-----
From: Yacine El Mghazli [mailto:yacine.elmghazli@ms.alcatel.fr]
Sent: Thursday, January 31, 2002 9:38 AM
To: rap@ops.ietf.org
Subject: Outsourcing and Provisioning common model


Hi everybody,

At the last IETF meeting, Shai Herzog launched a discussion about a COPS
common model merging both outsourcing and provisioning modes. The fact was
that some recent drafts use COPS-PR to OUTSOURCE information towards the PDP
(umts, access-bind, sls...) and the goal was to identify where the 2 modes
meets and to agree on a common model.

This discussion was supposed to take place on the mailing list, but...
doesnt come !

Speaking as a co-author of the cops-sls draft, I'm wondering about this
common model and, after all, about the future of such practises.

thanks,
yacine

-- Yacine El Mghazli
----------------------------------------------------------------------
  Alcatel R&I
  Next Generation Router & Cross-Connect - NGRX
  Marcoussis, France
  Tel  +33 1 6963 4187
  Fax  +33 1 6963 1169
  yacine.el_mghazli@alcatel.fr
----------------------------------------------------------------------




----- Original Message -----
From: "Da Silva, Pedro" <pdasilva@orchestream.com>
To: <rap@ops.ietf.org>
Sent: Thursday, January 31, 2002 12:08 PM
Subject: RE: A question about SPPI


> Man, Ravi,
>
> The uniqueness clause also states that "The specified set of attributes
> provide a necessary and sufficient set of values by which to identify an
> instance of this PRC".
> As such does an optional attribute constitute a necessary one for
UNIQUENESS
> clauses?
> Also, considering the case where a UNIQUENESS clause's set is composed of
> optional attributes only, does that constitute a sufficient set by which
to
> identify an instance of the PRC?
>
> cheers,
> Pedro
>
>
>
> -----Original Message-----
> From: Sahita, Ravi [mailto:ravi.sahita@intel.com]
> Sent: Wednesday, January 30, 2002 11:50 PM
> To: 'Man.M.Li@nokia.com'; rap@ops.ietf.org
> Subject: RE: A question about SPPI
>
>
> Man Li,
>
> The UNIQUENESS clause itself is optional. The SPPI does
> not put any constraints on which attributes can be specified
> in this clause other than -
> 1.  The attribute named in the PIB-INDEX clause may not be
>     present in the UNIQUENESS clause.
> 2. An attribute may not appear more than once in a UNIQUENESS
>    clause.
>
> So, yes, the UNIQUNESS clause can contain optional attributes
> as long as it follows the rules above.
>
> Hope that helps,
> Ravi
>
> -----Original Message-----
> From: Man.M.Li@nokia.com [mailto:Man.M.Li@nokia.com]
> Sent: Monday, January 28, 2002 11:08 AM
> To: rap@ops.ietf.org
> Subject: A question about SPPI
>
>
> In a PIB table, some attributes can be optional. Should the UNIQUENESS
> clause exclude optional attributes? In other words, can a UNIQUENESS
clause
> contain any optional attributes?
>
> Man Li
>