From majordomo@raleigh.ibm.com  Tue Jun  1 10:05:35 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA28974
	for <policy-archive@odin.ietf.org>; Tue, 1 Jun 1999 10:05:35 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id KAA32232;
	Tue, 1 Jun 1999 10:02:46 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id KAA24642;
	Tue, 1 Jun 1999 10:02:42 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA52672; Tue, 1 Jun 1999 09:48:24 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA52660; Tue, 1 Jun 1999 09:48:21 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id JAA22798
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 09:48:22 -0400
Received: from mail-blue.research.att.com (mail-blue.research.att.com [135.207.30.102])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id JAA17716
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 09:48:19 -0400
Received: from bigmail.research.att.com (bigmail.research.att.com [135.207.30.101])
	by mail-blue.research.att.com (Postfix) with ESMTP
	id 635EC4CE0F; Tue,  1 Jun 1999 09:48:16 -0400 (EDT)
Received: from pcraju (pc-raju.research.att.com [135.207.27.140])
	by bigmail.research.att.com (8.8.8/8.8.8) with SMTP id JAA10197;
	Tue, 1 Jun 1999 09:48:16 -0400 (EDT)
Message-Id: <00d001beac36$4e2dd2f0$8c1bcf87@research.att.com>
From: "Raju Rajan" <rajan@research.att.com>
To: "Durham, David" <david.durham@intel.com>, <policy@raleigh.ibm.com>
References: <4A043A1FE4B2D111AC3F00A0C96B513301B0F585@fmsmsx37.fm.intel.com>
Subject: Re: New Schema I-D Requested to be Posted to IETF Repository
Date: Tue, 1 Jun 1999 09:54:44 -0400
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Raju Rajan" <rajan@research.att.com>
Content-Transfer-Encoding: 7bit


----- Original Message -----
From: Durham, David <david.durham@intel.com>
To: 'Raju Rajan' <rajan@research.att.com>; <policy@raleigh.ibm.com>
Sent: Friday, May 28, 1999 8:32 PM
Subject: RE: New Schema I-D Requested to be Posted to IETF Repository

Hi David,

 That a device (PDP) should try to enforce a policy rule until it fetches
ALL the rules relevant to the PEP under consideration, is one of the primary
processing rules that are implied by the semantics of the current schema.
For instance, there could easily be (another) higher priority rule that
overrules the one under consideration. So, not only must the PDP get each
rule in its entireity befor acting, but it must get the entire rule database
(i.e., the portion relevant to its operation). I definitely see a need for
explicitly introducing this and other processing rules to clarify the
"correct" interpretation and enforcement of the rules. (I suggested a some
guidelines for this at the interim meeting.) Does this help with the second
example?

raju


> Hi Raju,
>
> > I dont still see why the issues you raise are couched (I'm referring to
> > the
> > earlier email) as "auxiliary class issues".
> >
> [[Dave]]  My dispute here is with section 5.3 "Flexibility
> Gained though Auxiliary Classes" and was relevant to my
> second example. I agree that auxiliary vs. structural classes
> is not the debate, it just depends on how they are used within
> the overall structure of the DIT. Let's see...
>
> > Your second example (below) is somewhat puzzling to me. Are you
suggesting
> > that a device ("console") will act on the policy before it retrieves a
> > full
> > LDAP object? Unless I misunderstand, this sounds more like a terrible
> > client
> > implementation (so sue me!) than an LDAP problem :)
> >
> [[Dave]]  Getting the full LDAP object is not the problem, it
> is the solution. My complaint is that when you go beyond a single
> LDAP entry to express a policy, then you are effectively opening
> up LDAP's non-transactional Achilles heel. Keeping all ANDed
> conditions within a single entry will help prevent clients from
> retrieving incomplete (=incorrect) policies. Attaching conditions
> as separate sibling entries to a policy rule implies that a client
> can retrieve a portion of the DIT that a console is in the process
> of editing. Rules need to be built from the conditions and actions
> up, or entries need to contain all attributes applicable to a rule,
> so that PDPs do not get incorrect policies. Basically, the core
> schema should be designed to take advantage of atomic transactions
> within an entry where LDAP does have limited transactional support.
> The DNF form of complex policy rules violates this principle as
> do simple policy rules (and correct me if I am wrong) where
> conditions and actions are attached as sibling entries.
>
> I have a host of other transactional issues with using auxiliary
> attachments as labels which will have to be edited across DNs all
> the while PDPs (and other consoles) are potentially querying
> (and editing) the DIT!
>




From majordomo@raleigh.ibm.com  Tue Jun  1 11:02:01 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA00381
	for <policy-archive@odin.ietf.org>; Tue, 1 Jun 1999 11:02:00 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id KAA22076;
	Tue, 1 Jun 1999 10:59:51 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id KAB30998;
	Tue, 1 Jun 1999 10:59:44 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA29912; Tue, 1 Jun 1999 10:44:15 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA28356; Tue, 1 Jun 1999 10:44:12 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id KAA31892
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 10:44:13 -0400
Received: from mail-blue.research.att.com (mail-blue.research.att.com [135.207.30.102])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id KAA34966
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 10:44:10 -0400
Received: from bigmail.research.att.com (bigmail.research.att.com [135.207.30.101])
	by mail-blue.research.att.com (Postfix) with ESMTP
	id 6025C4CE31; Tue,  1 Jun 1999 10:44:07 -0400 (EDT)
Received: from pcraju (pc-raju.research.att.com [135.207.27.140])
	by bigmail.research.att.com (8.8.8/8.8.8) with SMTP id KAA12329;
	Tue, 1 Jun 1999 10:44:07 -0400 (EDT)
Message-Id: <015501beac3e$1b4466d0$8c1bcf87@research.att.com>
From: "Raju Rajan" <rajan@research.att.com>
To: "John Schnizlein" <jschnizl@cisco.com>, "Keith McCloghrie" <kzm@cisco.com>
Cc: <policy@raleigh.ibm.com>
References: <005c01bea7ac$cb6f38f0$8c1bcf87@research.att.com> <4.1.19990527090353.00a23610@diablo.cisco.com>
Subject: Re: need for COPS - was "Suggested new charter ..."
Date: Tue, 1 Jun 1999 10:50:35 -0400
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Raju Rajan" <rajan@research.att.com>
Content-Transfer-Encoding: 7bit

Hi,

The interim policy meeting was very useful in clarifying several questions
that I had. Let me put out some of what I learned in the hope of clarifying
things for others with similar questions. (The official minutes will come
out whenever John/Ed put them out; if what I say conflicts with the minutes,
be sure I'm right :) )

1. There seemed to be considerable support for the position that policy
should be declarative and not procedural. This will be reflected in the
framework and the core documents.

2. The PIBs were not included as a work item on the charter, though it was
recognized that some means of recording PEP capabilities, policy errors and
policy usage information was vital to the correct operation of policy
servers/repositories/devices. It was thought that the PFWG had too much on
its plate already, and it would be better to deliver something and then
argue for the importance of more delivarables.

3. There seem to be two usages of the notion of PIB. One is an inclusive
notion that says that we need to store certain PEP related information
somewhere -- in the PEP, directories or elsewhere -- accessible through
multiple means -- LDAP, COPS, SNMP. I strongly support this inclusive PIB;
my personal interest is (primarily)  in using MIB formats for PIBs and
letting them be retrievable through SNMP. But I see the importance of other
protocols/information formats as well.
  The other exclusive  notion of PIBs is tied intimately to a data format
that is "slightly modified MIBs" retrievable through COPS alone. I believe
that we should not embark on such an effort. (Not that we are any danger of
doing so.)

Thanks,

raju rajan

----- Original Message -----
From: John Schnizlein <jschnizl@cisco.com>
To: Keith McCloghrie <kzm@cisco.com>; <rajan@research.att.com>
Cc: <policy@raleigh.ibm.com>
Sent: Thursday, May 27, 1999 9:26 AM
Subject: need for COPS - was "Suggested new charter ..."


> At 06:56 PM 5/26/99 -0700, Keith McCloghrie wrote:
> > in reply to Raju Rajan's:
> >> 3. If the use of PIBs is to put in hooks through the policy working
> >> group so that COPS may be the carrier of this information at a
> >> later date, I say great! But let's say so right up front, so that
> >> those of us interested in design/implementation can take a closer
> >> look at how it all fits together, what works, what doesn't and build
> >> it right. If PIBs are to be an essential component of the architecture,
> >> then let's not pretend in various architectural drawings that an
> >> LDAP-only router solution is even supported.
> >> We are saying that QoS policy is impossible without COPS.
> >> (I have no problem with such an architecture, and to the best of
> >> my knowledge, neither does AT&T :)
> >
> >I've not heard people say "QoS policy is impossible without COPS".
> >Rather, I think we've said that it's essential to allow the PEP and PDP
> >to be separated (to introduce hierarchy, and therefore scaleability),
> >and that COPS is being specified as a standard protocol for PEP-PDP
> >communication.
>
> We should recall the point of view that directories were designed
> around hierarchy, so that hierarchy was introduced with LDAP,
> although LDAP relaxes the obsession with hierarchy in X.500.
>
> A more central question is if the hierarchy of the directory
> (more attention needed on the containment issues Lee raised in MN)
> is sufficient. And the question of how dynamic policy propagation is
> needs to be resolved. Before MN, drafts differed in whether policy
> should be distributed through an event notification mechanism.
> Everyone seemed to agree signalling between nodes is necessary.
>
> It seems to me that policy can be static enough for the propagation
> speed of directory replication (which provides both reliability and
> distribution) if "policy" is defined as those rules the administrator
> wants enforced across all the state transitions in the network.
> This returns us to the question of "What is policy?" in a way that
> relates directly to the architecture, assuming the hierarchy in the
> directory can be used effectively to represent the semantics of policy.
>
> If we don't get consensus on the question if policy should be declarative
> or procedural, it is not clear we can get in on other questions.
>
> John
>



From majordomo@raleigh.ibm.com  Tue Jun  1 11:21:53 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA00843
	for <policy-archive@odin.ietf.org>; Tue, 1 Jun 1999 11:21:52 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA22350;
	Tue, 1 Jun 1999 11:20:52 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id LAA25076;
	Tue, 1 Jun 1999 11:20:50 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA22578; Tue, 1 Jun 1999 11:05:57 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA46378; Tue, 1 Jun 1999 11:05:55 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id LAA14186
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 11:05:56 -0400
Received: from mail-blue.research.att.com (mail-blue.research.att.com [135.207.30.102])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA40016
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 11:05:53 -0400
Received: from bigmail.research.att.com (bigmail.research.att.com [135.207.30.101])
	by mail-blue.research.att.com (Postfix) with ESMTP
	id C9B2F4CE0A; Tue,  1 Jun 1999 11:05:46 -0400 (EDT)
Received: from pcraju (pc-raju.research.att.com [135.207.27.140])
	by bigmail.research.att.com (8.8.8/8.8.8) with SMTP id LAA13407;
	Tue, 1 Jun 1999 11:05:46 -0400 (EDT)
Message-Id: <017f01beac41$21ade9d0$8c1bcf87@research.att.com>
From: "Raju Rajan" <rajan@research.att.com>
To: "Keith McCloghrie" <kzm@cisco.com>
Cc: <policy@raleigh.ibm.com>
References: <199905270156.SAA13513@foxhound.cisco.com>
Subject: Re: Suggested new charter for discussion at the interim meeting
Date: Tue, 1 Jun 1999 11:12:14 -0400
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Raju Rajan" <rajan@research.att.com>
Content-Transfer-Encoding: 7bit

Hi Keith,

Thanks for you informative reply. Your reply to my Question 3 says :

> I've not heard people say "QoS policy is impossible without COPS".
> Rather, I think we've said that it's essential to allow the PEP and PDP
> to be separated (to introduce hierarchy, and therefore scaleability),
> and that COPS is being specified as a standard protocol for PEP-PDP
> communication.

I agree with your answer about the neccessity of COPS for outsourced QoS
decision making. Further, your answer suggests that COPS is not required in
a network where PDPs are co-located with PEPs. It is this latter point that
I wish to see re-inforced. If data formats for policy errors and device
capabilities (for e.g., "Is RSVP turned on?") will be COPS-PIBS like and not
SNMP-MIB like, then I need COPS in my network, whether or not policy
decisions are outsourced. This is what I'm afraid of.

raju rajan

----- Original Message -----
From: Keith McCloghrie <kzm@cisco.com>
To: <rajan@research.att.com>
Cc: <policy@raleigh.ibm.com>
Sent: Wednesday, May 26, 1999 9:56 PM
Subject: Re: Suggested new charter for discussion at the interim meeting


> Raju,
>
> On day-one of SNMP, each device had at most one agent and that agent
> was for its device.  Not too many years later, we had multiple agents
> in a single chassis, agents with multiple logical devices, proxy agents
> handling requests for multiple devices, etc., etc.  I can't predict
> whether the same will the same happen for PEPs.  In any case, an issue
> when using both MIBs and PIBs will be that the COPS agent will not
> necessarily be one-to-one with an SNMP agent (or, a contextEngineID
> and contextName, see RFC 2571).
>
> > 1. I agree that we need means of recording PEP capabilities, so that
policy
> > servers and other management entities may query a network device (say a
> > router) and figure out whether it is RSVP aware, or DS capable or
whatever,
> > and the service types it supports (controlled load, expedited
forwarding..).
> > Are we using PIBs to be the common information model to record this
> > information? Why cant we use MIBs to do the same job?
>
> In this case, because you're asking about (static) "PEP capabilities",
> use of the PIB ensures that you get the right information, irrespective
> of the relationship between the PEP and associated SNMP agent(s).
>
> > 2. Further, during the operation of the device we may need to store
certain
> > device states that are germane to policy. These could include link
> > utilization or policy errors or warnings about imminent policy failure.
> > These too need to be stored and dynamically retrieved by policy
> > servers/management tools. Do we intend to use PIBs for this purpose too?
=
> > Why not MIBs?
>
> The COPS protocol contains error/warning codes about policies previously
> downloaded to that PEP.  Other network states which might be "germane to
> policy" seem to me to be appropriate for MIBs.
>
> > 3. If the use of PIBs is to put in hooks through the policy working
group so
> > that COPS may be the carrier of this information at a later date, I say
> > great! But let's say so right up front, so that those of us interested
in
> > design/implementation can take a closer look at how it all fits
together,
> > what works, what doesn't and build it right. If PIBs are to be an
essential
> > component of the architecture, then let's not pretend in various
> > architectural drawings that an LDAP-only router solution is even
supported.
> > We are saying that QoS policy is impossible without COPS. (I have no
problem
> > with such an architecture, and to the best of my knowledge, neither does
> > AT&T :)
>
>
> > 4. If we have PIBs in the architecture for QoS, one would wonder why
notfor
> > IPSec too? I believe there are suggestions to extend COPS to IPSec. It
would
> > be useful to design PIBs with IPSec in mind too.
>
> I presume you mean to define the PIB definition language with IPSec in
> mind, rather than to define a QoS PIB with IPSec in mind, yes ?
>
> Keith.
>



From majordomo@raleigh.ibm.com  Tue Jun  1 11:50:56 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA01593
	for <policy-archive@odin.ietf.org>; Tue, 1 Jun 1999 11:50:56 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA31156;
	Tue, 1 Jun 1999 11:47:12 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id LAB34818;
	Tue, 1 Jun 1999 11:47:07 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA47788; Tue, 1 Jun 1999 11:34:05 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA31904; Tue, 1 Jun 1999 11:34:02 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id LAA24588
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 11:34:04 -0400
Received: from dirty.research.bell-labs.com (dirty.research.bell-labs.com [204.178.16.6])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with SMTP id LAA29334
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 11:34:01 -0400
Received: from nova.dnrc.bell-labs.com ([135.180.131.5]) by dirty; Tue Jun  1 11:33:15 EDT 1999
Received: from dnrc.bell-labs.com (pingpanpc [135.180.130.92])
	by nova.dnrc.bell-labs.com (8.9.3/8.9.3) with ESMTP id LAA09377;
	Tue, 1 Jun 1999 11:33:14 -0400 (EDT)
Message-Id: <3753FCEF.72CD6C54@dnrc.bell-labs.com>
Date: Tue, 01 Jun 1999 11:31:59 -0400
From: Ping Pan <pingpan@dnrc.bell-labs.com>
Organization: Bell Labs
X-Mailer: Mozilla 4.51 [en] (Win98; I)
X-Accept-Language: en
Mime-Version: 1.0
To: Raju Rajan <rajan@research.att.com>
Cc: "Durham, David" <david.durham@intel.com>, policy@raleigh.ibm.com
Subject: Re: New Schema I-D Requested to be Posted to IETF Repository
References: <4A043A1FE4B2D111AC3F00A0C96B513301B0F585@fmsmsx37.fm.intel.com> <00d001beac36$4e2dd2f0$8c1bcf87@research.att.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Ping Pan <pingpan@dnrc.bell-labs.com>
Content-Transfer-Encoding: 7bit

Raju,

I agree with you that PDP's should try to do as much decision process as
possible. On the other hand, I'm not sure if a PDP can get ALL the rules
relevant to each PEP. 

For example, an edge router supports both classical RSVP and RSVP
extension for LSP establishment. On the router, one PDP is set for local
user's RSVP admission control, and another PDP is used for RSVP LSP
management with the ISP's. By the way, two PDP's belong to two separate
organizations. In this case, the PEP simply sends all received policy
information up to the corresponding PDP's and makes the final decision
(e.g., over-riding) locally.

Make sense?

I'm afraid that the concept of having one centralized PDP for each PEP
may constraint us for future services.

P.S.: Decision making on everything is centralized in SNA model. As a
result, it's rigid and slow. Let's avoid centralized model here, please.
;-)

Regards,

- Ping


Raju Rajan wrote:
> 
>  That a device (PDP) should try to enforce a policy rule until it fetches
> ALL the rules relevant to the PEP under consideration, is one of the primary
> processing rules that are implied by the semantics of the current schema.
> For instance, there could easily be (another) higher priority rule that
> overrules the one under consideration. So, not only must the PDP get each
> rule in its entireity befor acting, but it must get the entire rule database
> (i.e., the portion relevant to its operation). I definitely see a need for
> explicitly introducing this and other processing rules to clarify the
> "correct" interpretation and enforcement of the rules. (I suggested a some
> guidelines for this at the interim meeting.) Does this help with the second
> example?
> 
> raju
>


From majordomo@raleigh.ibm.com  Tue Jun  1 12:23:29 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA02343
	for <policy-archive@odin.ietf.org>; Tue, 1 Jun 1999 12:23:29 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id MAA26632;
	Tue, 1 Jun 1999 12:22:49 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id MAA24850;
	Tue, 1 Jun 1999 12:22:47 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA52624; Tue, 1 Jun 1999 12:10:03 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA42114; Tue, 1 Jun 1999 12:10:00 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id MAA21836
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 12:10:02 -0400
Received: from mail-blue.research.att.com (mail-blue.research.att.com [135.207.30.102])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id MAA34160
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 12:09:58 -0400
Received: from bigmail.research.att.com (bigmail.research.att.com [135.207.30.101])
	by mail-blue.research.att.com (Postfix) with ESMTP
	id E04F14CE05; Tue,  1 Jun 1999 12:09:52 -0400 (EDT)
Received: from pcraju (pc-raju.research.att.com [135.207.27.140])
	by bigmail.research.att.com (8.8.8/8.8.8) with SMTP id MAA15850;
	Tue, 1 Jun 1999 12:09:52 -0400 (EDT)
Message-Id: <021f01beac4a$15d46130$8c1bcf87@research.att.com>
From: "Raju Rajan" <rajan@research.att.com>
To: "Ping Pan" <pingpan@dnrc.bell-labs.com>
Cc: "Durham, David" <david.durham@intel.com>, <policy@raleigh.ibm.com>
References: <4A043A1FE4B2D111AC3F00A0C96B513301B0F585@fmsmsx37.fm.intel.com> <00d001beac36$4e2dd2f0$8c1bcf87@research.att.com> <3753FCEF.72CD6C54@dnrc.bell-labs.com>
Subject: Re: New Schema I-D Requested to be Posted to IETF Repository
Date: Tue, 1 Jun 1999 12:16:20 -0400
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Raju Rajan" <rajan@research.att.com>
Content-Transfer-Encoding: 7bit


> Raju,
>
> I agree with you that PDP's should try to do as much decision process as
> possible. On the other hand, I'm not sure if a PDP can get ALL the rules
> relevant to each PEP.

