RE: Question about COPS-PR model

["Rawlins, Diana" <Diana.Rawlins@wcom.com>]

Riccardo,

Please see below for my thoughts on the PPP (or whatever dial up)
service.

-Diana

-----Original Message-----
From: Riccardo Scandariato [mailto:riccardo.scandariato@polito.it]
Sent: Thursday, March 01, 2001 7:27 AM
To: rap@ops.ietf.org
Subject: Question about COPS-PR model


If I well understood COPS and COPS-PR, the first bases on a "bottom-up"
model where events arrive from the network (e.g. a RSVP req.), while the
second bases on a "top-down" model where events arrive from a control
system (e.g. a user interacting with a Web interface, or net-admin
operations).

The question is:
how can I arrange a scenario where events arrive from both the control
system and the network (e.g. a user dialing in an access router)?

FOR EXAMPLE, let us think to an hypothetical "Security Client-Type". Let
us suppose a router that must forward incoming traffic to many outgoing
tunnels. The router is placed at the border of a network cloud and
traffic arrives from statically attached sites. In my opinion, this is a
typical COPS-PR scenario:
- first I assign different labels (i.e. Roles) to incoming interfaces
(i.e. the attachment points)
- second, the router install a request-state (after having opened a COPS
session)
- third PDP asynchronously pushes the right PRIs (creation of outgoing
tunnels, forwarding rules relative to existent roles).
But what about a user that dials in the router? In this case user
traffic must be directed to the correct tunnel on per-user basis. I
guess the PEP should ask the PDP for a decision (which in turn can
interrogate e.g. a radius server). How can I manage this issue in
COPS-PR? Can the PEP install a new request state?

>>>djr The router detects an event "dial in". It outsources a REQ to
PDP and the PDP goes off to the radius server and then returns a DEC
to the PEP about the correct tunnel. This scenario is a new client
type with specific signaled clientSI objects defined. Another approach
to avoid the outsoarcing is to develop a model, say, "dial in" model. 
Develop a PIB that based on the model's semantics that can be pushed
to the PEP with COPS-PR. Just some thoughts.



I would like WG opinion on this point.

Regards,
Riccardo Scandariato