Thanks Ping. I just meant as many rules as it would normally get; I did not
intend to say that all the rules for one PEP must go to one PDP thus ruling
out distributed implementations. But even in a multi-PDP per PEP
environment, each PDP should get all the rules that it normally
would/should/must/may get before making starting to make decisions.

raju

>
> For example, an edge router supports both classical RSVP and RSVP
> extension for LSP establishment. On the router, one PDP is set for local
> user's RSVP admission control, and another PDP is used for RSVP LSP
> management with the ISP's. By the way, two PDP's belong to two separate
> organizations. In this case, the PEP simply sends all received policy
> information up to the corresponding PDP's and makes the final decision
> (e.g., over-riding) locally.
>
> Make sense?
>
> I'm afraid that the concept of having one centralized PDP for each PEP
> may constraint us for future services.
>
> P.S.: Decision making on everything is centralized in SNA model. As a
> result, it's rigid and slow. Let's avoid centralized model here, please.
> ;-)
>
> Regards,
>
> - Ping
>
>
> Raju Rajan wrote:
> >
> >  That a device (PDP) should try to enforce a policy rule until it
fetches
> > ALL the rules relevant to the PEP under consideration, is one of the
primary
> > processing rules that are implied by the semantics of the current
schema.
> > For instance, there could easily be (another) higher priority rule that
> > overrules the one under consideration. So, not only must the PDP get
each
> > rule in its entireity befor acting, but it must get the entire rule
database
> > (i.e., the portion relevant to its operation). I definitely see a need
for
> > explicitly introducing this and other processing rules to clarify the
> > "correct" interpretation and enforcement of the rules. (I suggested a
some
> > guidelines for this at the interim meeting.) Does this help with the
second
> > example?
> >
> > raju
> >
>



From majordomo@raleigh.ibm.com  Tue Jun  1 18:39:45 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA07308
	for <policy-archive@odin.ietf.org>; Tue, 1 Jun 1999 18:39:45 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id SAA21656;
	Tue, 1 Jun 1999 18:39:10 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id SAA06036;
	Tue, 1 Jun 1999 18:39:04 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA47588; Tue, 1 Jun 1999 17:24:11 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA43986; Tue, 1 Jun 1999 17:24:08 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id RAA31696
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 17:24:11 -0400
Received: from sol.extremenetworks.com (sol.extremenetworks.com [216.52.8.2])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA26504
	for <policy@raleigh.ibm.com>; Tue, 1 Jun 1999 17:24:03 -0400
Received: by ftp.extremenetworks.com with Internet Mail Service (5.5.2448.0)
	id <L8P72HC7>; Tue, 1 Jun 1999 14:23:32 -0700
Message-Id: <D0805D3B448BD211A7990008C7B18130142D8C@ftp.extremenetworks.com>
From: Andrew Smith <andrew@extremenetworks.com>
To: "'Raju Rajan'" <rajan@research.att.com>
Cc: policy@raleigh.ibm.com
Subject: RE: Suggested new charter for discussion at the interim meeting
Date: Tue, 1 Jun 1999 14:23:30 -0700 
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Andrew Smith <andrew@extremenetworks.com>

Raju,

Let's not get too hung up in this policy framework over what exactly the
device protocols/formats (PEP<->PDP) will be. COPS is good for some of it,
RADIUS for some, SNMPv1/v2/v3 for some. The really important part is that we
achieve real multi-vendor interoperability at this boundary *without*
pseudo-standards (e.g. enterprise SNMP MIBs, enterprise COPS PIBs, RADIUS
vendor-extensions).

Having said that, there does seem to be significant industry momentum
(concensus might be too strong a term) around some form of COPS PIB, at
least in the enterprise-product space, but that is meaningless until there
is a standards'-track proposal and more than a handful of vendors
implementing and shipping it.

Andrew

****************************************************************
Andrew Smith                              tel: +1 (408) 579-2821
Extreme Networks                          fax: +1 (408) 579-3000
3585 Monroe St.                   http://www.extremenetworks.com
Santa Clara CA 95051-1450        em:  andrew@extremenetworks.com
****************************************************************


> -----Original Message-----
> From: Raju Rajan [mailto:rajan@research.att.com]
> Sent: Tuesday, June 01, 1999 8:12 AM
> To: Keith McCloghrie
> Cc: policy@raleigh.ibm.com
> Subject: Re: Suggested new charter for discussion at the 
> interim meeting
> 
> 
> Hi Keith,
> 
> Thanks for you informative reply. Your reply to my Question 3 says :
> 
> > I've not heard people say "QoS policy is impossible without COPS".
> > Rather, I think we've said that it's essential to allow the 
> PEP and PDP
> > to be separated (to introduce hierarchy, and therefore 
> scaleability),
> > and that COPS is being specified as a standard protocol for PEP-PDP
> > communication.
> 
> I agree with your answer about the neccessity of COPS for 
> outsourced QoS
> decision making. Further, your answer suggests that COPS is 
> not required in
> a network where PDPs are co-located with PEPs. It is this 
> latter point that
> I wish to see re-inforced. If data formats for policy errors 
> and device
> capabilities (for e.g., "Is RSVP turned on?") will be 
> COPS-PIBS like and not
> SNMP-MIB like, then I need COPS in my network, whether or not policy
> decisions are outsourced. This is what I'm afraid of.
> 
> raju rajan
> 
> ----- Original Message -----
> From: Keith McCloghrie <kzm@cisco.com>
> To: <rajan@research.att.com>
> Cc: <policy@raleigh.ibm.com>
> Sent: Wednesday, May 26, 1999 9:56 PM
> Subject: Re: Suggested new charter for discussion at the 
> interim meeting
> 
> 
> > Raju,
> >
> > On day-one of SNMP, each device had at most one agent and that agent
> > was for its device.  Not too many years later, we had 
> multiple agents
> > in a single chassis, agents with multiple logical devices, 
> proxy agents
> > handling requests for multiple devices, etc., etc.  I can't predict
> > whether the same will the same happen for PEPs.  In any 
> case, an issue
> > when using both MIBs and PIBs will be that the COPS agent will not
> > necessarily be one-to-one with an SNMP agent (or, a contextEngineID
> > and contextName, see RFC 2571).
> >
> > > 1. I agree that we need means of recording PEP 
> capabilities, so that
> policy
> > > servers and other management entities may query a network 
> device (say a
> > > router) and figure out whether it is RSVP aware, or DS capable or
> whatever,
> > > and the service types it supports (controlled load, expedited
> forwarding..).
> > > Are we using PIBs to be the common information model to 
> record this
> > > information? Why cant we use MIBs to do the same job?
> >
> > In this case, because you're asking about (static) "PEP 
> capabilities",
> > use of the PIB ensures that you get the right information, 
> irrespective
> > of the relationship between the PEP and associated SNMP agent(s).
> >
> > > 2. Further, during the operation of the device we may 
> need to store
> certain
> > > device states that are germane to policy. These could include link
> > > utilization or policy errors or warnings about imminent 
> policy failure.
> > > These too need to be stored and dynamically retrieved by policy
> > > servers/management tools. Do we intend to use PIBs for 
> this purpose too?
> =
> > > Why not MIBs?
> >
> > The COPS protocol contains error/warning codes about 
> policies previously
> > downloaded to that PEP.  Other network states which might 
> be "germane to
> > policy" seem to me to be appropriate for MIBs.
> >
> > > 3. If the use of PIBs is to put in hooks through the 
> policy working
> group so
> > > that COPS may be the carrier of this information at a 
> later date, I say
> > > great! But let's say so right up front, so that those of 
> us interested
> in
> > > design/implementation can take a closer look at how it all fits
> together,
> > > what works, what doesn't and build it right. If PIBs are to be an
> essential
> > > component of the architecture, then let's not pretend in various
> > > architectural drawings that an LDAP-only router solution is even
> supported.
> > > We are saying that QoS policy is impossible without COPS. 
> (I have no
> problem
> > > with such an architecture, and to the best of my 
> knowledge, neither does
> > > AT&T :)
> >
> >
> > > 4. If we have PIBs in the architecture for QoS, one would 
> wonder why
> notfor
> > > IPSec too? I believe there are suggestions to extend COPS 
> to IPSec. It
> would
> > > be useful to design PIBs with IPSec in mind too.
> >
> > I presume you mean to define the PIB definition language 
> with IPSec in
> > mind, rather than to define a QoS PIB with IPSec in mind, yes ?
> >
> > Keith.
> >
> 


From majordomo@raleigh.ibm.com  Wed Jun  2 06:12:32 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id GAA26648
	for <policy-archive@odin.ietf.org>; Wed, 2 Jun 1999 06:12:32 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id GAA35342;
	Wed, 2 Jun 1999 06:11:55 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id GAA25388;
	Wed, 2 Jun 1999 06:11:50 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA54604; Wed, 2 Jun 1999 05:59:31 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA54594; Wed, 2 Jun 1999 05:59:28 -0400
Received: from nlvm1.benelux.ibm.com ([9.166.35.168])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id FAA29880
	for <policy@raleigh.ibm.com>; Wed, 2 Jun 1999 05:59:34 -0400
Message-Id: <199906020959.FAA29880@rtpmail01.raleigh.ibm.com>
Received: from UITVM1 by nlvm1.benelux.ibm.com (IBM VM SMTP V2R4)
   with BSMTP id 1150; Wed, 02 Jun 99 11:57:09 DST
Date: Wed, 2 Jun 99 11:57:09 DST
From: "Bert Wijnen" <WIJNEN@vnet.ibm.com>
To: gww@nortelnetworks.com
Cc: policy@raleigh.ibm.com
Subject: Policy Arch/framework
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Bert Wijnen" <WIJNEN@vnet.ibm.com>

Did you see draft-ietf-rap-framework-03.txt ???

Seems to address quite a few of the same architectural/framework issues.

Bert


From majordomo@raleigh.ibm.com  Wed Jun  2 20:13:58 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id UAA12605
	for <policy-archive@odin.ietf.org>; Wed, 2 Jun 1999 20:13:57 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id UAA21382;
	Wed, 2 Jun 1999 20:13:31 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id UAA34828;
	Wed, 2 Jun 1999 20:13:22 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA39740; Wed, 2 Jun 1999 20:00:43 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA50740; Wed, 2 Jun 1999 20:00:40 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id UAA34094
	for <policy@raleigh.ibm.com>; Wed, 2 Jun 1999 20:00:44 -0400
Received: from ganymede.or.intel.com (ganymede.or.intel.com [134.134.248.3])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id UAA28688
	for <policy@raleigh.ibm.com>; Wed, 2 Jun 1999 20:00:40 -0400
Received: from orsmsx28.jf.intel.com (orsmsx28.jf.intel.com [192.168.65.28])
	by ganymede.or.intel.com (8.9.1a+p1/8.9.1/d: relay.m4,v 1.6 1998/11/24 22:10:56 iwep Exp iwep $) with ESMTP id RAA11437;
	Wed, 2 Jun 1999 17:00:40 -0700 (PDT)
Received: by orsmsx28.jf.intel.com with Internet Mail Service (5.5.2448.0)
	id <M1XHTT9Q>; Wed, 2 Jun 1999 17:00:40 -0700
Message-Id: <4A043A1FE4B2D111AC3F00A0C96B513301B0F599@fmsmsx37.fm.intel.com>
From: "Durham, David" <david.durham@intel.com>
To: "'Raju Rajan'" <rajan@research.att.com>,
        "Durham, David"
	 <david.durham@intel.com>, policy@raleigh.ibm.com
Subject: RE: New Schema I-D Requested to be Posted to IETF Repository
Date: Wed, 2 Jun 1999 17:00:38 -0700 
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Durham, David" <david.durham@intel.com>

Raju,


>  That a device (PDP) should try to enforce a policy rule until it fetches
> ALL the rules relevant to the PEP under consideration, is one of the
> primary
> processing rules that are implied by the semantics of the current schema.
> For instance, there could easily be (another) higher priority rule that
> overrules the one under consideration. So, not only must the PDP get each
> rule in its entireity befor acting, but it must get the entire rule
> database
> (i.e., the portion relevant to its operation). I definitely see a need for
> explicitly introducing this and other processing rules to clarify the
> "correct" interpretation and enforcement of the rules. (I suggested a some
> guidelines for this at the interim meeting.) Does this help with the
> second
> example?
> 
	[[Dave]]  No, I don't think this helps the second example.
	Remember, the directory will be queried by PDP's at the
	same time consoles will be updating/entering entries which
	only cumulatively specify "complete" policies. So, before
	we proceed, a few questions to make sure my understanding
	is in line with your own...

	1. How does the PDP know when the policies specified in the
	directory are complete (aka. have been completely entered)?

	2. Are actions and conditions "attached" to rules in the
	simple form as a list of attributes contained within a
	single rule entry (derived from PolicyRule) or are they 
	"attached" as individual children entries in the DIT
	to a parent rule entry (as implied by figure 5)?

	2.5. Ditto for groupings of conditions/actions.

	3. When using the auxiliary attachments to label policies
	or when using the keywords attribute, how does the PDP
	know when all the relevant portions of the DIT have been
	properly labeled?

	3.5. Can labels be updated? How will PDPs know when such
	label updates have completed?


	... For a few starter questions. Clearly, my view is that
	vendor specific policy rules and groups as well as 
	conditions/actions should be DERIVED from the core schema.
	The vendor specific rule objects will then specify what 
	attributes (conditions and actions) are allowed (MAY) and
	required (MUST) for a given derived rule class. Derived
	groups can also be defined to aggregate specific conditions
	or action attributes. In all cases, a single derived rule 
	entry should specify all the relevant attributes completely
	composing a rule, and a policy object should aggregate all 
	relevant rules, etc. via reference. As all the relevant 
	information for a given level in this hierarchy is contained 
	within an LDAP entry, and entry level transactions are 
	possible, many of the current transactional problems across 
	DNs can be avoided altogether. Furthermore, the need for 
	labels can be avoided if the schema is better organized so,
	for example, all DHCP related policy/condition/action 
	instances are organized under a common root entry.

	-Dave


>  
> 


From majordomo@raleigh.ibm.com  Thu Jun  3 11:26:14 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA06884
	for <policy-archive@odin.ietf.org>; Thu, 3 Jun 1999 11:26:13 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA26796;
	Thu, 3 Jun 1999 11:22:06 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id LAA31074;
	Thu, 3 Jun 1999 11:22:02 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA39140; Thu, 3 Jun 1999 11:07:47 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA36248; Thu, 3 Jun 1999 11:07:42 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id LAA03920
	for <policy@raleigh.ibm.com>; Thu, 3 Jun 1999 11:07:44 -0400
Received: from mail-blue.research.att.com (mail-blue.research.att.com [135.207.30.102])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA31144
	for <policy@raleigh.ibm.com>; Thu, 3 Jun 1999 11:07:39 -0400
Received: from bigmail.research.att.com (bigmail.research.att.com [135.207.30.101])
	by mail-blue.research.att.com (Postfix) with ESMTP
	id B19E54CE01; Thu,  3 Jun 1999 11:07:29 -0400 (EDT)
Received: from pcraju (pc-raju.research.att.com [135.207.27.140])
	by bigmail.research.att.com (8.8.8/8.8.8) with SMTP id LAA26199;
	Thu, 3 Jun 1999 11:07:29 -0400 (EDT)
Message-Id: <005301beadd3$a8c0b9a0$8c1bcf87@research.att.com>
From: "Raju Rajan" <rajan@research.att.com>
To: "Durham, David" <david.durham@intel.com>, <policy@raleigh.ibm.com>
References: <4A043A1FE4B2D111AC3F00A0C96B513301B0F599@fmsmsx37.fm.intel.com>
Subject: Re: New Schema I-D Requested to be Posted to IETF Repository
Date: Thu, 3 Jun 1999 11:13:38 -0400
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Raju Rajan" <rajan@research.att.com>
Content-Transfer-Encoding: 7bit

Hi Dave,

Comments inline.


> Raju,
>
>
> >  That a device (PDP) should try to enforce a policy rule until it
fetches
> > ALL the rules relevant to the PEP under consideration, is one of the
> > primary
> > processing rules that are implied by the semantics of the current
schema.
> > For instance, there could easily be (another) higher priority rule that
> > overrules the one under consideration. So, not only must the PDP get
each
> > rule in its entireity befor acting, but it must get the entire rule
> > database
> > (i.e., the portion relevant to its operation). I definitely see a need
for
> > explicitly introducing this and other processing rules to clarify the
> > "correct" interpretation and enforcement of the rules. (I suggested a
some
> > guidelines for this at the interim meeting.) Does this help with the
> > second
> > example?
> >
> [[Dave]]  No, I don't think this helps the second example.
> Remember, the directory will be queried by PDP's at the
> same time consoles will be updating/entering entries which
> only cumulatively specify "complete" policies. So, before
> we proceed, a few questions to make sure my understanding
> is in line with your own...
>
> 1. How does the PDP know when the policies specified in the
> directory are complete (aka. have been completely entered)?



This is answered in the discussions surrounding the DIT and policy
retrieval. There are several mechanisms that are made available within LDAP
though I dont think any of them is mandatory at this point. To give you an
example the following could be the sequence of actions:

a) The network device (PEP/PXP) is configured with a root DN name (unique to
it, or to its group, or role, or organization, whatever).
b) The device communicates this DN to the policy server
c) The policy server (aggregating several such requests if need be) fetches
the specified DN object.
d) The DN object has the containedPolicyRule attribute with many values,
each the DN of a rule object relevant to the device.
e) The PDP/policy server fetches all these rules making sure that all the
references contained in the rule are fully "unwound". In other words, if you
fetch an IPSec action object and that further "points" to a proposal object,
then go and fetch the proposal object too (unless you got it before for some
other rule). Same for condition objects and other action objects
f) That's it. You're done.

Note: There are other grouping objects. See the core schema for more.
Note: The above is for illustration only. There are other ways of doing the
same.The point is that the schema provides the hooks, and not a method.

>
> 2. Are actions and conditions "attached" to rules in the
> simple form as a list of attributes contained within a
> single rule entry (derived from PolicyRule) or are they
> "attached" as individual children entries in the DIT
> to a parent rule entry (as implied by figure 5)?

The former, if I understand your question correctly.

>
> 2.5. Ditto for groupings of conditions/actions.

Ditto for groupings of conditions/actions.

>
> 3. When using the auxiliary attachments to label policies
> or when using the keywords attribute, how does the PDP
> know when all the relevant portions of the DIT have been
> properly labeled?
>
> 3.5. Can labels be updated? How will PDPs know when such
> label updates have completed?

I'll leave the last question to better informed people.

Hope this helps

raju

>
> -Dave
>
>
> >
> >
>



From majordomo@raleigh.ibm.com  Thu Jun  3 14:26:58 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA10865
	for <policy-archive@odin.ietf.org>; Thu, 3 Jun 1999 14:26:57 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id OAA35710;
	Thu, 3 Jun 1999 14:22:05 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id OAA23628;
	Thu, 3 Jun 1999 14:22:02 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA34518; Thu, 3 Jun 1999 14:12:56 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA34506; Thu, 3 Jun 1999 14:12:53 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id OAA27796
	for <policy@raleigh.ibm.com>; Thu, 3 Jun 1999 14:12:56 -0400
Received: from sol.extremenetworks.com (sol.extremenetworks.com [216.52.8.2])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id OAA28980
	for <policy@raleigh.ibm.com>; Thu, 3 Jun 1999 14:12:50 -0400
Received: by ftp.extremenetworks.com with Internet Mail Service (5.5.2448.0)
	id <L8P72N5Y>; Thu, 3 Jun 1999 11:12:24 -0700
Message-Id: <D0805D3B448BD211A7990008C7B18130142D9F@ftp.extremenetworks.com>
From: Andrew Smith <andrew@extremenetworks.com>
To: "'Bert Wijnen'" <WIJNEN@vnet.ibm.com>
Cc: policy@raleigh.ibm.com
Subject: RE: Policy Arch/framework
Date: Thu, 3 Jun 1999 11:12:23 -0700 
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Andrew Smith <andrew@extremenetworks.com>


The RAP Framework certainly addresses parts of the problem space of the
policy WG. But it does focus primarily on the requirements for RSVP policy
control and not QoS in general. I think that John'n'Ed's terminology draft
already incorporate some of the discussion from the RAP Framework, by
reference or by value. I certainly hope that there is nothing inconsistent
between the RAP Framework and the proposed Policy Framework - please feel
free to include any appropriate material in the Policy Framework document,
either by reference or by value (we expect draft-ietf-rap-framework-03.txt
to achieve Informational RFC status very shortly) - I'm all in favour of
providing as much tutorial material as possible in these areas.

Andrew Smith
RAP WG co-chair

> -----Original Message-----
> From: Bert Wijnen [mailto:WIJNEN@vnet.ibm.com]
> Sent: Wednesday, June 02, 1999 4:57 AM
> To: gww@nortelnetworks.com
> Cc: policy@raleigh.ibm.com
> Subject: Policy Arch/framework
> 
> 
> Did you see draft-ietf-rap-framework-03.txt ???
> 
> Seems to address quite a few of the same 
> architectural/framework issues.
> 
> Bert
> 


From majordomo@raleigh.ibm.com  Fri Jun  4 11:30:21 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA07112
	for <policy-archive@odin.ietf.org>; Fri, 4 Jun 1999 11:30:21 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA40450;
	Fri, 4 Jun 1999 11:29:09 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id LAA28898;
	Fri, 4 Jun 1999 11:28:54 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA36352; Fri, 4 Jun 1999 11:17:30 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA57848; Fri, 4 Jun 1999 11:17:28 -0400
Received: from southrelay02.raleigh.ibm.com (southrelay02.raleigh.ibm.com [9.37.3.209])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id LAA25534
	for <policy@raleigh.ibm.com>; Fri, 4 Jun 1999 11:17:32 -0400
From: remoore@us.ibm.com
Received: from d54mta04.raleigh.ibm.com (d54mta04.raleigh.ibm.com [9.67.228.36])
	by southrelay02.raleigh.ibm.com (8.8.8m2/NCO v1.8) with SMTP id LAA34222;
	Fri, 4 Jun 1999 11:17:29 -0400
Received: by d54mta04.raleigh.ibm.com(Lotus SMTP MTA v4.6.4  (830.2 3-23-1999))  id 85256786.0053FDEA ; Fri, 4 Jun 1999 11:17:24 -0400
X-Lotus-Fromdomain: IBMUS
To: policy@raleigh.ibm.com
Cc: "Durham, David" <david.durham@intel.com>,
        "'Raju Rajan'" <rajan@research.att.com>
Message-Id: <85256786.0053FB7B.00@d54mta04.raleigh.ibm.com>
Date: Fri, 4 Jun 1999 11:09:34 -0400
Subject: Transactional issues with the LDAP Core Schema
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: remoore@us.ibm.com



Dave,

You wrote the following in a response to Raju:

[[Dave]]  Getting the full LDAP object is not the problem, it
     is the solution. My complaint is that when you go beyond a single
     LDAP entry to express a policy, then you are effectively opening
     up LDAP's non-transactional Achilles heel. Keeping all ANDed
     conditions within a single entry will help prevent clients from
     retrieving incomplete (=incorrect) policies. Attaching conditions
     as separate sibling entries to a policy rule implies that a client
     can retrieve a portion of the DIT that a console is in the process
     of editing. Rules need to be built from the conditions and actions
     up, or entries need to contain all attributes applicable to a rule,
     so that PDPs do not get incorrect policies. Basically, the core
     schema should be designed to take advantage of atomic transactions
     within an entry where LDAP does have limited transactional support.
     The DNF form of complex policy rules violates this principle as
     do simple policy rules (and correct me if I am wrong) where
     conditions and actions are attached as sibling entries.

     I have a host of other transactional issues with using auxiliary
     attachments as labels which will have to be edited across DNs all
     the while PDPs (and other consoles) are potentially querying
     (and editing) the DIT!

While I understand the general point about LDAP's transactional
weakness and the resulting exposure arising from the use of multiple
entries to store related pieces of information, I'm having a hard time
finding the exposures in the specific case of the Core Policy Schema
classes.

For example, there's no chance for a client to get a "half-baked"
policy rule if the entries that make up the rule are added to the
directory (or, if they're already there, updated to say the right
thing) in the following order:

1. Entries like Lee's addressRange object, that are pointed to by
   policy condition and/or policy action objects.
2. All policyCondition and policyAction entries that the policyRule
   will point to.
3. The policyRule entry itself.

It makes no difference whether the rule combines the conditions in DNF
or CNF, since by the time the rule is there to be retrieved, all the
conditions are already present in the directory.

Ordinarily this sequence of updates isn't something that the directory
administrator will need to worry about:  the Policy Tool should take
care of it automatically.  The standard should probably contain an
implementation note saying that a Policy Tool needs to write rule-
related information to the directory in this order, to avoid the
transactional pitfalls that Dave is worried about.

Another set of semantic relationships (besides those between a rule
and its conditions and actions) in the Core Schema are those between
a policy group and the policy rules or policy groups that it
aggregates.  Once again I don't see how a client can be misled if the
order of creation (or update) for these entries is:

1. All the policyRules or policyGroups to be aggregated.
2. The policyGroup instance that aggregates them.

The jurisdiction associations, realized by attaching the policy rule
or policy group pointers to something other than policyGroups, will
work in the same way:  first create the rules / groups being pointed
to, then create the pointers to them.  A client may come to the
directory before this process has been completed, in which case it
will fail to find any of its policy rules.  But this is the same
thing that would happen if it came to the directory before any of
the updates related to its policy rules had been done.  There is, once
again, no point in time when a client might retrieve some of the rules
related to its jurisdiction, while believing that it had retrieved all
of them.

The one case that does strike me as problematic is where a client
is retrieving all the policy rules whose priorities need to be
compared against each other to decide what to do in a particular set
of circumstances.  If the client gets to these rules via a policyGroup,
then there shouldn't be a problem.  The client will know that it's in
the middle of retrieving a set of rules, and so it knows that it must
not do a priority comparison until it finishes retrieving all the
rules.  (The preceding sentence becomes more complicated if we factor
in a PDP / PEP split, but I think the principle still applies.)

Where a client can't be sure it's gotten a complete picture is when
the policy rules whose priorities compete against each other are
identified only by a keyword.  In this case an unlucky client might
search a portion of the DIT for policy rules containing the keyword,
and get back only a partial set of rules, because it happened to do
the search while the rules were being added to the directory.

In this case, if the Policy Tool adds the rules to the directory in
order of decreasing priority, the likelihood of a mishap can be
reduced.  It isn't eliminated completely, though, because of the
following possible sequence of events:

1. The client searches the portion of the DIT that will contain the
   highest priority rule, and it finds nothing.
2. The Policy Tool adds the highest priority rule to the portion of
   the DIT that the client just searched.
3. The Policy Tool adds a lower priority rule to a different portion
   of the DIT.
4. The client searches this second portion of the DIT, and finds the
   lower priority rule.

At this point the client is set up to do the "wrong" thing (i.e., the
action(s) associated with the lower priority rule) if it's
presented with a stimulus that satisfies the conditions of both rules.
Notice, though, that this has nothing to do with splitting a single
policy rule into rule, condition, and action entries.  Exactly the
same thing could happen if each rule were represented by a single
entry.  I understand that Raju has raised some additional questions /
concerns related to prioritization of policy rules.  As we address
these concerns, we should also consider how to insure that a client
doesn't do a priority comparison when it has retrieved only a subset
of the rules that belong in the set to be compared.

Regards,
Bob

Bob Moore
IBM Networking Software
+1-919-254-4436
remoore@us.ibm.com




From majordomo@raleigh.ibm.com  Fri Jun  4 16:48:15 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA10953
	for <policy-archive@odin.ietf.org>; Fri, 4 Jun 1999 16:48:15 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id QAA34608;
	Fri, 4 Jun 1999 16:46:22 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id QAA25230;
	Fri, 4 Jun 1999 16:46:03 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA57650; Fri, 4 Jun 1999 16:36:52 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA59420; Fri, 4 Jun 1999 16:36:49 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id QAA26340
	for <policy@raleigh.ibm.com>; Fri, 4 Jun 1999 16:36:53 -0400
Received: from ganymede.or.intel.com (ganymede.or.intel.com [134.134.248.3])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id QAA29046
	for <policy@raleigh.ibm.com>; Fri, 4 Jun 1999 16:36:50 -0400
Received: from orsmsx28.jf.intel.com (orsmsx28.jf.intel.com [192.168.65.28])
	by ganymede.or.intel.com (8.9.1a+p1/8.9.1/d: relay.m4,v 1.6 1998/11/24 22:10:56 iwep Exp iwep $) with ESMTP id NAA04918;
	Fri, 4 Jun 1999 13:36:50 -0700 (PDT)
Received: by orsmsx28.jf.intel.com with Internet Mail Service (5.5.2448.0)
	id <M1XHVT1D>; Fri, 4 Jun 1999 13:36:50 -0700
Message-Id: <4A043A1FE4B2D111AC3F00A0C96B513301B0F5A0@fmsmsx37.fm.intel.com>
From: "Durham, David" <david.durham@intel.com>
To: "'remoore@us.ibm.com'" <remoore@us.ibm.com>, policy@raleigh.ibm.com
Cc: "Durham, David" <david.durham@intel.com>,
        "'Raju Rajan'"
	 <rajan@research.att.com>
Subject: RE: Transactional issues with the LDAP Core Schema
Date: Fri, 4 Jun 1999 13:36:47 -0700 
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Durham, David" <david.durham@intel.com>

Thanks Bob, this helps clear things up a bit with my second
example. I see that by following a well-defined procedure 
a policy editor tool can enter policy in a way that will 
avoid many of the "half-baked" policy problems despite the lack 
of transactions. This information should probably find its 
way into the specification because policy editors that do 
not follow the "correct" procedure will create problems.
It would be even better if the LDAPized core schema could
somehow force the correct procedure. 

I still think the issue remains for policies that already 
exist but are instead being edited/updated. Let me see if 
I can conjure up an illustrative example for such a case...
Say, policy "re-ordering" in the case rules are prioritied, 
or redefining two groups already ANDed together within a 
policy rule. Seems as if even more restrictions will have
to be placed on the policy editor for cases as these where
dependencies cross the boundaries of a single entry.

Also, can you explain if there is a procedure for modifying 
auxiliary attachments that act as labels (e.g. through
the policyKeywords attribute) such that lack of transactional
support is not a pitfall here?

-Dave


> -----Original Message-----
> From:	remoore@us.ibm.com [SMTP:remoore@us.ibm.com]
> Sent:	Friday, June 04, 1999 8:10 AM
> To:	policy@raleigh.ibm.com
> Cc:	Durham, David; 'Raju Rajan'
> Subject:	Transactional issues with the LDAP Core Schema
> 
> 
> 
> Dave,
> 
> You wrote the following in a response to Raju:
> 
> [[Dave]]  Getting the full LDAP object is not the problem, it
>      is the solution. My complaint is that when you go beyond a single
>      LDAP entry to express a policy, then you are effectively opening
>      up LDAP's non-transactional Achilles heel. Keeping all ANDed
>      conditions within a single entry will help prevent clients from
>      retrieving incomplete (=incorrect) policies. Attaching conditions
>      as separate sibling entries to a policy rule implies that a client
>      can retrieve a portion of the DIT that a console is in the process
>      of editing. Rules need to be built from the conditions and actions
>      up, or entries need to contain all attributes applicable to a rule,
>      so that PDPs do not get incorrect policies. Basically, the core
>      schema should be designed to take advantage of atomic transactions
>      within an entry where LDAP does have limited transactional support.
>      The DNF form of complex policy rules violates this principle as
>      do simple policy rules (and correct me if I am wrong) where
>      conditions and actions are attached as sibling entries.
> 
>      I have a host of other transactional issues with using auxiliary
>      attachments as labels which will have to be edited across DNs all
>      the while PDPs (and other consoles) are potentially querying
>      (and editing) the DIT!
> 
> While I understand the general point about LDAP's transactional
> weakness and the resulting exposure arising from the use of multiple
> entries to store related pieces of information, I'm having a hard time
> finding the exposures in the specific case of the Core Policy Schema
> classes.
> 
> For example, there's no chance for a client to get a "half-baked"
> policy rule if the entries that make up the rule are added to the
> directory (or, if they're already there, updated to say the right
> thing) in the following order:
> 
> 1. Entries like Lee's addressRange object, that are pointed to by
>    policy condition and/or policy action objects.
> 2. All policyCondition and policyAction entries that the policyRule
>    will point to.
> 3. The policyRule entry itself.
> 
> It makes no difference whether the rule combines the conditions in DNF
> or CNF, since by the time the rule is there to be retrieved, all the
> conditions are already present in the directory.
> 
> Ordinarily this sequence of updates isn't something that the directory
> administrator will need to worry about:  the Policy Tool should take
> care of it automatically.  The standard should probably contain an
> implementation note saying that a Policy Tool needs to write rule-
> related information to the directory in this order, to avoid the
> transactional pitfalls that Dave is worried about.
> 
> Another set of semantic relationships (besides those between a rule
> and its conditions and actions) in the Core Schema are those between
> a policy group and the policy rules or policy groups that it
> aggregates.  Once again I don't see how a client can be misled if the
> order of creation (or update) for these entries is:
> 
> 1. All the policyRules or policyGroups to be aggregated.
> 2. The policyGroup instance that aggregates them.
> 
> The jurisdiction associations, realized by attaching the policy rule
> or policy group pointers to something other than policyGroups, will
> work in the same way:  first create the rules / groups being pointed
> to, then create the pointers to them.  A client may come to the
> directory before this process has been completed, in which case it
> will fail to find any of its policy rules.  But this is the same
> thing that would happen if it came to the directory before any of
> the updates related to its policy rules had been done.  There is, once
> again, no point in time when a client might retrieve some of the rules
> related to its jurisdiction, while believing that it had retrieved all
> of them.
> 
> The one case that does strike me as problematic is where a client
> is retrieving all the policy rules whose priorities need to be
> compared against each other to decide what to do in a particular set
> of circumstances.  If the client gets to these rules via a policyGroup,
> then there shouldn't be a problem.  The client will know that it's in
> the middle of retrieving a set of rules, and so it knows that it must
> not do a priority comparison until it finishes retrieving all the
> rules.  (The preceding sentence becomes more complicated if we factor
> in a PDP / PEP split, but I think the principle still applies.)
> 
> Where a client can't be sure it's gotten a complete picture is when
> the policy rules whose priorities compete against each other are
> identified only by a keyword.  In this case an unlucky client might
> search a portion of the DIT for policy rules containing the keyword,
> and get back only a partial set of rules, because it happened to do
> the search while the rules were being added to the directory.
> 
> In this case, if the Policy Tool adds the rules to the directory in
> order of decreasing priority, the likelihood of a mishap can be
> reduced.  It isn't eliminated completely, though, because of the
> following possible sequence of events:
> 
> 1. The client searches the portion of the DIT that will contain the
>    highest priority rule, and it finds nothing.
> 2. The Policy Tool adds the highest priority rule to the portion of
>    the DIT that the client just searched.
> 3. The Policy Tool adds a lower priority rule to a different portion
>    of the DIT.
> 4. The client searches this second portion of the DIT, and finds the
>    lower priority rule.
> 
> At this point the client is set up to do the "wrong" thing (i.e., the
> action(s) associated with the lower priority rule) if it's
> presented with a stimulus that satisfies the conditions of both rules.
> Notice, though, that this has nothing to do with splitting a single
> policy rule into rule, condition, and action entries.  Exactly the
> same thing could happen if each rule were represented by a single
> entry.  I understand that Raju has raised some additional questions /
> concerns related to prioritization of policy rules.  As we address
> these concerns, we should also consider how to insure that a client
> doesn't do a priority comparison when it has retrieved only a subset
> of the rules that belong in the set to be compared.
> 
> Regards,
> Bob
> 
> Bob Moore
> IBM Networking Software
> +1-919-254-4436
> remoore@us.ibm.com
> 


From majordomo@raleigh.ibm.com  Fri Jun  4 18:02:48 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA11890
	for <policy-archive@odin.ietf.org>; Fri, 4 Jun 1999 18:02:47 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id SAA21708;
	Fri, 4 Jun 1999 18:02:33 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id SAA27512;
	Fri, 4 Jun 1999 18:02:32 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA30876; Fri, 4 Jun 1999 17:54:59 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA30864; Fri, 4 Jun 1999 17:54:55 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id RAA28016
	for <policy@raleigh.ibm.com>; Fri, 4 Jun 1999 17:55:00 -0400
Received: from nsm-mail2.cisco.com (nsm-mail2.cisco.com [171.71.236.25])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA33958
	for <policy@raleigh.ibm.com>; Fri, 4 Jun 1999 17:54:56 -0400
Received: from jstrassnpc.cisco.com ([171.69.108.130])
	by nsm-mail2.cisco.com (8.8.8-Cisco List Logging/8.8.8) with SMTP id OAA21238
	for <policy@raleigh.ibm.com>; Fri, 4 Jun 1999 14:54:27 -0700 (PDT)
Message-Id: <036801beaed4$f2240820$826c45ab@cisco.com>
From: "John Strassner" <jstrassn@cisco.com>
To: <policy@raleigh.ibm.com>
Subject: NEW TIMES FOR OSLO
Date: Fri, 4 Jun 1999 14:55:22 -0700
Organization: Cisco Systems
Mime-Version: 1.0
Content-Type: text/plain;
	charset="utf-8"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "John Strassner" <jstrassn@cisco.com>
Content-Transfer-Encoding: 7bit

Our new times our now:

Monday, July 12
    1930-2200  Evening Sessions
Tuesday, July 13
    1545-1645  Afternoon Sessions III

regards,
John



From majordomo@raleigh.ibm.com  Fri Jun  4 21:07:44 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA13572
	for <policy-archive@odin.ietf.org>; Fri, 4 Jun 1999 21:07:43 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id VAA24916;
	Fri, 4 Jun 1999 21:06:01 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id VAA32976;
	Fri, 4 Jun 1999 21:05:53 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA48472; Fri, 4 Jun 1999 20:58:10 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA42320; Fri, 4 Jun 1999 20:58:07 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id UAA24700;
	Fri, 4 Jun 1999 20:58:13 -0400
Received: from nsm-mail2.cisco.com (nsm-mail2.cisco.com [171.71.236.25])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id UAA23918;
	Fri, 4 Jun 1999 20:58:11 -0400
Received: from jstrassnpc.cisco.com ([171.69.108.130])
	by nsm-mail2.cisco.com (8.8.8-Cisco List Logging/8.8.8) with SMTP id RAA03489;
	Fri, 4 Jun 1999 17:57:33 -0700 (PDT)
Message-Id: <00de01beaeee$86532d00$826c45ab@cisco.com>
From: "John Strassner" <jstrassn@cisco.com>
To: <policy@raleigh.ibm.com>
Cc: <fred@cisco.com>, <johns@cisco.com>, <ellesson@raleigh.ibm.com>,
        <wijnen@vnet.ibm.com>, <randy@psg.com>
Subject: Minutes - Interim Policy Framework Meeting
Date: Fri, 4 Jun 1999 17:58:27 -0700
Organization: Cisco Systems
Mime-Version: 1.0
Content-Type: text/plain;
	charset="utf-8"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "John Strassner" <jstrassn@cisco.com>
Content-Transfer-Encoding: 7bit

Proposed Minutes are attached directly below. Please send
comments to the list if you have any additions.

regards,
John and Ed

------------------------------ snip away from here
down ---------------------------------

Policy Framework WG Interim Meeting
Ft. Collins, Colorado
May 27-28, 1999

Thursday, May 27:
-----------------

Attendees:

1. Ed Ellesson, IBM, ellesson@raleigh.ibm.com
2. John Strassner, Cisco, johns@cisco.com
3. Scott Hahn, Intel, scott.hahn@intel.com
4. Hugh Mahon, HP, hugh_mahon@hp.com
5. Chris Martin, Xylan, chris.martin@xylan.com
6. Raju Rajan, ATT, rajan@research.att.com
7. Mark Stevens, Lucent, markstevens@lucent.com
8. Fran Reichmeyer, Nortel, freichmeyer@baynetworks.com
9. Bert Wijnen, IBM, wijnen@vnet.ibm.com
10. Patrick Thompson, Microsoft, pthomp@microsoft.com
11. Shigeru Miyake, Hitachi, yake@sdl.hitachi.co.jp
12. Tomoichi Ebata, Hitachi, ebata@sdl.hitachi.co.jp
13. Ryan Moats, ATT, jayhawk@att.com
14. Glenn Waters, Nortel, gww@nortelnetworks.com
15. David Partain, Ericsson, David.Partain@ericsson.com
16. Andrew Smith, Extreme, andrew@extremenetworks.com
17. Lee Rafalow, IBM, rafalow@raleigh.ibm.com
18. Yzhak Ronen, ATT, ronen@att.com
19. Randy Bush, Verio, randy@psg.com


Agenda Bash (Ed)
----------------

Agenda approved.

Thank you, to Hugh Mahon and HP, for sponsoring this interim
meeting,
and for arranging the logistics. The conference room, lunch, and
refreshments were most appreciated!

Discussion of "Missing Piece" to Working Group's Efforts (Straz)
----------------------------------------------------------------

 -What is the missing piece?
 -What document/deliverable does it need to be in?
 -What context does the working group need, in order to make
progress?

Summary of John Strassner's presentation, and associated
discussion:

SLA/SLO's in the "Policy Meta Model":

We can talk about the existence of high level business policies
as
part of a total solution, but business level semantics are not
the
purview of the IETF.

We can show a block diagram in the policy architecture that
reflects
the business level, but we can not and should not address the
business models that are enabled at this level, in our
discussions
and documents.  We can only address the services enabled at this
level in the broadest terms, such that we understand the support
required for the interface between the device-level, device-
independent policy building blocks (e.g., the structure that we
are
standardizing) and the next level up (i.e., the business level
policy).

In the application of policy to networks, the content of this
business level view of policy is termed SLA's (Service Level
Agreements), and SLO's (Service Level Objectives). The SLA/SLO
(business) level of policy is the "business contract" level which
we are not standardizing in the IETF.

Raju: Agreed that you can't dictate the business model, or the
services, which is what the business level policies deal with.

Bert: Suggests that the arrows in the picture need to go upward,
rather than downward, on the Policy Meta Model. (This was agreed
by consensus of those present.)

More about the Meta Model:  Some confusion expressed about what
each
level in the diagram represents, at an abstraction level. Need
more
description in the work group documents about what we mean by
policy and schema.

Clarification:  The Core Policy Schema and the
discipline-specific
sub-classed schemata that will follow from this (QOS, IPSEC,
etc.)
are both at the same level in the Policy Meta Model. This level
is
the technology-independent, vendor-independent, device-level
policy
in the middle block, that we are standardizing. The PIB (Policy
Information Base) exists as a connection to the device-specific
MIB,
which is at the lower-level in the meta model.

Hugh:  Stressed need to reflect in this schema what is do-able,
in
terms of existing technologies and protocols, at the device
level.

Straz:  The schema at the "Policy" level in this diagram needs to
know about groups of users and devices, that apply to policy, but
the
PIB does not. The PIB needs to understand groups of ports,
interfaces, etc. that represent the user at the device level.

Lee:  Recommends that we don't spend time on defining the upper
boundary of policy. Too different from company to company.
Vendors
will innovate there. Can't agree on this in this group. Work on
Policy level instead, and then discipline-specific work will
iterate on this.

Hugh: Agrees we don't want to standardize the SLA/SLO's. But
suggests
that we do need to define what we are not defining. Need a
description of it with examples. This captures the influence
without standardizing it.

Mark: Agrees with Hugh. This process will help us understand the
level of abstraction we need. Starting from the top with examples
will help. Need to work through the examples by deconstructing
them.

Fran: We need to describe what we mean by "up" and "down" in the
meta model.

John:  Example: Let's say the goal is to provide intranet access.
SLO describes VPN tunnels with some set of characteristics.
"Policy", as defined by the policy building block in the Meta
Model,
is the device-independent specification of device-level
configuration,
which is specific to the parameters that technically define the
means
of carrying out the policy with conditions and actions. The
schema
can not, and will not, specify specific configuration
information,
since that is device-specific. The PIB is intended to first
announce
the capabilities of a device to the PDP, and translate policy
specifications to device-specific MIBs. That way, the same policy
can be applied to heterogeneous devices, and those devices can
give
relatively the same service to a given type of traffic even if
their
capabilities and mechanisms are different.

Bert: agreed. 1 Mb/s is the device-independent param, but Gold
Service
is higher level, and may be different by ISP for the same
bandwidth.

Raju: Concerned with the level of abstraction at the lowest level
in
the Meta Model, while still being device-independent and
implementation method-independent. This is hard.

Patrick: Volunteering to bring in 30 or 40 examples of SLA level
definitions, as a means to describe what the Policy module in the
meta model needs to support.

John: Great, but need to be QOS Networking examples. Others can
be
archived for future work.

Lee: Should we have breadth across disciplines in the schema
examples?

John: No, we need to dive into QOS as a worked example. Our goal
should focus on showing that we can define how to translate SLAs
to a
form implementable by heterogeneous devices that delivers end2end
services.

Raju: Problem with the process: doing the examples in multiple
disciplines is critical.

Mark: Don't wait until the car is finished before you take it out
on
the test track.

John: We must demonstrate a full end2end solution. This has more
value that portions of solutions that aren't end2end.

Bert: If you go to proposed standard, the IETF is saying you
should
implement to this, so we can see if it works. The IETF does not
guarantee that things will not change at Proposed Standard. The
vendor takes a risk that the standard won't change.

Procedures are clear. You have to show multiple implementations
of
every feature, in order to move it to draft standard. That's when
the
IETF takes a stand that there are multiple independent
implementations. Recycling at proposed is ok and expected.
Retooling required.

Yzhak Ronen: Agrees with the levels we have identified. Core
Policy
independent of discipline, discipline-specific is next. Build on
that
with specific types of QOS schemata, and then vendor builds on
that.
When you change things in the device level, with new version, or
new
version of RSVP, don't want to have to change the next level up.

The challenge of abstracting. Also, informaton sharing is a
factor
here. That is why we are not standardizing the business level.

Raju: May not be able to do RSVP and DiffServ "under" QOS, but
maybe
have parallel RSVP, DiffServ, and generic QOS discipline that is
common, and potentially beyond these two. Maybe add this to the
charter.  E.g., maybe needs to be inclusive of ATM, FR CIR, end
system concepts of QOS, so you can get to a end-to-end QOS.

Andrew: Do vendors subclass further?

Ryan: For vendor-specific things, you might want to use another
protocol besides LDAP.

Andrew: Does the schema to PIB arrow carry the vendor-specific
stuff?

Mark: What about vendor-interoperability? Need to specify the
policy
in vendor-independent way, and let the vendors innovate and
differentiate in how they accomplish and carryout the policy.

Andrew: The vendor-specific stuff needs to be opaque. Must accept
the
fact that there will be product differentiation. Does not need to
prevent interoperability if the abstraction is general enough.

Scott Hahn: Don't want CLI commands in the schema! Let's defer to
the schema discussion.


Summary of what we agreed to, during the above discussion:

1. Work group will talk about high level (SLA/SLO) level as
examples
   only, to create and validate the building blocks in the core
and
   QoS schemata. Patrick volunteered to provide 30-40 SLA
examples;
   John and Ed to filter these into a smaller set specific to
this
   working group's intentions. We are not standardizing this
level
   (SLA/SLOs) in this group.

2. Show the arrows going up in the meta model, from policy to the
   higher level. Ensure that this is put in the terminology
draft,
   and perhaps rename the terminology draft. Explain specifically
   what the arrows mean.

3. Policy, as expressed in higher- to lower-levels, needs to
reflect
   what is practical in technology, within a particular
discipline,
   such as QOS. The end goal is to develop policy schemata
related
   specifically to protocols standardized in the IETF.

4. We focus on IETF QOS in this charter as the worked example.
This
   includes RSVP and DiffServ. We are open as to whether we
consider
   specific QoS mechanisms of specific media (e.g., FR CIR).

5. We need to formalize a responsive process of getting input
from
   other working groups. Schema Doctors and Chairs to implement a
   formal process as a feedback mechanism. In Oslo, we should
have a
   meeting of relevant working group chairs and editors and
   Schema Doctors.

6. We can advance the core schema to proposed standard, when the
   worked example proves that it works in that one case, at
least.
   This case is defined to be QoS. It is likely that no one
vendor
   will implement the entire core and QOS schema; however, it is
   certain that a set of vendors will do so, thus enabling the
   interoperability proof for further advancement.

7. Vendor interoperability: target is to provide interoperable
   policy, with implementation differences masked by the level
   of abstraction in the policy schema that is standardized.

8. The QOS schema has multiple pieces, including RSVP/INTSERV,
   DiffServ, and generic QOS. Structuring these will be part of
the
   on-going process.

Thursday Charter Discussion (Straz):
------------------------------------

Note:  See also Friday's Charter Discussion, which was documented
in
the form of the changes to the proposed charter. John Strassner
to
post this revised proposed charter to the mailing list for
further
comment and wordsmithing.

John presented Fran's architecture chart, to help us understand
how
to position the role of architecture in the process of the
working
group. John's point is: policy drives architecture, not the other
way around.

Raju: wants the "architecture" document to expand and elaborate
on
what is in the charter, to further explain the purpose of policy,
and the overall approach we are taking as a working group.

John would like to reposition the "architecture" document
deliverable
as a "framework" document instead. Hopefully, this re-missioning
of
the document will re-focus the work to a different level, and
therefore enable a greater degree of consensus.

It was agreed that we must scope the framework doc to include
elaboration on the charter, and to carefully describe the purpose
of
the policy work we are doing, at a minimum. The framework doc
should
also go forward at the same time as the core schema to wg last
call.
Glenn Waters will put the doc together, and take input.
Working group must agree on its content.

Policy for QOS: We are using networked QOS as the validation case
for
the core schema, as well as for serving the policy needs of the
multiple qos protocols and disciplines developed by IETF working
groups, including RAP, DiffServ, IntServ, RSVP, and ISSLL.
The framework doc should reflect this.

Declarative vs. Procedural: We agreed long ago (Chicago), that
this
working group would base its work on a declarative model, not a
procedural one.  Framework doc must state this.

PIB and Related Issues:
-----------------------

Add bullet to charter: We are capturing dynamic data in the
framework, but not in the schema. This allows the RAP group to
proceed. This is one place where the PIB comes in.

Edited draft charter to include "PIB-like" mission and an
associated
document. Some disagree with the addition of PIB to the mission.
Does
it de-focus us? John argued that it is a necessary piece of
accomplishing our mission (end2end policy control and
expression).

Bert: The IESG AD's may not agree to this PIB expansion of the
charter, until the Policy wg has advanced some drafts. Other
groups,
such as RAP, have a proven track record of accomplishing last
calls.
This could be a later piece of work for the Policy Framework WG,
just the same as MIB's are done in other working groups, or,
alternatively, the RAP wg could do it. Needs more discussion.

In the end, there was not consensus to add the PIB to the
charter.
Expect further discussion on this topic on the list.

One idea: change the name of the core policy schema to read
information model and core schema. Randy believes we need to
clearly
delineate the information model separate from the storage model.
(Note: On Friday, we agreed to split the "core schema" document,
currently a single monolithic document, into two separate
documents,
one for the information model, and one for the core policy
schema.)

Further discussion on the charter was tabled until Friday
morning,
in order to give Lee time to present the schema changes.


Lee's presentation: Schema
--------------------------

-Simple Policy Rule:

Lee described the new approach to simple and complex policy
rules,
using aux classes. Note that the underlying information model is
independent of these changes.

The simple policy rule representation maximizes lookup efficiency
(which meets a requirement of those working on a schema for dhcp
and other wgs). The complex policy rule, on the other hand, is
oriented toward data efficiency, in that it enables re-use of
condition and action instances, (which meets a requirement of
many
of those working on pre-standard versions of a schema for other
wgs, such as QoS). The complex rule mechanism also enables the
DNF
and CNF mechanisms currently described in the core schema, which
the
simple policy rule approach does not support.

Raju's proposal in Minneapolis: By way of reference, Raju's
proposal
for the structure of the policy rule included both simple and
complex
policy rules as separate classes.

-Problems with the proposed structure:

Lee reviewed problems that have been pointed out with the
proposed
policy schema (-03 version), which makes extensive use of
auxilliary
classes (this from email, DMTF, and Schema Doctors discussions):

  1. Objectclass attribute values: There is a problem with
searching
     on the object class type in one important directory
     implementation, when the object class is picked up with the
     aux class that is attached to it. One way around this
problem
     is to use keywords for searching and put a word like
"policy"
     in the keyword list.

     -DHCP example was presented, with the address range class
      playing multiple policy functions. Address range can be
both
      a condition semantic and an action semantic. That is, when
      a DHCP system is choosing an address range to apply to a
      given user, the address range object is an "action" in the
      DHCP policy schema. When an element of a QOS system is
      choosing a queuing strategy based on the address range of a
      packet, that same address range object is a "condition" in
the
      QOS policy schema. If each policy subtree (DHCP and QOS)
had
      its own address range object instances, policy
administration
      would be more difficult. Thus, object re-use simplifies the
      job of the policy administrator. The PolicyElement aux
class
      object serves as simply a flag that says this object class
to
      which it is attached, in this example, address range
object,
      is a policy thingie, so that directory searches on "policy"
      will find it, and so that policy parsers will identify it
      easily and properly.

     -Note: Search on keywords vs. object class type may have
      security implications, since, in mainstream directory
      implementations, one  can enable only certain users to
query
      on certain object classes using Access Control Lists
(ACL's),
      but there are no limits on keyword searches.)

  2. Attach binding time:

     -One directory implementation (the same one as noted in item
#1,
      above), binds the aux class to its associated structural
classes
      at schema definition time. This would require that the
      PolicyElement aux class be bound to all structural classes
for
      which it could possibly serve as a flag, at the time of
      initially loading the schema into the directory.
Potentially,
      this could mean attaching the PolicyElement aux class to
all
      structural elements in the directory. For those instances
which
      are not policy objects at the time, the PolicyElement
      attributes would be assigned null values. Unfortunately,
this
      eliminates the searching efficiency of the PolicyElement
flag
      function. For this specific directory implementation, one
would
      have to use keyword instead.

  3. Derivation from top.

     The same directory implementation noted in items # 1 and #2,
     above, requires that aux classes can only inherit from
"top".
     This prevents our schema aux classes from inheriting
attributes
     from higher level policy classes. The work around for this
     directory implementation is that one must redefine what
would
     otherwise be inherited attributes, in each of the aux
classes.

     Another implication: every time you subclass
PolicyCondition,
     for discipline-specific attributes, one would have to repeat
the
     PolicyCondition attributes again, in the subclass. This is
     vendor-specific, for that directory implementation.

     -Discussion (of schema changes and associated
implementation-
      specific problems):

      Ryan: 1. Parsing structured pointers is hard for thin ldap
               clients
            2. LDAP-accessible directories should have an ability
               to unwind references.

      Response to Ryan's comment #1: it is true that dn pointers
and
      structured pointers require code that is not in a standard
      ldap client api implementation, but people are already
writing
      that code for their pre-standard implementations. It works.
      Discipline-specific code is required in any case.

      Response to Ryan's comment #2: Reference chasing
(unwinding)
      suggestion: this should be proposed to the LDAPEXT wg, but
when
      will LDAPEXT get to this? This could be DOA for us if we
are
      dependent on it, leading to a long time lag when
directories
      finally support reference unwinding.

      Raju: 1. Reviewed PolicyRule, which subclassed into
               SimplePolicyRule and ComplexPolicyRule.

            2. Why we need aux classes: Without multiple
inheritance,
               you have to create new subclasses every time you
have
               new attributes to add, when you start with
structural
               objects. With aux classes, you get the "effect" of
               multiple inheritance, you get to mix and match
what
               attributes you want into an instance, without
               attribute clashes.

      Lee: a byproduct of the implementation-specific set of
           problems, is that, all the attributes must be "mays"
in
           the aux classes. There are many attributes that should
be
           "musts", per the semantic of the attribute.

      The use of aux classes requires that our working group
create
      an appendix of implementation notes with the above
      implications documented.

      -Discussion of using LDAP-accessible directories as the
       IETF-standardized "point of interoperability" for policy
(as
       opposed to, for example, a relational database approach):

      Randy: Are these ldap only issues? Are we commiting to ldap
as
             the only representation? How do we feel about the
             impedance mismatch of policy with ldap?

      Lee: Marketplace is demanding that we use ldap at this
point
           in time.

      Chris: Customers like what ldap-based directories already
do.
             We have implementation experience with them for
policy,
             and we also have work-arounds for many of ldap's
             shortcomings.

      Randy: Are we putting kludges in the schema as a result of
the
             binding to ldap?

      Patrick: Maybe we should have an info model instead, or in
               addition to, the ldap schema.

      John: We do. We're separating that out into two documents
to
            make this clearer.

      Agreement: Call the current schema an ldap schema, and the
                 info model something separate. (Later, we
decided
                 to make these separate documents. See the new
                 proposed charter.)

      Suggestion: (Randy and Ryan): The working group should
create
                  a separate information model draft, a draft
based
                  on current ldap standards, and a later
                  "ldapfuture" draft. (Base the ldapfuture on
                  ability to unwind references, for example. All
                  three will need to be standards track
documents.
                  One of the problems this addresses is the fact
                  that we are introducing kludges being made to
the
                  schema in the name of ldap limitations. We need
                  consensus on this approach on the mailing list.

  4. Policygroup and Policyrule associations:
     PolicyXXXContainmentAuxClasses. These are simply pointers to
     link groups to groups and groups to rules.

     Discussion and preliminary conclusion: One could use DIT
     containment to create these associations; however, this
     restricts re-use. (DIT containment was suggested by Steven
Judd
     on a DMTF phone call.) DIT containment also restricts
     implementors to not use the DIT to create other
associations.
     Finally, we have to realize that we do not have control of
the
     DIT! Customers have already defined their DITs, and >90% of
the
     time, the DIT was defined without taking Policy needs into
     account. DITs are already being used for administrative
     access control, by administrator. Therefore, the wg members
     present at this meeting reject this notion, and decide to
     stick with what we have for linking groups to groups and
groups
     to rules, subject to consensus on the mailing list.

  5. Efficiency of retrieval issue, using the subtree pointer
     mechanism in the schema:

     If you want to get some significant set of policy entries
     identified to the pdp, you can do this thru the subtree
     pointer. One operation, lots of objects. One retrieval per
     policy "sub-domain".

     Example shown. If you never implement it, you don't break
     interoperability, you just have less efficiency in lookups.

     PDP's need to be able to be configured with key words
defined
     by the customer, and by the discipline-specific schemas,
like
     dhcp.

     How do PDPs learn what their starting points are? Point high
     in the tree and have it point down. Blue, short, fat and
west,
     determines the set of objects I need to go to.

     Ryan wants an attribute at the root, defined in an I-D, so
that
     a new PDP can come up and find out where to go in the
subtrees.
     It's a well-defined attribute, perhaps a well-defined dn.
     Ldapext would have to document this. Ryan is proposing this
to
     the protocol group. It goes into the root of the tree.

     This is a change to the protocol spec for ldap. This solves
the
     PDP starting point problem.

     Could do this with a well-known dn, and register that. Blue,
     red:dn.

     Raju: Security problem, if you don't trust your PDPs. There
     are no ACLs on individual entries. You can put ACLs on the
     attributes. Needs more thought. Suggestion: Let the MIB be
     configured by the mgmnt tool, and it tells the PDP where its
     ldap directory is, and also can tell it what its starting
     points are.

  6. Choice of RDN:

     Discussion:

     We added multiple attribute names because we thought that
typing
     was a Good Thing. Following the CIM model, we thought that
we
     would equate CIM's CreationClassName to the LDAP-accessible
     directory's ObjectClass and CIM's Name to LDAP's cn. Problem
is
     that certain directory implementations don't support
multi-part
     RDNs, and many people don't like them. So we use a typed
name.

     Some people object to typed names due to GUI issues, and to
     overlapping semantics.

     In most implementations of ldap-accessible directories, any
     attribute can be a dn. It does not matter.

     Ryan: Must have multiple attributes in forming an rdn. (Lee:
     But the one implementation, the same one as noted in
previous
     comments as an exception to the rule, does not allow this.
     Without a lot of extra effort, that implementation requires
     that the rdn be the cn.)

     Lee: Proposes either policyClassname or cn to be the rdn.

     Typed names: should we have them? Does this put a
restriction
     on CIM to LDAP mapping? Note: XML mapping effort is
proposing
     that CIM change in 3.0 to drop creation class name.
(Patrick)

     Agreement (subject to mailing list consensus):

     Ryan wants "any attribute suitable for naming to be used to
     form the rdn". This group should define what those
attributes
     are. This does not imply a multi-part rdn as a requirement.

     Class-specific names. Consensus in this room:
     problem for IETF is resolved. (It is understood that this
     does not solve the CIM-to-ldap mapping problem.)

     Ryan: Customers will decide their own dit naming scheme.

     Lee: Policy gui's will probably not present ldap-looking
names.


Friday discussion:
------------------

  -Not recorded in meeting yesterday, but recalled from
yesterday:
   Discussion about finding your server, and finding an object or
   attribute in the directory that will tell it where its policy
is.
   Lee will prepare an update to the schema that involves subtree
   pointers in the root of the suffix.

Security discussion (Ed)

Need to identify users and user requirements in each of the
following
disciplines:
  - integrity
  - authentication
  - authorization
  - access control
  - confidentiality
  - non-repudiation

Russ talked about the three forms that we need to consider -
client,
on the wire, and end2end at the Minneapolis meeting.

Discussion on access control. Security through obscurity will not
fly.
But what about super-users, do we need to encrypt the data or can
we
just rely on directory ACLs?

Also denial-of-service concerns. Note that integrity means
verifying
that the data at the destination matches that of the source -
this is
different than encrypting on the wire.

What about security concerns when using LDAP? We either need to
get
LDAP to fix it or not use LDAP. Non-repudiation is what the
directory
service does in terms of an audit trail, and may be out-of-scope
with
respect to the protocol. Thus, it's a requirement for the
solution,
but not of LDAP (e.g., the audit trail would be stored in
the directory).

Protocols don't do this, implementers should be warned. We need
to say
that such auditing should be done, and that the customer should
consider this and choose a directory server that meets his or her
needs. We don't need a statement defining a minimum set of
functions
that must be logged. (Note that SNMPv3 didn't).

Seems like we should use LDAPv3 over TLS. Note that LDAP has as
its
mandatory-to-implement mechanism HTTP Digest, which is
insufficient
for large-scale deployment. We should say SHOULD for using TLS,
and
MAY for using HTTP Digest.

Three types of users - administrators of the schema,
administrators
of the schema content, and consumers like PDPs. Each needs their
own
specific security guidelines and permissions.

Directory service access control is problematic, since the
ACL/ACI
model is not yet standard and is still being worked on in the
LDAPEXT WG. The other problem, of course, is interoperability
between multiple directory solutions. So we need to document the
two administrative user cases. Note that ACLs are bound to the
user identity that logs onto the directory.

Problem in using LDAP security is that if we decide to use TLS,
there is no standard key distribution mechanism. One
recommendation
is to say that we do nothing to break this. We could also use
Kerberos. Perhaps the best solution is to recommend a strong
security mechanism (e.g., TLS OR Kerberos) and to advise against
using weaker mechanisms (clear text, HTTP Digest, etc.).

Denial-of-service is a serious problem, and we recommend the use
of multiple directories. However, this just makes it a "little
bit better" and doesn't really solve the problem. Note that the
schema is unaffected - we're not adding new ways for people to
set up a denial-of-service attack. We should document the fact
that such an attack, if executed successfully, could prevent PDPs
from getting to the directory. Note that DNSSEC is a similar
model
- zone files are not encrypted, but it does do on-the-wire
encryption.

We need to have boilerplate that says that the machine should be
put in a secure location, etc., etc.

We should recommend mutual authentication. The protocol supports
this, but the functionality varies with directory implementation.
We put text similar to what is in MIBs in the schema, and the
detailed comments in the framework document. We'll involve Russ
and organize a discussion on the list, possibly supplemented with
a conference call.

Suggestion: beyond what is transmitted on the wire, everything
else is out-of-scope. We need to document these things, but not
solve them. For example, DNSSEC and SNMPv3 did not talk about
how data is stored. We should say that we do nothing to
break anything or make it worse.

Why should we be treating policy schema any different than other
schema or MIBs?


  -Charter discussion - wordsmithed it in the meeting. See the
   proposed revised charter, to be posted by John Strassner.

  -Procedural issues:
   All ldap schema disccussion that occurs on other lists, such
as
   the DMTF SLA working group list needs to be cross-posted to
the
   IETF policy frameworks list. There are IETF members that are
not
   DMTF members. Solution: Use policy list for everything except
   pure CIM stuff.

   Comment: Need to put slides on a web site for the group

   -03 schema is different that the current visio. Needs
updating.
    (Action:  Bob Moore.)

   -New Framework draft: Glenn will put table of contents
together
    on the mailing list

Schema issues presentation (by Raju Rajan)
------------------------------------------

List of Tradeoffs:

Expressibility: closer to natural language, kind of conditions
and
                actions

Functionality: Analysis needed of constraints that protocol is
               capable of.

Query Optimization: Scalability issues

Data Optimization: Size and storage of data required for policy
Schema sanity checks and managability: Dynamic pointers and
loops.
If delete a user, how many places do I have to go and change
things

Parsing: How much effort required on part of client to parse the
policy info into something it can use.

Execution:  How much burden is added to the data path due to
policy.
(Use of terms execution and limits was controversial.)

Lee: Policy is the device-independent specification of
configuration.
It's not on the execution path.

The PDP and PEP have the job of turning this into execution
instructions.

Raju: The policies we enable should be reasonably compilable and
executable, and flag that this is a problem with some pep
devices.

Chris: where is this documented?

Raju: this is a list of trade-offs. Agrees that this is not
necessarily quantifiable as a set constraints in the wg
documentation.

  1. Expressibility: Raju objects to using the term "finite state
     machine" as the way we characterize what we are doing. It
was
     agreed to define this more strictly in the framework draft.

  2. Declarative vs. procedural. Declarative says if you see this
     condition, then go to this state. Procedural says, if this
state
     then do this. We want to use declarative. (Agreed to by
     all present; to be added to the framework draft.

     Action: Raju will write some words that describe that we are
     restricting our work to declarative, and what that means.

  3. Policy Scope: Make it multi-valued. Check this. A number of
     people have concerns about what it means, anyway.

  4. Hybrid Rules: not prevented by the syntax (combined simple
and
     complex). Raju wants simple and complex rules as separate
     objects. This would enforce the constraint without just text
     caveats. Lee: you have to do validity checks anyway.

     How can a particular discipline schema prevent or require
     simple vs. complex, if it wants to constrain this choice?
     Attribute collisions can result. Wants to give this
alternative
     to specify this by the discipline-specific schema.

     Needs more discussion on the list

  5. Priority and Policy Action Ordering

     -What happens when rules have actions for two different
      disciplines (e.g., both ipsec and qos actions)? How does
      rule priority resolve this, when the priority is on the
      rule and not the action?

      One solution would be for the discipline specific-schemas
      to resolve this.

      The expressibility needs to be there for the disciplines.
      Do we want to constrain the expressibility mechanisms in
      the core schema with structures that allow/constrain what
      is needed?

      The core document must say that certain policies can
conflict
      with the same priority and log an error.

      Andrew suggests that priority should be within the context
of
      the action discipline.

      Lee: Maybe should limit rules to having only one action
      discipline. Rules need to be complete for a particular
action
      and discipline. Requires careful definition of a
discipline.
      Pick and choose from different rules you have to be
complete.

      We need this discussion on the mailing list. Raju will
write
      this up for the list.

      Raju wants to remove action ordering from the spec. Let the
      action ordering be determined by the requirements of the
      disciplines. Requires rules about how to resolve ordering
in
      multiple rules when they both apply.

  6. Complicated Booleans:  DNF, CNF and Negation

     -Compiling of rules: processing time, lookup speed (search
      depth), and memory tree size. Decision tree may be in pdp,
or
      pep, if diffserv.  Could impact data path complexity.

      What is the sizing impact. Just raising the flag.

     -Suggests that we might want to re-examine CNF?.

Raju's presentation and associated discussion was the last agenda
item for this interim meeting. Further discussion prior to Oslo
will
be on the Policy Frameworks mailing list, policy@raleigh.ibm.com.

Thanks to all who attended, and especially to Hugh Mahon, and HP,
Ft. Collins, who sponsored the meeting.

regards,
John and Ed



From majordomo@raleigh.ibm.com  Fri Jun  4 21:18:24 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA13615
	for <policy-archive@odin.ietf.org>; Fri, 4 Jun 1999 21:18:23 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id VAA29998;
	Fri, 4 Jun 1999 21:17:56 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id VAA11006;
	Fri, 4 Jun 1999 21:17:53 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA57568; Fri, 4 Jun 1999 21:10:27 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA30936; Fri, 4 Jun 1999 21:10:25 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id VAA14330;
	Fri, 4 Jun 1999 21:10:31 -0400
Received: from nsm-mail2.cisco.com (nsm-mail2.cisco.com [171.71.236.25])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id VAA33806;
	Fri, 4 Jun 1999 21:10:29 -0400
Received: from jstrassnpc.cisco.com ([171.69.108.130])
	by nsm-mail2.cisco.com (8.8.8-Cisco List Logging/8.8.8) with SMTP id SAA04282;
	Fri, 4 Jun 1999 18:09:55 -0700 (PDT)
Message-Id: <00f601beaef0$40bf1040$826c45ab@cisco.com>
From: "John Strassner" <jstrassn@cisco.com>
To: <policy@raleigh.ibm.com>
Cc: <fred@cisco.com>, <johns@cisco.com>, <ellesson@raleigh.ibm.com>,
        <wijnen@vnet.ibm.com>, <randy@psg.com>
Subject: Draft charter agreed to at the Interim Meeting
Date: Fri, 4 Jun 1999 18:10:49 -0700
Organization: Cisco Systems
Mime-Version: 1.0
Content-Type: text/plain;
	charset="utf-8"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "John Strassner" <jstrassn@cisco.com>
Content-Transfer-Encoding: 7bit

Everyone,

below is the proposed draft charter that was agreed to by the
participants of the Interim Meeting. This draft modifies the
original draft sent earlier on this list.

Note that one new addition is item 7, proposed by Lee Rafalow as
promised at the interim meeting. Also note that the PIB is not
officially proposed as added to the charter yet, pending further
discussion.

regards,
John and Ed

-------------------------  begin new proposed charter
text -----------------------------

1. Identify a set of representative use cases to guide us in
defining a policy framework, information model, and schemata to
store, retrieve and process policies. These use cases should map
to a set of policy rules, and aid us in defining the composition
of policies.

2. Define a framework for intra-domain policy definition and
administration for a heterogeneous set of Policy Decision and
Enforcement Points. The framework will be shown to be able to
represent and manage policy information in an unambiguous,
interoperable manner in a single administrative domain. The
framework has the ability to store policy information in a policy
repository. This will be documented in the framework and
information model drafts. This framework will be applied to
network QoS. It is envisioned that this framework can also
support other network services based on protocols defined by the
IETF, such as IPSec, DHCP, and VPNs.

3. The main result of this work will be an LDAP schema for
storing policy information in a directory. This will be
documented in the core and network QoS schema drafts. Care will
be taken to ensure that this work is extensible to other types of
repositories, such as relational databases.

4. The schemata developed will focus on generic network policies
for both provisioned and signaled QoS services.

This working group will work with the DiffServ, IntServ, RAP, and
ISSLL working groups to ensure that the QoS schema meets their
technical needs, and is suitable for both IPv4 and IPv6 standards
track work.

5. To ensure that security issues are addressed, Russ Mundy will
serve as this working group's security adviser.

6. The charter specifically excludes:

  -protocol definition
  -schema attributes or classes that are vendor-specific
   (although the schema defined in this group will be defined in
    a way that is extensible by specific vendors)

// NEW ITEM - see discussion for this new item at the end
7. Because policy is a device-independent specification of
configuration
information, the policy information model and schema provide for
infrequently
changing declarative information that reflect network rules
rather than network
state.  The Policy Framework WG may request that other IETF
Working
Groups provide enhancements to protocols that they develop to
communicate
network state information (e.g., current network user
identification) so that
rules can be written to use additional dynamic stateinformation
in applying the
policy rules.

Goals and Milestones:

1. Policy terminology draft (Informational Track)
- Working Group Last Call just after Oslo IETF
2. Core information model draft (Informational Track?)
- Working group Last Call September 1999
3. Core LDAP schema draft (Standards Track)
- Working group Last Call September 1999
4. Framework draft (Informational Track)
- Working group Last Call, December 1999
5. QoS Schema draft(s) (Standards Track)
- First issuance, June, 1999
- Working group Last Call, just before Adelaide IETF, Spring 2000

-----------  end new proposed charter text, start discussion of
item 7 -----------

Some supporting discussion for item #7 (from Lee):

We need to be clearer about the difference between the
interpretation of policy rules and the data path execution logic
associated with the policy rules.
Policy rules specify the *configuration* of the network that is
required to meet
SLOs in a device-independent way.   Policy, then, is the source
language that is compiled (by the PDP) into a target language
specification (device
configuration) of the logic needed to implement those policies on
a particular
platform. I suspect we all agree, more or less, with these
statements.

Where clarity and consensus are replaced by confusion and
disagreement, I think, is that some of the data on which the
policy target language description
operates is not currently available to the PEPs using existing
protocols. Therefore, I think, people may be tempted to think of
dynamically changing the
existing configuration parameters to make up for this missing
state information and, thus, conclude that they are implementing
"dynamic policy" rather than providing data on which the policy
can operate through the same interfaces that the policy itself is
input into the target system.

The last point to make now is that providing this state data as
configuration
changes may be appealing (and may even be necessary in the
short-term) but it may be asking for trouble.  Whenever there are
two paths along which information about a flow and the flow
itself may traverse, there are potential race conditions.  To
solve these race conditions, one must either have a
synchronization mechanism or must be able to tolerate loose
consistency.  I'll
assert without defence that policies can live with loose
consistency (hence LDAP is an adequate choice for communicating
them) but network state information must be consistent in order
for the network to behave in a consistent and predictable
fashion.  Therefore, inline communication of applicable state
changes will normally be the better technical answer; i.e.,
protocols need to be enhanced to synchronize awareness of events
that are input to policy enforcement.



From majordomo@raleigh.ibm.com  Mon Jun  7 11:12:33 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA23553
	for <policy-archive@odin.ietf.org>; Mon, 7 Jun 1999 11:12:33 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA32604;
	Mon, 7 Jun 1999 11:06:39 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id LAA26528;
	Mon, 7 Jun 1999 11:06:33 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA44742; Mon, 7 Jun 1999 10:44:46 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA44724; Mon, 7 Jun 1999 10:44:41 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id KAA25840;
	Mon, 7 Jun 1999 10:44:37 -0400
Received: from mail-blue.research.att.com (mail-blue.research.att.com [135.207.30.102])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id KAA34602;
	Mon, 7 Jun 1999 10:44:31 -0400
Received: from bigmail.research.att.com (bigmail.research.att.com [135.207.30.101])
	by mail-blue.research.att.com (Postfix) with ESMTP
	id 66A7F4CE06; Mon,  7 Jun 1999 10:44:28 -0400 (EDT)
Received: from pcraju (pc-raju.research.att.com [135.207.27.140])
	by bigmail.research.att.com (8.8.8/8.8.8) with SMTP id KAA03553;
	Mon, 7 Jun 1999 10:44:27 -0400 (EDT)
Message-Id: <019001beb0f5$19f44eb0$8c1bcf87@research.att.com>
From: "Raju Rajan" <rajan@research.att.com>
To: "John Strassner" <jstrassn@cisco.com>, <policy@raleigh.ibm.com>
Cc: <fred@cisco.com>, <johns@cisco.com>, <ellesson@raleigh.ibm.com>,
        <wijnen@vnet.ibm.com>, <randy@psg.com>
References: <00f601beaef0$40bf1040$826c45ab@cisco.com>
Subject: Re: Draft charter agreed to at the Interim Meeting
Date: Mon, 7 Jun 1999 10:50:35 -0400
Mime-Version: 1.0
Content-Type: text/plain;
	charset="utf-8"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Raju Rajan" <rajan@research.att.com>
Content-Transfer-Encoding: 7bit

Hi,

I had a comment about the newly added item on the policy charter:

> // NEW ITEM - see discussion for this new item at the end
> 7. Because policy is a device-independent specification of configuration
> information, the policy information model and schema provide for
infrequently
> changing declarative information that reflect network rules rather than
network
> state.  The Policy Framework WG may request that other IETF Working
> Groups provide enhancements to protocols that they develop to communicate
> network state information (e.g., current network user identification) so
that
> rules can be written to use additional dynamic stateinformation in
applying the
> policy rules.

I agree  with the first sentence. The second is a vague non-sequitur. Is it
intended to give us a mandate to *demand* or to *suggest* ? For the latter
we don't need a mandate. We *may* always ask another working group if
they'll do so and so for us. If the former, we pretty much allow ourselves
to improve any protocol in IETF. The example given (current network user
identification) is a part of the PDP decision making process. The PDP has to
bind policy categories -- user name, application name, interface type,
whatever -- with dynamic information -- user name to source address, for
instance. Then it may instruct the PEP to carry out the right operations on
the packet stream. If my understanding is in keeping with Lee's explanation
below (which I believe it is), then the protocol enhancements that could
potentially be requested are a very large set. We could impact protocols
between a route server--network devices, Kerberos authentication, Radius
server, billing servers, etc.

I would plead that we postpone this mandate till we have a clear idea of
which protocol/protocols will be impacted, what exactly we need, and how we
intend to achieve it. In other words, I request that we clearly state that
we will not be involved with binding of policy categories (conditions) to
header information as yet, and just drop the second sentence. Thanks for
your patience,

raju


>
> Some supporting discussion for item #7 (from Lee):
>
> We need to be clearer about the difference between the
> interpretation of policy rules and the data path execution logic
> associated with the policy rules.
> Policy rules specify the *configuration* of the network that is
> required to meet
> SLOs in a device-independent way.   Policy, then, is the source
> language that is compiled (by the PDP) into a target language
> specification (device
> configuration) of the logic needed to implement those policies on
> a particular
> platform. I suspect we all agree, more or less, with these
> statements.
>
> Where clarity and consensus are replaced by confusion and
> disagreement, I think, is that some of the data on which the
> policy target language description
> operates is not currently available to the PEPs using existing
> protocols. Therefore, I think, people may be tempted to think of
> dynamically changing the
> existing configuration parameters to make up for this missing
> state information and, thus, conclude that they are implementing
> "dynamic policy" rather than providing data on which the policy
> can operate through the same interfaces that the policy itself is
> input into the target system.
>
> The last point to make now is that providing this state data as
> configuration
> changes may be appealing (and may even be necessary in the
> short-term) but it may be asking for trouble.  Whenever there are
> two paths along which information about a flow and the flow
> itself may traverse, there are potential race conditions.  To
> solve these race conditions, one must either have a
> synchronization mechanism or must be able to tolerate loose
> consistency.  I'll
> assert without defence that policies can live with loose
> consistency (hence LDAP is an adequate choice for communicating
> them) but network state information must be consistent in order
> for the network to behave in a consistent and predictable
> fashion.  Therefore, inline communication of applicable state
> changes will normally be the better technical answer; i.e.,
> protocols need to be enhanced to synchronize awareness of events
> that are input to policy enforcement.
>
>



From majordomo@raleigh.ibm.com  Mon Jun  7 15:27:54 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA29503
	for <policy-archive@odin.ietf.org>; Mon, 7 Jun 1999 15:27:54 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA31146;
	Mon, 7 Jun 1999 15:27:11 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id PAA26222;
	Mon, 7 Jun 1999 15:26:51 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA32634; Mon, 7 Jun 1999 15:11:27 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA38258; Mon, 7 Jun 1999 15:11:23 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id PAA29390
	for <policy@raleigh.ibm.com>; Mon, 7 Jun 1999 15:11:25 -0400
Received: from sol.extremenetworks.com (sol.extremenetworks.com [216.52.8.2])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA21170
	for <policy@raleigh.ibm.com>; Mon, 7 Jun 1999 15:11:13 -0400
Received: by ftp.extremenetworks.com with Internet Mail Service (5.5.2448.0)
	id <L8P72VVX>; Mon, 7 Jun 1999 12:10:28 -0700
Message-Id: <D0805D3B448BD211A7990008C7B18130142DC5@ftp.extremenetworks.com>
From: =?utf-8?B?QW5kcmV3IFNtaXRo?= <andrew@extremenetworks.com>
To: =?utf-8?B?J1JhanUgUmFqYW4n?= <rajan@research.att.com>,
        policy@raleigh.ibm.com
Cc: fred@cisco.com, wijnen@vnet.ibm.com, randy@psg.com
Subject: =?utf-8?B?UkU6IERyYWZ0IGNoYXJ0ZXIgYWdyZWVkIHRvIGF0IHRoZSBJbnRl?=
	=?utf-8?B?cmltIE1lZXRpbmc=?=
Date: Mon, 7 Jun 1999 12:10:27 -0700 
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="utf-8"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: =?utf-8?B?QW5kcmV3IFNtaXRo?= <andrew@extremenetworks.com>


I don't see that the proposed new charter item 7 adds much useful guidance.
I also do not think that we need any explicit non-goals regarding "binding
of policy categories (conditions) to header information". But it would be
good to capture some of Lee's discussion/argument in the framework document.

Andrew

****************************************************************
Andrew Smith                              tel: +1 (408) 579-2821
Extreme Networks                          fax: +1 (408) 579-3000
3585 Monroe St.                   http://www.extremenetworks.com
Santa Clara CA 95051-1450        em:  andrew@extremenetworks.com
****************************************************************


> -----Original Message-----
> From: Raju Rajan [mailto:rajan@research.att.com]
> Sent: Monday, June 07, 1999 7:51 AM
> To: John Strassner; policy@raleigh.ibm.com
> Cc: fred@cisco.com; johns@cisco.com; ellesson@raleigh.ibm.com;
> wijnen@vnet.ibm.com; randy@psg.com
> Subject: Re: Draft charter agreed to at the Interim Meeting
> 
> 
> Hi,
> 
> I had a comment about the newly added item on the policy charter:
> 
> > // NEW ITEM - see discussion for this new item at the end
> > 7. Because policy is a device-independent specification of 
> configuration
> > information, the policy information model and schema provide for
> infrequently
> > changing declarative information that reflect network rules 
> rather than
> network
> > state.  The Policy Framework WG may request that other IETF Working
> > Groups provide enhancements to protocols that they develop 
> to communicate
> > network state information (e.g., current network user 
> identification) so
> that
> > rules can be written to use additional dynamic stateinformation in
> applying the
> > policy rules.
> 
> I agree  with the first sentence. The second is a vague 
> non-sequitur. Is it
> intended to give us a mandate to *demand* or to *suggest* ? 
> For the latter
> we don't need a mandate. We *may* always ask another working group if
> they'll do so and so for us. If the former, we pretty much 
> allow ourselves
> to improve any protocol in IETF. The example given (current 
> network user
> identification) is a part of the PDP decision making process. 
> The PDP has to
> bind policy categories -- user name, application name, interface type,
> whatever -- with dynamic information -- user name to source 
> address, for
> instance. Then it may instruct the PEP to carry out the right 
> operations on
> the packet stream. If my understanding is in keeping with 
> Lee's explanation
> below (which I believe it is), then the protocol enhancements 
> that could
> potentially be requested are a very large set. We could 
> impact protocols
> between a route server--network devices, Kerberos 
> authentication, Radius
> server, billing servers, etc.
> 
> I would plead that we postpone this mandate till we have a 
> clear idea of
> which protocol/protocols will be impacted, what exactly we 
> need, and how we
> intend to achieve it. In other words, I request that we 
> clearly state that
> we will not be involved with binding of policy categories 
> (conditions) to
> header information as yet, and just drop the second sentence. 
> Thanks for
> your patience,
> 
> raju
> 
> 
> >
> > Some supporting discussion for item #7 (from Lee):
> >
> > We need to be clearer about the difference between the
> > interpretation of policy rules and the data path execution logic
> > associated with the policy rules.
> > Policy rules specify the *configuration* of the network that is
> > required to meet
> > SLOs in a device-independent way.   Policy, then, is the source
> > language that is compiled (by the PDP) into a target language
> > specification (device
> > configuration) of the logic needed to implement those policies on
> > a particular
> > platform. I suspect we all agree, more or less, with these
> > statements.
> >
> > Where clarity and consensus are replaced by confusion and
> > disagreement, I think, is that some of the data on which the
> > policy target language description
> > operates is not currently available to the PEPs using existing
> > protocols. Therefore, I think, people may be tempted to think of
> > dynamically changing the
> > existing configuration parameters to make up for this missing
> > state information and, thus, conclude that they are implementing
> > "dynamic policy" rather than providing data on which the policy
> > can operate through the same interfaces that the policy itself is
> > input into the target system.
> >
> > The last point to make now is that providing this state data as
> > configuration
> > changes may be appealing (and may even be necessary in the
> > short-term) but it may be asking for trouble.  Whenever there are
> > two paths along which information about a flow and the flow
> > itself may traverse, there are potential race conditions.  To
> > solve these race conditions, one must either have a
> > synchronization mechanism or must be able to tolerate loose
> > consistency.  I'll
> > assert without defence that policies can live with loose
> > consistency (hence LDAP is an adequate choice for communicating
> > them) but network state information must be consistent in order
> > for the network to behave in a consistent and predictable
> > fashion.  Therefore, inline communication of applicable state
> > changes will normally be the better technical answer; i.e.,
> > protocols need to be enhanced to synchronize awareness of events
> > that are input to policy enforcement.
> >
> >
> 


From majordomo@raleigh.ibm.com  Mon Jun  7 15:28:40 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA29519
	for <policy-archive@odin.ietf.org>; Mon, 7 Jun 1999 15:28:40 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA22292;
	Mon, 7 Jun 1999 15:27:17 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id PAA28700;
	Mon, 7 Jun 1999 15:26:51 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA12004; Mon, 7 Jun 1999 15:12:41 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA11994; Mon, 7 Jun 1999 15:12:38 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id PAA22958;
	Mon, 7 Jun 1999 15:12:40 -0400
Received: from bmailnj.iphighway.com ([209.3.6.76])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA30782;
	Mon, 7 Jun 1999 15:12:35 -0400
Received: from shai-notebook (SHAI_NOTEBOOK [209.3.6.99]) by bmailnj.iphighway.com with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2448.0)
	id KN70SRNV; Mon, 7 Jun 1999 15:07:47 -0400
Message-Id: <4.1.19990607150621.04e25100@bmailnj.iphighway.com>
X-Sender: shai_pop@bmailnj.iphighway.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.1 
Date: Mon, 07 Jun 1999 15:11:18 -0400
To: "Raju Rajan" <rajan@research.att.com>,
        "John Strassner" <jstrassn@cisco.com>, <policy@raleigh.ibm.com>
From: Shai Herzog <herzog@iphighway.com>
Subject: Re: Draft charter agreed to at the Interim Meeting
Cc: <fred@cisco.com>, <johns@cisco.com>, <ellesson@raleigh.ibm.com>,
        <wijnen@vnet.ibm.com>, <randy@psg.com>
In-Reply-To: <019001beb0f5$19f44eb0$8c1bcf87@research.att.com>
References: <00f601beaef0$40bf1040$826c45ab@cisco.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Shai Herzog <herzog@iphighway.com>

At 10:50 AM 6/7/99, Raju Rajan wrote:

>I would plead that we postpone this mandate till we have a clear idea of
>which protocol/protocols will be impacted, what exactly we need, and how we
>intend to achieve it. In other words, I request that we clearly state that
>we will not be involved with binding of policy categories (conditions) to
>header information as yet, and just drop the second sentence. 

I am not sure if I misunderstand, or disagree with you, but let me
try to clarify it.

Are you saying that the WG should work on a schema that includes
dynamic binding (e.g., user name, path info, etc), but shouldn't 
worry about how this is done (COPS, etc.) for now? OR, are you saying
that we shouldn't allow dynamic binding criteria into the schema? (which
I don't think you meant to say).

Shai

>Thanks for your patience,

What patience???? ;-)

Shai


__________________________________________________________
Shai Herzog                  Founder & CTO, IPHighway Inc.         
Cell: (408) 390-3045       http://www.iphighway.com/herzog           
Tel : (201) 585-0800                  Fax : (212) 656-1006 


From majordomo@raleigh.ibm.com  Mon Jun  7 15:40:16 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA29752
	for <policy-archive@odin.ietf.org>; Mon, 7 Jun 1999 15:40:16 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA26106;
	Mon, 7 Jun 1999 15:39:56 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id PAA28630;
	Mon, 7 Jun 1999 15:39:54 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA27710; Mon, 7 Jun 1999 15:22:06 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA38190; Mon, 7 Jun 1999 15:22:04 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id PAA28492;
	Mon, 7 Jun 1999 15:22:06 -0400
Received: from mail-blue.research.att.com (mail-blue.research.att.com [135.207.30.102])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA31994;
	Mon, 7 Jun 1999 15:22:03 -0400
Received: from bigmail.research.att.com (bigmail.research.att.com [135.207.30.101])
	by mail-blue.research.att.com (Postfix) with ESMTP
	id 9A6BF4CE06; Mon,  7 Jun 1999 15:22:04 -0400 (EDT)
Received: from pcraju (pc-raju.research.att.com [135.207.27.140])
	by bigmail.research.att.com (8.8.8/8.8.8) with SMTP id PAA16171;
	Mon, 7 Jun 1999 15:22:04 -0400 (EDT)
Message-Id: <004a01beb11b$e0810de0$8c1bcf87@research.att.com>
From: "Raju Rajan" <rajan@research.att.com>
To: "John Strassner" <jstrassn@cisco.com>, <policy@raleigh.ibm.com>,
        "Shai Herzog" <herzog@iphighway.com>
Cc: <fred@cisco.com>, <johns@cisco.com>, <ellesson@raleigh.ibm.com>,
        <wijnen@vnet.ibm.com>, <randy@psg.com>
References: <00f601beaef0$40bf1040$826c45ab@cisco.com> <4.1.19990607150621.04e25100@bmailnj.iphighway.com>
Subject: Re: Draft charter agreed to at the Interim Meeting
Date: Mon, 7 Jun 1999 15:28:09 -0400
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Raju Rajan" <rajan@research.att.com>
Content-Transfer-Encoding: 7bit

I'm saying that we should not put anything into the charter till we
understand what it entails. Otherwise we'll have a second PFDL disaster.

raju

----- Original Message -----
From: Shai Herzog <herzog@iphighway.com>
To: Raju Rajan <rajan@research.att.com>; John Strassner
<jstrassn@cisco.com>; <policy@raleigh.ibm.com>
Cc: <fred@cisco.com>; <johns@cisco.com>; <ellesson@raleigh.ibm.com>;
<wijnen@vnet.ibm.com>; <randy@psg.com>
Sent: Monday, June 07, 1999 3:11 PM
Subject: Re: Draft charter agreed to at the Interim Meeting


> At 10:50 AM 6/7/99, Raju Rajan wrote:
>
> >I would plead that we postpone this mandate till we have a clear idea of
> >which protocol/protocols will be impacted, what exactly we need, and how
we
> >intend to achieve it. In other words, I request that we clearly state
that
> >we will not be involved with binding of policy categories (conditions) to
> >header information as yet, and just drop the second sentence.
>
> I am not sure if I misunderstand, or disagree with you, but let me
> try to clarify it.
>
> Are you saying that the WG should work on a schema that includes
> dynamic binding (e.g., user name, path info, etc), but shouldn't
> worry about how this is done (COPS, etc.) for now? OR, are you saying
> that we shouldn't allow dynamic binding criteria into the schema? (which
> I don't think you meant to say).
>
> Shai
>
> >Thanks for your patience,
>
> What patience???? ;-)
>
> Shai
>
>
> __________________________________________________________
> Shai Herzog                  Founder & CTO, IPHighway Inc.
> Cell: (408) 390-3045       http://www.iphighway.com/herzog
> Tel : (201) 585-0800                  Fax : (212) 656-1006
>



From majordomo@raleigh.ibm.com  Mon Jun  7 16:16:29 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA00278
	for <policy-archive@odin.ietf.org>; Mon, 7 Jun 1999 16:16:28 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id QAA30110;
	Mon, 7 Jun 1999 16:15:59 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id QAA25552;
	Mon, 7 Jun 1999 16:15:55 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA38302; Mon, 7 Jun 1999 15:58:31 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA31858; Mon, 7 Jun 1999 15:58:20 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id PAA23832;
	Mon, 7 Jun 1999 15:58:23 -0400
Received: from nsm-mail2.cisco.com (nsm-mail2.cisco.com [171.71.236.25])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA27712;
	Mon, 7 Jun 1999 15:58:09 -0400
Received: from jstrassnpc.cisco.com (newyork-dhcp-128.cisco.com [171.68.50.128])
	by nsm-mail2.cisco.com (8.8.8-Cisco List Logging/8.8.8) with SMTP id MAA12553;
	Mon, 7 Jun 1999 12:57:17 -0700 (PDT)
Message-Id: <027601beb120$14e563c0$803244ab@cisco.com>
From: "John Strassner" <jstrassn@cisco.com>
To: "Raju Rajan" <rajan@research.att.com>,
        "John Strassner" <jstrassn@cisco.com>, <policy@raleigh.ibm.com>
Cc: <fred@cisco.com>, <johns@cisco.com>, <ellesson@raleigh.ibm.com>,
        <wijnen@vnet.ibm.com>, <randy@psg.com>
References: <00f601beaef0$40bf1040$826c45ab@cisco.com> <019001beb0f5$19f44eb0$8c1bcf87@research.att.com>
Subject: Re: Draft charter agreed to at the Interim Meeting
Date: Mon, 7 Jun 1999 12:58:08 -0700
Organization: Cisco Systems
Mime-Version: 1.0
Content-Type: text/plain;
	charset="utf-8"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "John Strassner" <jstrassn@cisco.com>
Content-Transfer-Encoding: 7bit

Raju,

we agreed in the Interim meeting to the first 6 items, hence
their inclusion in the charter. We also agreed to let Lee suggest
his addition, which is why I included Lee's discussion that gave
his rationalization for requesting this addition to the charter.

I'll let Lee answer your non-sequitur issues. I think we can make
a decision to add or not add Lee's suggestion once feedback roles
in. Right now, this is the first message I've seen (though I'm on
travel with spotty Internet access, and am sure yours will
generate others).

regards,
John

----- Original Message -----
From: Raju Rajan <rajan@research.att.com>
To: John Strassner <jstrassn@cisco.com>; <policy@raleigh.ibm.com>
Cc: <fred@cisco.com>; <johns@cisco.com>;
<ellesson@raleigh.ibm.com>; <wijnen@vnet.ibm.com>;
<randy@psg.com>
Sent: Monday, June 07, 1999 7:50 AM
Subject: Re: Draft charter agreed to at the Interim Meeting


> Hi,
>
> I had a comment about the newly added item on the policy
charter:
>
> > // NEW ITEM - see discussion for this new item at the end
> > 7. Because policy is a device-independent specification of
configuration
> > information, the policy information model and schema provide
for
> infrequently
> > changing declarative information that reflect network rules
rather than
> network
> > state.  The Policy Framework WG may request that other IETF
Working
> > Groups provide enhancements to protocols that they develop to
communicate
> > network state information (e.g., current network user
identification) so
> that
> > rules can be written to use additional dynamic
stateinformation in
> applying the
> > policy rules.
>
> I agree  with the first sentence. The second is a vague
non-sequitur. Is it
> intended to give us a mandate to *demand* or to *suggest* ? For
the latter
> we don't need a mandate. We *may* always ask another working
group if
> they'll do so and so for us. If the former, we pretty much
allow ourselves
> to improve any protocol in IETF. The example given (current
network user
> identification) is a part of the PDP decision making process.
The PDP has to
> bind policy categories -- user name, application name,
interface type,
> whatever -- with dynamic information -- user name to source
address, for
> instance. Then it may instruct the PEP to carry out the right
operations on
> the packet stream. If my understanding is in keeping with Lee's
explanation
> below (which I believe it is), then the protocol enhancements
that could
> potentially be requested are a very large set. We could impact
protocols
> between a route server--network devices, Kerberos
authentication, Radius
> server, billing servers, etc.
>
> I would plead that we postpone this mandate till we have a
clear idea of
> which protocol/protocols will be impacted, what exactly we
need, and how we
> intend to achieve it. In other words, I request that we clearly
state that
> we will not be involved with binding of policy categories
(conditions) to
> header information as yet, and just drop the second sentence.
Thanks for
> your patience,
>
> raju
>
>
> >
> > Some supporting discussion for item #7 (from Lee):
> >
> > We need to be clearer about the difference between the
> > interpretation of policy rules and the data path execution
logic
> > associated with the policy rules.
> > Policy rules specify the *configuration* of the network that
is
> > required to meet
> > SLOs in a device-independent way.   Policy, then, is the
source
> > language that is compiled (by the PDP) into a target language
> > specification (device
> > configuration) of the logic needed to implement those
policies on
> > a particular
> > platform. I suspect we all agree, more or less, with these
> > statements.
> >
> > Where clarity and consensus are replaced by confusion and
> > disagreement, I think, is that some of the data on which the
> > policy target language description
> > operates is not currently available to the PEPs using
existing
> > protocols. Therefore, I think, people may be tempted to think
of
> > dynamically changing the
> > existing configuration parameters to make up for this missing
> > state information and, thus, conclude that they are
implementing
> > "dynamic policy" rather than providing data on which the
policy
> > can operate through the same interfaces that the policy
itself is
> > input into the target system.
> >
> > The last point to make now is that providing this state data
as
> > configuration
> > changes may be appealing (and may even be necessary in the
> > short-term) but it may be asking for trouble.  Whenever there
are
> > two paths along which information about a flow and the flow
> > itself may traverse, there are potential race conditions.  To
> > solve these race conditions, one must either have a
> > synchronization mechanism or must be able to tolerate loose
> > consistency.  I'll
> > assert without defence that policies can live with loose
> > consistency (hence LDAP is an adequate choice for
communicating
> > them) but network state information must be consistent in
order
> > for the network to behave in a consistent and predictable
> > fashion.  Therefore, inline communication of applicable state
> > changes will normally be the better technical answer; i.e.,
> > protocols need to be enhanced to synchronize awareness of
events
> > that are input to policy enforcement.
> >
> >
>
>



From majordomo@raleigh.ibm.com  Mon Jun  7 16:31:00 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA00457
	for <policy-archive@odin.ietf.org>; Mon, 7 Jun 1999 16:30:59 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id QAA25378;
	Mon, 7 Jun 1999 16:30:23 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id QAA35162;
	Mon, 7 Jun 1999 16:30:19 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA37324; Mon, 7 Jun 1999 16:16:36 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA41156; Mon, 7 Jun 1999 16:16:33 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id QAA31756
	for <policy@raleigh.ibm.com>; Mon, 7 Jun 1999 16:16:37 -0400
Received: from diablo.cisco.com (diablo.cisco.com [171.68.224.210])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id QAA24370
	for <policy@raleigh.ibm.com>; Mon, 7 Jun 1999 16:16:31 -0400
Received: from jschnizl-pc.cisco.com (jschnizl-isdn.cisco.com [171.70.238.115]) by diablo.cisco.com (8.8.6 (PHNE_14041)/CISCO.SERVER.1.2) with SMTP id NAA00120 for <policy@raleigh.ibm.com>; Mon, 7 Jun 1999 13:15:58 -0700 (PDT)
Message-Id: <4.1.19990607160537.00a7b2e0@diablo.cisco.com>
X-Sender: jschnizl@diablo.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.1 
Date: Mon, 07 Jun 1999 16:07:45 -0400
To: <policy@raleigh.ibm.com>
From: John Schnizlein <jschnizl@cisco.com>
Subject: Fwd: Re: IP Bandwidth Management
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: John Schnizlein <jschnizl@cisco.com>

I don't know how this thread got separated from the list.
Forwarding it back in hopes it happens here instead of diff-serv

}From: Keith McCloghrie <kzm@cisco.com>
}Subject: Re: IP Bandwidth Management
}To: herzog@iphighway.com (Shai Herzog)
}Date: Mon, 7 Jun 1999 13:04:08 -0700 (PDT)
}Cc: jschnizl@cisco.com, andrew@extremenetworks.com,
}        diffserv@external.cisco.com, rsvp-test@ISI.EDU, 
}        sob@newdev.harvard.edu, hadi@americasm01.nt.com
}X-Mailer: ELM [version 2.4 PL25]
}
}> >>The Policy WG doesn't do protocols and I think RAP isn't about
}> >>policy schema, but only the protocol+PIB.
}> 
}> There is the "PIB" mechanism (which belongs in the protocol, since it
}> is tied to a lot of transactional aspects), and the actual PIB details,
}> which should belong with each WG interested in using low level policy.
}> (In the same way MIBs are produced by each WG, for its own needs).
}
}The MIB analogy may or may not be appropriate, but if it is, it points
}in the other direction to what you suggest: the first MIBs were
}produced by the same WG as produced the protocol (the SNMP WG).  It
}was only later (when both the protocol and the SMI were stable) that
}each WG started doing its own MIBs.
}
}Keith.



From majordomo@raleigh.ibm.com  Mon Jun  7 16:31:20 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA00538
	for <policy-archive@odin.ietf.org>; Mon, 7 Jun 1999 16:31:19 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id QAA25624;
	Mon, 7 Jun 1999 16:30:31 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id QAA26424;
	Mon, 7 Jun 1999 16:30:18 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA32882; Mon, 7 Jun 1999 16:15:34 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA31850; Mon, 7 Jun 1999 16:15:32 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id QAA30052;
	Mon, 7 Jun 1999 16:15:33 -0400
Received: from bmailnj.iphighway.com ([209.3.6.76])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id QAA20190;
	Mon, 7 Jun 1999 16:15:29 -0400
Received: from shai-notebook (SHAI_NOTEBOOK [209.3.6.99]) by bmailnj.iphighway.com with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2448.0)
	id KN70SR3F; Mon, 7 Jun 1999 16:10:41 -0400
Message-Id: <4.1.19990607161102.04be14c0@bmailnj.iphighway.com>
X-Sender: shai_pop@bmailnj.iphighway.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.1 
Date: Mon, 07 Jun 1999 16:14:08 -0400
To: "Raju Rajan" <rajan@research.att.com>,
        "John Strassner" <jstrassn@cisco.com>, <policy@raleigh.ibm.com>
From: Shai Herzog <herzog@iphighway.com>
Subject: Re: Draft charter agreed to at the Interim Meeting
Cc: <fred@cisco.com>, <johns@cisco.com>, <ellesson@raleigh.ibm.com>,
        <wijnen@vnet.ibm.com>, <randy@psg.com>
In-Reply-To: <004a01beb11b$e0810de0$8c1bcf87@research.att.com>
References: <00f601beaef0$40bf1040$826c45ab@cisco.com>
 <4.1.19990607150621.04e25100@bmailnj.iphighway.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Shai Herzog <herzog@iphighway.com>

It was my impression that the charter preceeds the development
of solutions, so by definition, it includes things we THINK we know
what to do with, but haven't done or fully understand yet.

However, I think the charter should describe a problem space
and not a solution. Solutions come later.

Shai

At 03:28 PM 6/7/99, Raju Rajan wrote:
>I'm saying that we should not put anything into the charter till we
>understand what it entails. Otherwise we'll have a second PFDL disaster.
>
>raju
>
>----- Original Message -----
>From: Shai Herzog <herzog@iphighway.com>
>To: Raju Rajan <rajan@research.att.com>; John Strassner
><jstrassn@cisco.com>; <policy@raleigh.ibm.com>
>Cc: <fred@cisco.com>; <johns@cisco.com>; <ellesson@raleigh.ibm.com>;
><wijnen@vnet.ibm.com>; <randy@psg.com>
>Sent: Monday, June 07, 1999 3:11 PM
>Subject: Re: Draft charter agreed to at the Interim Meeting
>
>
>> At 10:50 AM 6/7/99, Raju Rajan wrote:
>>
>> >I would plead that we postpone this mandate till we have a clear idea of
>> >which protocol/protocols will be impacted, what exactly we need, and how
>we
>> >intend to achieve it. In other words, I request that we clearly state
>that
>> >we will not be involved with binding of policy categories (conditions) to
>> >header information as yet, and just drop the second sentence.
>>
>> I am not sure if I misunderstand, or disagree with you, but let me
>> try to clarify it.
>>
>> Are you saying that the WG should work on a schema that includes
>> dynamic binding (e.g., user name, path info, etc), but shouldn't
>> worry about how this is done (COPS, etc.) for now? OR, are you saying
>> that we shouldn't allow dynamic binding criteria into the schema? (which
>> I don't think you meant to say).
>>
>> Shai
>>
>> >Thanks for your patience,
>>
>> What patience???? ;-)
>>
>> Shai
>>
>>
>> __________________________________________________________
>> Shai Herzog                  Founder & CTO, IPHighway Inc.
>> Cell: (408) 390-3045       http://www.iphighway.com/herzog
>> Tel : (201) 585-0800                  Fax : (212) 656-1006
>>


__________________________________________________________
Shai Herzog                  Founder & CTO, IPHighway Inc.         
Cell: (408) 390-3045       http://www.iphighway.com/herzog           
Tel : (201) 585-0800                  Fax : (212) 656-1006 


From majordomo@raleigh.ibm.com  Mon Jun  7 17:38:41 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA01575
	for <policy-archive@odin.ietf.org>; Mon, 7 Jun 1999 17:38:41 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA39344;
	Mon, 7 Jun 1999 17:30:50 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id RAA33166;
	Mon, 7 Jun 1999 17:30:50 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA49698; Mon, 7 Jun 1999 17:17:16 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA41754; Mon, 7 Jun 1999 17:17:14 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id RAA32470
	for <policy@raleigh.ibm.com>; Mon, 7 Jun 1999 17:17:17 -0400
Received: from rerun.lucentctc.com (rerun.lucentctc.com [199.93.237.2])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA21296
	for <policy@raleigh.ibm.com>; Mon, 7 Jun 1999 17:17:13 -0400
Received: by rerun.lucentctc.com with Internet Mail Service (5.5.2448.0)
	id <L71M9SY6>; Mon, 7 Jun 1999 17:16:43 -0400
Message-Id: <75ADD7496F0BD211ADC000104B8846CFC9B1DA@rerun.lucentctc.com>
From: "Stevens, Mark" <markstevens@lucent.com>
To: "'Andrew Smith'" <andrew@extremenetworks.com>,
        "'Raju Rajan'"
	 <rajan@research.att.com>, policy@raleigh.ibm.com
Cc: fred@cisco.com, wijnen@vnet.ibm.com, randy@psg.com
Subject: RE: Draft charter agreed to at the Interim Meeting
Date: Mon, 7 Jun 1999 17:16:42 -0400 
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Stevens, Mark" <markstevens@lucent.com>

I think that I may be inclined to agree with Andrew, but I would offer that
item number 7 seems to say,

Let's not bite off more than we can chew.
Let's define an vendor-independent way to express packet filters in a
directory.

Does anyone else see this subtext within item #7?

Below, I include the original item and what I see when a read it.

Item 7 from the charter:
Because policy is a device-independent specification of configuration
information, the policy information model and schema provide for
infrequently changing declarative information that reflect network rules
rather than network state. The Policy Framework WG may request that other
IETF Working Groups provide enhancements to protocols that they develop  to
communicate network state information (e.g., current network user
identification) so that rules can be written to use additional dynamic
state information in applying the policy rules.  

Item 7 seems to break down into:
 - Policies are expressions of device-independent
   configuration information.

 - Our existing information model and schema
   accommodates the expression of policy in terms
   of static rules.

 - Sophisticated rules that consider the state of
   the network should not be expressed in the
   existing schema unless or until protocols are
   modified to contain state information needed 
   to make the decision.

Am I reading too much, or perhaps too little into this item?

-Mark

> -----Original Message-----
> From:	Andrew Smith [SMTP:andrew@extremenetworks.com]
> Sent:	Monday, June 07, 1999 3:10 PM
> To:	'Raju Rajan'; policy@raleigh.ibm.com
> Cc:	fred@cisco.com; wijnen@vnet.ibm.com; randy@psg.com
> Subject:	RE: Draft charter agreed to at the Interim Meeting
> 
> 
> I don't see that the proposed new charter item 7 adds much useful
> guidance.
> I also do not think that we need any explicit non-goals regarding "binding
> of policy categories (conditions) to header information". But it would be
> good to capture some of Lee's discussion/argument in the framework
> document.
> 
> Andrew
> 
> ****************************************************************
> Andrew Smith                              tel: +1 (408) 579-2821
> Extreme Networks                          fax: +1 (408) 579-3000
> 3585 Monroe St.                   http://www.extremenetworks.com
> Santa Clara CA 95051-1450        em:  andrew@extremenetworks.com
> ****************************************************************
> 
> 
> > -----Original Message-----
> > From: Raju Rajan [mailto:rajan@research.att.com]
> > Sent: Monday, June 07, 1999 7:51 AM
> > To: John Strassner; policy@raleigh.ibm.com
> > Cc: fred@cisco.com; johns@cisco.com; ellesson@raleigh.ibm.com;
> > wijnen@vnet.ibm.com; randy@psg.com
> > Subject: Re: Draft charter agreed to at the Interim Meeting
> > 
> > 
> > Hi,
> > 
> > I had a comment about the newly added item on the policy charter:
> > 
> > > // NEW ITEM - see discussion for this new item at the end
> > > 7. Because policy is a device-independent specification of 
> > configuration
> > > information, the policy information model and schema provide for
> > infrequently
> > > changing declarative information that reflect network rules 
> > rather than
> > network
> > > state.  The Policy Framework WG may request that other IETF Working
> > > Groups provide enhancements to protocols that they develop 
> > to communicate
> > > network state information (e.g., current network user 
> > identification) so
> > that
> > > rules can be written to use additional dynamic stateinformation in
> > applying the
> > > policy rules.
> > 
> > I agree  with the first sentence. The second is a vague 
> > non-sequitur. Is it
> > intended to give us a mandate to *demand* or to *suggest* ? 
> > For the latter
> > we don't need a mandate. We *may* always ask another working group if
> > they'll do so and so for us. If the former, we pretty much 
> > allow ourselves
> > to improve any protocol in IETF. The example given (current 
> > network user
> > identification) is a part of the PDP decision making process. 
> > The PDP has to
> > bind policy categories -- user name, application name, interface type,
> > whatever -- with dynamic information -- user name to source 
> > address, for
> > instance. Then it may instruct the PEP to carry out the right 
> > operations on
> > the packet stream. If my understanding is in keeping with 
> > Lee's explanation
> > below (which I believe it is), then the protocol enhancements 
> > that could
> > potentially be requested are a very large set. We could 
> > impact protocols
> > between a route server--network devices, Kerberos 
> > authentication, Radius
> > server, billing servers, etc.
> > 
> > I would plead that we postpone this mandate till we have a 
> > clear idea of
> > which protocol/protocols will be impacted, what exactly we 
> > need, and how we
> > intend to achieve it. In other words, I request that we 
> > clearly state that
> > we will not be involved with binding of policy categories 
> > (conditions) to
> > header information as yet, and just drop the second sentence. 
> > Thanks for
> > your patience,
> > 
> > raju
> > 
> > 
> > >
> > > Some supporting discussion for item #7 (from Lee):
> > >
> > > We need to be clearer about the difference between the
> > > interpretation of policy rules and the data path execution logic
> > > associated with the policy rules.
> > > Policy rules specify the *configuration* of the network that is
> > > required to meet
> > > SLOs in a device-independent way.   Policy, then, is the source
> > > language that is compiled (by the PDP) into a target language
> > > specification (device
> > > configuration) of the logic needed to implement those policies on
> > > a particular
> > > platform. I suspect we all agree, more or less, with these
> > > statements.
> > >
> > > Where clarity and consensus are replaced by confusion and
> > > disagreement, I think, is that some of the data on which the
> > > policy target language description
> > > operates is not currently available to the PEPs using existing
> > > protocols. Therefore, I think, people may be tempted to think of
> > > dynamically changing the
> > > existing configuration parameters to make up for this missing
> > > state information and, thus, conclude that they are implementing
> > > "dynamic policy" rather than providing data on which the policy
> > > can operate through the same interfaces that the policy itself is
> > > input into the target system.
> > >
> > > The last point to make now is that providing this state data as
> > > configuration
> > > changes may be appealing (and may even be necessary in the
> > > short-term) but it may be asking for trouble.  Whenever there are
> > > two paths along which information about a flow and the flow
> > > itself may traverse, there are potential race conditions.  To
> > > solve these race conditions, one must either have a
> > > synchronization mechanism or must be able to tolerate loose
> > > consistency.  I'll
> > > assert without defence that policies can live with loose
> > > consistency (hence LDAP is an adequate choice for communicating
> > > them) but network state information must be consistent in order
> > > for the network to behave in a consistent and predictable
> > > fashion.  Therefore, inline communication of applicable state
> > > changes will normally be the better technical answer; i.e.,
> > > protocols need to be enhanced to synchronize awareness of events
> > > that are input to policy enforcement.
> > >
> > >
> > 


From majordomo@raleigh.ibm.com  Mon Jun  7 18:01:13 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA01911
	for <policy-archive@odin.ietf.org>; Mon, 7 Jun 1999 18:01:12 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id SAA27620;
	Mon, 7 Jun 1999 18:00:19 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id SAA25578;
	Mon, 7 Jun 1999 18:00:19 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA42854; Mon, 7 Jun 1999 17:45:33 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA41822; Mon, 7 Jun 1999 17:45:30 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id RAA25728
	for <policy@raleigh.ibm.com>; Mon, 7 Jun 1999 17:45:33 -0400
Received: from sol.extremenetworks.com (sol.extremenetworks.com [216.52.8.2])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA34258
	for <policy@raleigh.ibm.com>; Mon, 7 Jun 1999 17:45:29 -0400
Received: by ftp.extremenetworks.com with Internet Mail Service (5.5.2448.0)
	id <L8P72WHB>; Mon, 7 Jun 1999 14:45:01 -0700
Message-Id: <D0805D3B448BD211A7990008C7B18130142DCC@ftp.extremenetworks.com>
From: Andrew Smith <andrew@extremenetworks.com>
To: "'Stevens, Mark'" <markstevens@lucent.com>, policy@raleigh.ibm.com
Cc: fred@cisco.com, wijnen@vnet.ibm.com, randy@psg.com
Subject: RE: Draft charter agreed to at the Interim Meeting
Date: Mon, 7 Jun 1999 14:45:00 -0700 
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Andrew Smith <andrew@extremenetworks.com>

Mark,

I think that is certainly a deliverable from this WG: but I would consider
it to be covered already by both the original and new charters' "...
provisioned and signaled QoS ... DiffServ, IntServ, RAP, and
ISSLL working groups ... to develop schemata for both IPv4 and IPv6 ...". I
guess it does not hurt to make it more explicit, although we ought to limit
the scope to include only those packet filters necessary for
IntServ/ISSLL/DiffServ sorts of things.

Andrew

> -----Original Message-----
> From: Stevens, Mark [mailto:markstevens@lucent.com]
> Sent: Monday, June 07, 1999 2:17 PM
> To: 'Andrew Smith'; 'Raju Rajan'; policy@raleigh.ibm.com
> Cc: fred@cisco.com; wijnen@vnet.ibm.com; randy@psg.com
> Subject: RE: Draft charter agreed to at the Interim Meeting
> 
...
> Let's define an vendor-independent way to express packet filters in a
> directory.


From majordomo@raleigh.ibm.com  Thu Jun 10 08:41:39 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id IAA27102
	for <policy-archive@odin.ietf.org>; Thu, 10 Jun 1999 08:41:39 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id IAA20298;
	Thu, 10 Jun 1999 08:40:31 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id IAA33168;
	Thu, 10 Jun 1999 08:40:27 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA30754; Thu, 10 Jun 1999 08:16:44 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA39350; Thu, 10 Jun 1999 08:16:20 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id IAA33164
	for <policy@raleigh.ibm.com>; Thu, 10 Jun 1999 08:15:52 -0400
Received: from duettech.com (snt.snt.com [209.24.228.10])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id IAA30424
	for <policy@raleigh.ibm.com>; Thu, 10 Jun 1999 08:15:44 -0400
Received: (from root@localhost)
	by duettech.com (8.8.8/8.8.8) id FAA27323
	for <policy@raleigh.ibm.com>; Thu, 10 Jun 1999 05:15:44 -0700 (PDT)
Received: from snt002(199.172.181.2) by duet.duettech.com via smap (V2.1)
	id xma027312; Thu, 10 Jun 99 05:15:13 -0700
Received: from duettech.com ([199.172.183.52])
	by casnt.ca.duettech.com (8.9.0/8.9.0) with SMTP id FAA04002
	for <policy@raleigh.ibm.com>; Thu, 10 Jun 1999 05:12:01 -0700 (PDT)
Received: from surat.snt.com by  duettech.com  (4.1/SMI-4.1-S&T)
	id AA02154; Thu, 10 Jun 99 17:44:38 IST
Received: from duettech.com (benz) by surat.snt.com (4.1/SMI-4.1-iS&T)
        id AA17558; Thu, 10 Jun 99 17:40:27 IST
Message-Id: <375FAB3B.2DE7F3C0@duettech.com>
Date: Thu, 10 Jun 1999 17:40:35 +0530
From: Atul Jain <atulj@duettech.com>
Organization: Duet Technologies
X-Mailer: Mozilla 4.5 [en] (WinNT; I)
X-Accept-Language: en
Mime-Version: 1.0
To: policy@raleigh.ibm.com
Cc: Atul Jain <atulj@duettech.com>
Subject: Policy vs. service?
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Atul Jain <atulj@duettech.com>
Content-Transfer-Encoding: 7bit

Hi,

I have confusion about "Policy vs Service" as defined in section 4.7 of
"draft-strassner-policy-terms-01.txt".
It says
"A service is the expression of a relationship between a set of objects.

A policy is a statement about a set of relationships between objects
that provide a particular service.  "

But this definition is not very clear. To me even a "PolicyAction" (as
defined in draft-rajan-policy-qosschema-01.txt) looks like a "service"
definition.
Can somebody elaborate on it with the help of some examples (or give
pointers to documents/resource) as to how a service can be defined and
how does it relate to the policies.

Thanks
Atul

--
*****************************************
Atul Jain
Duet Technologies(I) Pvt. Ltd.
SDF #B1, NEPZ, Noida (U.P.) - 201305
India
Phone(Office): +(91)-(1191)-567001 to 567005
Phone(Home):   +(91)-(1191)-541919
mailto:atulj@duettech.com
*****************************************




From majordomo@raleigh.ibm.com  Thu Jun 10 10:09:49 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA28369
	for <policy-archive@odin.ietf.org>; Thu, 10 Jun 1999 10:09:47 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id KAA28412;
	Thu, 10 Jun 1999 10:08:53 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id KAA34826;
	Thu, 10 Jun 1999 10:08:44 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA40876; Thu, 10 Jun 1999 09:58:00 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA40868; Thu, 10 Jun 1999 09:57:58 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id JAA31924;
	Thu, 10 Jun 1999 09:58:00 -0400
Received: from nsm-mail2.cisco.com (nsm-mail2.cisco.com [171.71.236.25])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id JAA23452;
	Thu, 10 Jun 1999 09:57:57 -0400
Received: from jstrassnpc.cisco.com ([171.69.108.130])
	by nsm-mail2.cisco.com (8.8.8-Cisco List Logging/8.8.8) with SMTP id GAA25529;
	Thu, 10 Jun 1999 06:57:19 -0700 (PDT)
Message-Id: <000701beb349$48bec940$826c45ab@cisco.com>
From: "John Strassner" <jstrassn@cisco.com>
To: "Jean Christophe Martin" <jcmartin@france.sun.com>,
        "Raju Rajan" <rajan@research.att.com>
Cc: "John Strassner" <jstrassn@cisco.com>, <policy@raleigh.ibm.com>,
        <fred@cisco.com>, <johns@cisco.com>, <ellesson@raleigh.ibm.com>,
        <wijnen@vnet.ibm.com>, <randy@psg.com>
References: <00f601beaef0$40bf1040$826c45ab@cisco.com> <019001beb0f5$19f44eb0$8c1bcf87@research.att.com> <375E730A.E2B8F432@france.sun.com>
Subject: Re: Draft charter agreed to at the Interim Meeting
Date: Wed, 9 Jun 1999 23:14:04 -0700
Organization: Cisco Systems
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "John Strassner" <jstrassn@cisco.com>
Content-Transfer-Encoding: 7bit

I don't agree.

This working group is NOT solely concerned with producing an
optimized schema that can be implemented in a directory that uses
LDAPv3 as its access protocol. This working group IS concerned
with developing an information model, from which specific
schemata aimed at specific implementations can be derived. The
fact that we are spending a lot of time on a directory
implementation does not mean that the information model should
ignore dynamic information.

That being said, I think that we should walk before we run.
Certainly, focusing on non-dynamic information is simpler,
easier, and will lead to tangible results more quickly. I'm just
objecting to the black-and-white statement of "no, we should not
care about how the dynamic information (or network state) is
updated.

Could you please explain what you mean to being "able to bind a
policy to an object partially defined"? I can't parse this. Does
this mean I'm binding a policy to an object which is then going
to change? If it changes, how do I know that the binding is still
valid?

Finally, the examples you gave are LDAP-specific. Remember, we
have two work items for the core and QoS schemata: an information
model and an LDAP mapping of the information model. So yes, we
should consider all those, but we should ensure that the
information model is complete before we go off into a specific
implementation. Otherwise, we may never be able to get back to
the information model cleanly, which will then inhibit other
mappings (e.g., to an RDBMS).

John

----- Original Message -----
From: Jean Christophe Martin <jcmartin@france.sun.com>
To: Raju Rajan <rajan@research.att.com>
Cc: John Strassner <jstrassn@cisco.com>;
<policy@raleigh.ibm.com>; <fred@cisco.com>; <johns@cisco.com>;
<ellesson@raleigh.ibm.com>; <wijnen@vnet.ibm.com>;
<randy@psg.com>
Sent: Wednesday, June 09, 1999 6:58 AM
Subject: Re: Draft charter agreed to at the Interim Meeting


> Raju
>
> I agree with you that this group should not care about how the
> dynamic information (or network state) is updated, however,
> I think that this is important to be able to bind a policy
> to an object partially defined because some of its parameters
> are dynamically allocated.
> This group should define if the binding is done by DN pointers,
> by specific DIT, by specific attributes defined as placeholder
> for dynamic valus, or placeholder for pointers to the dynamic
> attributes.
>
> JC.
>
> > > // NEW ITEM - see discussion for this new item at the end
> > > 7. Because policy is a device-independent specification of
configuration
> > > information, the policy information model and schema
provide for
> > infrequently
> > > changing declarative information that reflect network rules
rather than
> > network
> > > state.  The Policy Framework WG may request that other IETF
Working
> > > Groups provide enhancements to protocols that they develop
to communicate
> > > network state information (e.g., current network user
identification) so
> > that
> > > rules can be written to use additional dynamic
stateinformation in
> > applying the
> > > policy rules.
> >
> > I agree  with the first sentence. The second is a vague
non-sequitur. Is it
> > intended to give us a mandate to *demand* or to *suggest* ?
For the latter
> > we don't need a mandate. We *may* always ask another working
group if
> > they'll do so and so for us. If the former, we pretty much
allow ourselves
> > to improve any protocol in IETF. The example given (current
network user
> > identification) is a part of the PDP decision making process.
The PDP has to
> > bind policy categories -- user name, application name,
interface type,
> > whatever -- with dynamic information -- user name to source
address, for
> > instance. Then it may instruct the PEP to carry out the right
operations on
> > the packet stream. If my understanding is in keeping with
Lee's explanation
> > below (which I believe it is), then the protocol enhancements
that could
> > potentially be requested are a very large set. We could
impact protocols
> > between a route server--network devices, Kerberos
authentication, Radius
> > server, billing servers, etc.
> >
> > I would plead that we postpone this mandate till we have a
clear idea of
> > which protocol/protocols will be impacted, what exactly we
need, and how we
> > intend to achieve it. In other words, I request that we
clearly state that
> > we will not be involved with binding of policy categories
(conditions) to
> > header information as yet, and just drop the second sentence.
Thanks for
> > your patience,
> >
> > raju
> >
> > >
> > > Some supporting discussion for item #7 (from Lee):
> > >
> > > We need to be clearer about the difference between the
> > > interpretation of policy rules and the data path execution
logic
> > > associated with the policy rules.
> > > Policy rules specify the *configuration* of the network
that is
> > > required to meet
> > > SLOs in a device-independent way.   Policy, then, is the
source
> > > language that is compiled (by the PDP) into a target
language
> > > specification (device
> > > configuration) of the logic needed to implement those
policies on
> > > a particular
> > > platform. I suspect we all agree, more or less, with these
> > > statements.
> > >
> > > Where clarity and consensus are replaced by confusion and
> > > disagreement, I think, is that some of the data on which
the
> > > policy target language description
> > > operates is not currently available to the PEPs using
existing
> > > protocols. Therefore, I think, people may be tempted to
think of
> > > dynamically changing the
> > > existing configuration parameters to make up for this
missing
> > > state information and, thus, conclude that they are
implementing
> > > "dynamic policy" rather than providing data on which the
policy
> > > can operate through the same interfaces that the policy
itself is
> > > input into the target system.
> > >
> > > The last point to make now is that providing this state
data as
> > > configuration
> > > changes may be appealing (and may even be necessary in the
> > > short-term) but it may be asking for trouble.  Whenever
there are
> > > two paths along which information about a flow and the flow
> > > itself may traverse, there are potential race conditions.
To
> > > solve these race conditions, one must either have a
> > > synchronization mechanism or must be able to tolerate loose
> > > consistency.  I'll
> > > assert without defence that policies can live with loose
> > > consistency (hence LDAP is an adequate choice for
communicating
> > > them) but network state information must be consistent in
order
> > > for the network to behave in a consistent and predictable
> > > fashion.  Therefore, inline communication of applicable
state
> > > changes will normally be the better technical answer; i.e.,
> > > protocols need to be enhanced to synchronize awareness of
events
> > > that are input to policy enforcement.
> > >
> > >
>



From majordomo@raleigh.ibm.com  Fri Jun 11 17:26:17 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA07610
	for <policy-archive@odin.ietf.org>; Fri, 11 Jun 1999 17:26:16 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA06752;
	Fri, 11 Jun 1999 17:26:08 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id RAA26558;
	Fri, 11 Jun 1999 17:25:59 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA51056; Fri, 11 Jun 1999 17:13:56 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA40296; Fri, 11 Jun 1999 17:13:53 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id RAA26806
	for <policy@raleigh.ibm.com>; Fri, 11 Jun 1999 17:13:54 -0400
Received: from mail-blue.research.att.com (mail-blue.research.att.com [135.207.30.102])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA27636
	for <policy@raleigh.ibm.com>; Fri, 11 Jun 1999 17:13:50 -0400
Received: from bigmail.research.att.com (bigmail.research.att.com [135.207.30.101])
	by mail-blue.research.att.com (Postfix) with ESMTP
	id AA8704CE03; Fri, 11 Jun 1999 17:13:52 -0400 (EDT)
Received: from pcraju (pc-raju.research.att.com [135.207.27.140])
	by bigmail.research.att.com (8.8.8/8.8.8) with SMTP id RAA02078;
	Fri, 11 Jun 1999 17:13:52 -0400 (EDT)
Message-Id: <003801beb450$3097ca30$8c1bcf87@research.att.com>
From: "Raju Rajan" <rajan@research.att.com>
To: "Atul Jain" <atulj@duettech.com>, <policy@raleigh.ibm.com>
Cc: "Atul Jain" <atulj@duettech.com>
References: <375FAB3B.2DE7F3C0@duettech.com>
Subject: Re: Policy vs. service?
Date: Fri, 11 Jun 1999 17:20:11 -0400
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Raju Rajan" <rajan@research.att.com>
Content-Transfer-Encoding: 7bit

Hi Atul,

 The word "service" has been used somewhat differently at the two companies
that I have worked for; at IBM I have heard people refer to IPSec and QoS as
services, while at AT&T nothing short of VPN seems to be a service. So I
guess a service is whatever your company is selling :) But let me make some
observations on usage:

1) Service is an abstraction that is closer to "what the customer pays for".
So it is cobbled together from a variety of different protocols and
mechanisms; for instance, I can sell a QoS service that is assembled from
DiffServ, Frame Relay, RSVP etc. in different parts of the network.

2) Policy is a lower layer of abstraction, and is interior to your network.
I cannot have the same policy or policy rule for FR and RSVP. It is aimed at
a particular PEP or set of PEPs playing a particular role.

My 2 bits,

raju
----- Original Message -----
From: Atul Jain <atulj@duettech.com>
To: <policy@raleigh.ibm.com>
Cc: Atul Jain <atulj@duettech.com>
Sent: Thursday, June 10, 1999 8:10 AM
Subject: Policy vs. service?


> Hi,
>
> I have confusion about "Policy vs Service" as defined in section 4.7 of
> "draft-strassner-policy-terms-01.txt".
> It says
> "A service is the expression of a relationship between a set of objects.
>
> A policy is a statement about a set of relationships between objects
> that provide a particular service.  "
>
> But this definition is not very clear. To me even a "PolicyAction" (as
> defined in draft-rajan-policy-qosschema-01.txt) looks like a "service"
> definition.
> Can somebody elaborate on it with the help of some examples (or give
> pointers to documents/resource) as to how a service can be defined and
> how does it relate to the policies.
>
> Thanks
> Atul
>
> --
> *****************************************
> Atul Jain
> Duet Technologies(I) Pvt. Ltd.
> SDF #B1, NEPZ, Noida (U.P.) - 201305
> India
> Phone(Office): +(91)-(1191)-567001 to 567005
> Phone(Home):   +(91)-(1191)-541919
> mailto:atulj@duettech.com
> *****************************************
>
>
>



From majordomo@raleigh.ibm.com  Sun Jun 13 15:25:10 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA20880
	for <policy-archive@odin.ietf.org>; Sun, 13 Jun 1999 15:25:10 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA33786;
	Sun, 13 Jun 1999 15:24:50 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id PAA31510;
	Sun, 13 Jun 1999 15:24:36 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA41236; Sun, 13 Jun 1999 15:17:10 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA40456; Sun, 13 Jun 1999 15:17:08 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id PAA29260
	for <policy@raleigh.ibm.com>; Sun, 13 Jun 1999 15:17:05 -0400
Received: from bmailnj.iphighway.com ([209.3.6.76])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA41974
	for <policy@raleigh.ibm.com>; Sun, 13 Jun 1999 15:17:02 -0400
Received: from shai-notebook (slip-32-100-251-253.ny.us.IBM.NET [32.100.251.253]) by bmailnj.iphighway.com with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2448.0)
	id KN70SS8G; Sun, 13 Jun 1999 15:11:47 -0400
Message-Id: <4.1.19990613151123.044a76b0@bmailnj.iphighway.com>
X-Sender: herzog@bmailnj.iphighway.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.1 
Date: Sun, 13 Jun 1999 15:12:48 -0400
To: "Raju Rajan" <rajan@research.att.com>, "Atul Jain" <atulj@duettech.com>,
        <policy@raleigh.ibm.com>
From: Shai Herzog <herzog@iphighway.com>
Subject: Re: Policy vs. service?
Cc: "Atul Jain" <atulj@duettech.com>
In-Reply-To: <003801beb450$3097ca30$8c1bcf87@research.att.com>
References: <375FAB3B.2DE7F3C0@duettech.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Shai Herzog <herzog@iphighway.com>

Raju,

I tend to agree with your definition, and I think it is well put.
I wish the terminology ducument would replace the stuff that is there
with a clearer definition (following your explanation).

Shai

At 05:20 PM 6/11/99, Raju Rajan wrote:
>Hi Atul,
>
> The word "service" has been used somewhat differently at the two companies
>that I have worked for; at IBM I have heard people refer to IPSec and QoS as
>services, while at AT&T nothing short of VPN seems to be a service. So I
>guess a service is whatever your company is selling :) But let me make some
>observations on usage:
>
>1) Service is an abstraction that is closer to "what the customer pays for".
>So it is cobbled together from a variety of different protocols and
>mechanisms; for instance, I can sell a QoS service that is assembled from
>DiffServ, Frame Relay, RSVP etc. in different parts of the network.
>
>2) Policy is a lower layer of abstraction, and is interior to your network.
>I cannot have the same policy or policy rule for FR and RSVP. It is aimed at
>a particular PEP or set of PEPs playing a particular role.
>
>My 2 bits,
>
>raju
>----- Original Message -----
>From: Atul Jain <atulj@duettech.com>
>To: <policy@raleigh.ibm.com>
>Cc: Atul Jain <atulj@duettech.com>
>Sent: Thursday, June 10, 1999 8:10 AM
>Subject: Policy vs. service?
>
>
>> Hi,
>>
>> I have confusion about "Policy vs Service" as defined in section 4.7 of
>> "draft-strassner-policy-terms-01.txt".
>> It says
>> "A service is the expression of a relationship between a set of objects.
>>
>> A policy is a statement about a set of relationships between objects
>> that provide a particular service.  "
>>
>> But this definition is not very clear. To me even a "PolicyAction" (as
>> defined in draft-rajan-policy-qosschema-01.txt) looks like a "service"
>> definition.
>> Can somebody elaborate on it with the help of some examples (or give
>> pointers to documents/resource) as to how a service can be defined and
>> how does it relate to the policies.
>>
>> Thanks
>> Atul
>>
>> --
>> *****************************************
>> Atul Jain
>> Duet Technologies(I) Pvt. Ltd.
>> SDF #B1, NEPZ, Noida (U.P.) - 201305
>> India
>> Phone(Office): +(91)-(1191)-567001 to 567005
>> Phone(Home):   +(91)-(1191)-541919
>> mailto:atulj@duettech.com
>> *****************************************
>>
>>
>>


__________________________________________________________
Shai Herzog                  Founder & CTO, IPHighway Inc.         
Cell: (408) 390-3045       http://www.iphighway.com/herzog           
Tel : (201) 585-0800                  Fax : (212) 656-1006 


From majordomo@raleigh.ibm.com  Sun Jun 13 15:47:45 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA20987
	for <policy-archive@odin.ietf.org>; Sun, 13 Jun 1999 15:47:44 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA28318;
	Sun, 13 Jun 1999 15:47:40 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id PAA33480;
	Sun, 13 Jun 1999 15:47:38 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA51534; Sun, 13 Jun 1999 15:39:55 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA35142; Sun, 13 Jun 1999 15:39:51 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id PAA07758
	for <policy@raleigh.ibm.com>; Sun, 13 Jun 1999 15:39:51 -0400
Received: from bmailnj.iphighway.com ([209.3.6.76])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA41534
	for <policy@raleigh.ibm.com>; Sun, 13 Jun 1999 15:39:50 -0400
Received: from shai-notebook (slip-32-100-112-90.ny.us.ibm.net [32.100.112.90]) by bmailnj.iphighway.com with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2448.0)
	id KN70SS82; Sun, 13 Jun 1999 15:34:34 -0400
Message-Id: <4.1.19990613152827.0444fc50@bmailnj.iphighway.com>
X-Sender: herzog@bmailnj.iphighway.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.1 
Date: Sun, 13 Jun 1999 15:35:49 -0400
To: "Durham, David" <david.durham@intel.com>,
        "'remoore@us.ibm.com'" <remoore@us.ibm.com>, policy@raleigh.ibm.com
From: Shai Herzog <herzog@iphighway.com>
Subject: RE: Transactional issues with the LDAP Core Schema
Cc: "Durham, David" <david.durham@intel.com>,
        "'Raju Rajan'" <rajan@research.att.com>
In-Reply-To: <4A043A1FE4B2D111AC3F00A0C96B513301B0F5A0@fmsmsx37.fm.intel
 .com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Shai Herzog <herzog@iphighway.com>

Dave and Bob,

Admiting to not being an LDAP/Directory guru, I'm wondering if we 
are doing the right thing by puting the burden of preventing half 
baked rules (groups) on the gui/tool. This opens up a lot of room 
for interpretation and attempt to close loopholes, which in turn 
lead to incompatible results.

Wouldn't a better approach would be to put a "semaphore lock" on
the root of the rule. Given the infrequent update, overhead is NOT
a problem, but it may solve virtually all the updating problems.

1. As long as the rule is being updated, it is locked, and PDPs
   should wait until the update is complete.

2. If multiple rules are updated as a transactional group, then
   all of them are locked, until they are all in, and then the
   lock is removed. This way, a keyword search would discover
   a few locked rules, and would cause the PDP to wait until
   they are cleared, and then re-issueing the keyword search
   to get all the rules.

What do you think?

Shai

At 01:36 PM 6/4/99, Durham, David wrote:
>Thanks Bob, this helps clear things up a bit with my second
>example. I see that by following a well-defined procedure 
>a policy editor tool can enter policy in a way that will 
>avoid many of the "half-baked" policy problems despite the lack 
>of transactions. This information should probably find its 
>way into the specification because policy editors that do 
>not follow the "correct" procedure will create problems.
>It would be even better if the LDAPized core schema could
>somehow force the correct procedure. 
>
>I still think the issue remains for policies that already 
>exist but are instead being edited/updated. Let me see if 
>I can conjure up an illustrative example for such a case...
>Say, policy "re-ordering" in the case rules are prioritied, 
>or redefining two groups already ANDed together within a 
>policy rule. Seems as if even more restrictions will have
>to be placed on the policy editor for cases as these where
>dependencies cross the boundaries of a single entry.
>
>Also, can you explain if there is a procedure for modifying 
>auxiliary attachments that act as labels (e.g. through
>the policyKeywords attribute) such that lack of transactional
>support is not a pitfall here?
>
>-Dave
>
>
>> -----Original Message-----
>> From:	remoore@us.ibm.com [SMTP:remoore@us.ibm.com]
>> Sent:	Friday, June 04, 1999 8:10 AM
>> To:	policy@raleigh.ibm.com
>> Cc:	Durham, David; 'Raju Rajan'
>> Subject:	Transactional issues with the LDAP Core Schema
>> 
>> 
>> 
>> Dave,
>> 
>> You wrote the following in a response to Raju:
>> 
>> [[Dave]]  Getting the full LDAP object is not the problem, it
>>      is the solution. My complaint is that when you go beyond a single
>>      LDAP entry to express a policy, then you are effectively opening
>>      up LDAP's non-transactional Achilles heel. Keeping all ANDed
>>      conditions within a single entry will help prevent clients from
>>      retrieving incomplete (=incorrect) policies. Attaching conditions
>>      as separate sibling entries to a policy rule implies that a client
>>      can retrieve a portion of the DIT that a console is in the process
>>      of editing. Rules need to be built from the conditions and actions
>>      up, or entries need to contain all attributes applicable to a rule,
>>      so that PDPs do not get incorrect policies. Basically, the core
>>      schema should be designed to take advantage of atomic transactions
>>      within an entry where LDAP does have limited transactional support.
>>      The DNF form of complex policy rules violates this principle as
>>      do simple policy rules (and correct me if I am wrong) where
>>      conditions and actions are attached as sibling entries.
>> 
>>      I have a host of other transactional issues with using auxiliary
>>      attachments as labels which will have to be edited across DNs all
>>      the while PDPs (and other consoles) are potentially querying
>>      (and editing) the DIT!
>> 
>> While I understand the general point about LDAP's transactional
>> weakness and the resulting exposure arising from the use of multiple
>> entries to store related pieces of information, I'm having a hard time
>> finding the exposures in the specific case of the Core Policy Schema
>> classes.
>> 
>> For example, there's no chance for a client to get a "half-baked"
>> policy rule if the entries that make up the rule are added to the
>> directory (or, if they're already there, updated to say the right
>> thing) in the following order:
>> 
>> 1. Entries like Lee's addressRange object, that are pointed to by
>>    policy condition and/or policy action objects.
>> 2. All policyCondition and policyAction entries that the policyRule
>>    will point to.
>> 3. The policyRule entry itself.
>> 
>> It makes no difference whether the rule combines the conditions in DNF
>> or CNF, since by the time the rule is there to be retrieved, all the
>> conditions are already present in the directory.
>> 
>> Ordinarily this sequence of updates isn't something that the directory
>> administrator will need to worry about:  the Policy Tool should take
>> care of it automatically.  The standard should probably contain an
>> implementation note saying that a Policy Tool needs to write rule-
>> related information to the directory in this order, to avoid the
>> transactional pitfalls that Dave is worried about.
>> 
>> Another set of semantic relationships (besides those between a rule
>> and its conditions and actions) in the Core Schema are those between
>> a policy group and the policy rules or policy groups that it
>> aggregates.  Once again I don't see how a client can be misled if the
>> order of creation (or update) for these entries is:
>> 
>> 1. All the policyRules or policyGroups to be aggregated.
>> 2. The policyGroup instance that aggregates them.
>> 
>> The jurisdiction associations, realized by attaching the policy rule
>> or policy group pointers to something other than policyGroups, will
>> work in the same way:  first create the rules / groups being pointed
>> to, then create the pointers to them.  A client may come to the
>> directory before this process has been completed, in which case it
>> will fail to find any of its policy rules.  But this is the same
>> thing that would happen if it came to the directory before any of
>> the updates related to its policy rules had been done.  There is, once
>> again, no point in time when a client might retrieve some of the rules
>> related to its jurisdiction, while believing that it had retrieved all
>> of them.
>> 
>> The one case that does strike me as problematic is where a client
>> is retrieving all the policy rules whose priorities need to be
>> compared against each other to decide what to do in a particular set
>> of circumstances.  If the client gets to these rules via a policyGroup,
>> then there shouldn't be a problem.  The client will know that it's in
>> the middle of retrieving a set of rules, and so it knows that it must
>> not do a priority comparison until it finishes retrieving all the
>> rules.  (The preceding sentence becomes more complicated if we factor
>> in a PDP / PEP split, but I think the principle still applies.)
>> 
>> Where a client can't be sure it's gotten a complete picture is when
>> the policy rules whose priorities compete against each other are
>> identified only by a keyword.  In this case an unlucky client might
>> search a portion of the DIT for policy rules containing the keyword,
>> and get back only a partial set of rules, because it happened to do
>> the search while the rules were being added to the directory.
>> 
>> In this case, if the Policy Tool adds the rules to the directory in
>> order of decreasing priority, the likelihood of a mishap can be
>> reduced.  It isn't eliminated completely, though, because of the
>> following possible sequence of events:
>> 
>> 1. The client searches the portion of the DIT that will contain the
>>    highest priority rule, and it finds nothing.
>> 2. The Policy Tool adds the highest priority rule to the portion of
>>    the DIT that the client just searched.
>> 3. The Policy Tool adds a lower priority rule to a different portion
>>    of the DIT.
>> 4. The client searches this second portion of the DIT, and finds the
>>    lower priority rule.
>> 
>> At this point the client is set up to do the "wrong" thing (i.e., the
>> action(s) associated with the lower priority rule) if it's
>> presented with a stimulus that satisfies the conditions of both rules.
>> Notice, though, that this has nothing to do with splitting a single
>> policy rule into rule, condition, and action entries.  Exactly the
>> same thing could happen if each rule were represented by a single
>> entry.  I understand that Raju has raised some additional questions /
>> concerns related to prioritization of policy rules.  As we address
>> these concerns, we should also consider how to insure that a client
>> doesn't do a priority comparison when it has retrieved only a subset
>> of the rules that belong in the set to be compared.
>> 
>> Regards,
>> Bob
>> 
>> Bob Moore
>> IBM Networking Software
>> +1-919-254-4436
>> remoore@us.ibm.com
>> 


__________________________________________________________
Shai Herzog                  Founder & CTO, IPHighway Inc.         
Cell: (408) 390-3045       http://www.iphighway.com/herzog           
Tel : (201) 585-0800                  Fax : (212) 656-1006 


From majordomo@raleigh.ibm.com  Sun Jun 13 21:48:29 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA23368
	for <policy-archive@odin.ietf.org>; Sun, 13 Jun 1999 21:48:29 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id VAA23870;
	Sun, 13 Jun 1999 21:48:22 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id VAA31524;
	Sun, 13 Jun 1999 21:44:54 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA51708; Sun, 13 Jun 1999 21:29:26 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA46832; Sun, 13 Jun 1999 21:29:23 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id VAA31702
	for <policy@raleigh.ibm.com>; Sun, 13 Jun 1999 21:29:24 -0400
Received: from rip.psg.com (rip.psg.com [147.28.0.39])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id VAA33308
	for <policy@raleigh.ibm.com>; Sun, 13 Jun 1999 21:29:20 -0400
Received: from localhost (1774 bytes) by rip.psg.com
	via sendmail with P:stdio/R:inet_resolve/T:smtp
	(sender: <randy>) (ident <randy> using unix)
	id <m10tLYp-0008G4C@rip.psg.com>
	for <policy@raleigh.ibm.com>; Sun, 13 Jun 1999 18:29:15 -0700 (PDT)
	(Smail-3.2.0.101 1997-Dec-17 #1 built 1999-Apr-1)
Message-Id: <m10tLYp-0008G4C@rip.psg.com>
Date: Sun, 13 Jun 1999 18:29:15 -0700 (PDT)
From: Randy Bush <randy@psg.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: Shai Herzog <herzog@iphighway.com>
Cc: policy@raleigh.ibm.com
Subject: RE: Transactional issues with the LDAP Core Schema
References: <4A043A1FE4B2D111AC3F00A0C96B513301B0F5A0@fmsmsx37.fm.intel
 .com>
	<4.1.19990613152827.0444fc50@bmailnj.iphighway.com>
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Randy Bush <randy@psg.com>
Content-Transfer-Encoding: 7bit

[ making trouble and very definitely not speaking as an iesg member ]

> Wouldn't a better approach would be to put a "semaphore lock" on
> the root of the rule. Given the infrequent update, overhead is NOT
> a problem, but it may solve virtually all the updating problems.

if the lock is stored in ldap, then you have the problem of a holder
crashing or spacing out.  if it is not in ldap, then how can one share the
data across client workstations?  

we are about to learn all the lessons of opti/pessimistic concurrency,
rollback, two phase commit, ...  and there will be real fun if we want
multiple copies or otherwise distributed data.

"The first thing you should not do is treat the directory like a general
database."  -- _LDAP: Programming Directory-Enabled Applications with
Lightweight Directory Access Protocol_; T Howes, M Smith.

let me stir further trouble.  yes, ldap is the 'in thing' this year (well,
at least dns is not the victim any more:-).  but i hypothize that the vast
majority of users who will wanto to deploy the level of tools contemplated
here already have real relational database systems in-house.  and, not only
do those systems have transaction models, but they also won't require
embarrassing kludging to map cim's relational model onto a transaction-free
hierarchic dictionary.

randy


From majordomo@raleigh.ibm.com  Mon Jun 14 12:59:34 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA18278
	for <policy-archive@odin.ietf.org>; Mon, 14 Jun 1999 12:59:33 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id MAA34348;
	Mon, 14 Jun 1999 12:59:21 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id MAA24292;
	Mon, 14 Jun 1999 12:58:58 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA50396; Mon, 14 Jun 1999 12:35:25 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA42960; Mon, 14 Jun 1999 12:35:23 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id MAA27158
	for <policy@raleigh.ibm.com>; Mon, 14 Jun 1999 12:35:25 -0400
Received: from research.solidum.com (IDENT:root@gigabit.solidum.com [216.13.130.242] (may be forged))
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA18608
	for <policy@raleigh.ibm.com>; Mon, 14 Jun 1999 11:39:46 -0400
Received: from venus.solidum.com (venus.solidum.com [192.168.1.1])
          by research.solidum.com (8.8.4/8.8.4) with ESMTP
	  id LAA08842 for <policy@raleigh.ibm.com>; Mon, 14 Jun 1999 11:22:55 -0400
Received: from phobos.solidum.com (mcr@phobos.solidum.com [192.168.1.13])
	by venus.solidum.com (8.8.7/8.8.7) with ESMTP id LAA17559
	for <policy@raleigh.ibm.com>; Mon, 14 Jun 1999 11:39:40 -0400
Message-Id: <199906141539.LAA17559@venus.solidum.com>
To: policy@raleigh.ibm.com
Subject: Re: Transactional issues with the LDAP Core Schema 
In-Reply-To: Your message of "Sun, 13 Jun 1999 15:35:49 EDT."
             <4.1.19990613152827.0444fc50@bmailnj.iphighway.com> 
Mime-Version: 1.0 (generated by tm-edit 7.108)
Content-Type: text/plain; charset=US-ASCII
Date: Mon, 14 Jun 1999 11:39:40 -0400
From: Michael Richardson <mcr@solidum.com>
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Michael Richardson <mcr@solidum.com>


>>>>> "Shai" == Shai Herzog <herzog@iphighway.com> writes:
    Shai> Admiting to not being an LDAP/Directory guru, I'm wondering if we 
    Shai> are doing the right thing by puting the burden of preventing half 
    Shai> baked rules (groups) on the gui/tool. This opens up a lot of room 
    Shai> for interpretation and attempt to close loopholes, which in turn 
    Shai> lead to incompatible results.

  Aren't we being rather LDAP specific?

  Most databases that I know of permit the concept of multiple record update.
Shouldn't an LDAP specific method be created within the LDAP group to deal
with this requirement and let's get on with defining how policies are
described, not how they are stored.

    Shai> What do you think?

  I think that you wouldn't want to wait for the LDAP lock to be released
due to network congestion in order to call 911.

   :!mcr!:            |  Solidum Systems Corporation, http://www.solidum.com
   Michael Richardson |For a better connected world,where data flows faster<tm>
 Personal: <A HREF="http://www.sandelman.ottawa.on.ca/People/Michael_Richardson/Bio.html">mcr@sandelman.ottawa.on.ca</A>. PGP key available.
 Corporate: <A HREF="mailto:mcr@solidum.com">mcr@solidum.com</A>. 




From majordomo@raleigh.ibm.com  Sat Jun 19 19:37:53 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA24469
	for <policy-archive@odin.ietf.org>; Sat, 19 Jun 1999 19:37:53 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id TAA23722;
	Sat, 19 Jun 1999 19:37:40 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id TAA05516;
	Sat, 19 Jun 1999 19:37:39 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA50436; Sat, 19 Jun 1999 19:13:20 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA34556; Sat, 19 Jun 1999 19:13:16 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id TAA39564;
	Sat, 19 Jun 1999 19:13:20 -0400
Received: from nsm-mail2.cisco.com (nsm-mail2.cisco.com [171.71.236.25])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id TAA23656;
	Sat, 19 Jun 1999 19:13:17 -0400
Received: from jstrassnpc.cisco.com ([171.69.108.130])
	by nsm-mail2.cisco.com (8.8.8-Cisco List Logging/8.8.8) with SMTP id QAA09972;
	Sat, 19 Jun 1999 16:12:45 -0700 (PDT)
Message-Id: <001b01bebaa9$58b2ba20$826c45ab@cisco.com>
From: "John Strassner" <jstrassn@cisco.com>
To: <policy@raleigh.ibm.com>
Cc: <johns@cisco.com>, <ellesson@raleigh.ibm.com>, <wijnen@vnet.ibm.com>,
        <randy@psg.com>, <fred@cisco.com>
Subject: NEWLY REVISED Policy Framework Working Group Charter
Date: Sat, 19 Jun 1999 16:13:29 -0700
Organization: Cisco Systems
Mime-Version: 1.0
Content-Type: text/plain;
	charset="utf-8"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "John Strassner" <jstrassn@cisco.com>
Content-Transfer-Encoding: 7bit

Policy Enthusiasts,

First, a heartfelt welcome to Bert and Randy for being our new
ADs. And many thanks to Fred for taking us this far, we really
appreciate it.

Below is a new version of our proposed new charter. This version
takes into account the new IESG position on the PIB, plus talks
with both Fred and Bert and other working group chairs and ADs.

Please reply with comments on this charter asap - we'd like to
get working group consensus before Oslo.

regards,
John and Ed

-------------------------------------- start
here ---------------------------------

Policy Framework Working Group Charter

Chairs:

Ed Ellesson <ellesson@raleigh.ibm.com>
John Strassner <johns@cisco.com>

O&M Area Directors:

Bert Wijnen <wijnen@vnet.ibm.com>
Randy Bush <randy@psg.com>


Mailing Lists:

  List: policy@raleigh.ibm.com

  To Subscribe: policy-request@raleigh.ibm.com
  In Body:      subscribe
  Archive:      http//www.raleigh.ibm.com/maillists/policy/


Description of the Working Group:

Problem Statement:

There is a need to represent, manage, share and reuse policies
and policy information in a vendor-independent, interoperable,
and scalable manner. This working group has three main goals.
First, to provide a framework that will meet these needs. Second,
to define an extensible information model and specific schemata
compliant with that framework that can be used for general policy
representation (called the core information model and schema).
For now, only a directory schema will be defined. Third, to
extend
the core information model and schema to address the needs of
QoS traffic management (called the QoS information model and
schemata).

The viability of the framework will be proven by demonstrating
that high-level policy information can be translated into device
configuration information for network QoS applications. This
requires the coordination of the core and QoS schemata, the
PIB and MIB being developed in DiffServ, and possibly
extensions to COPS provisioning, which is being developed
in RAP. A secondary goal of this framework is to show that this
general development process can be extended to other
application domains.


Objectives:

The objectives of this working group are to:

1. Identify a set of representative use cases to guide us in
defining a policy framework, information model, and schemata
to store, retrieve, distribute and process policies. These use
cases should map to a set of policy rules, and aid us in defining
the composition of policies.

2. Define a framework for intra-domain policy definition and
administration for a heterogeneous set of Policy Decision and
Enforcement Points. The framework will be shown to be able to
be used to represent, distribute, and manage policies and
policy information in an unambiguous, interoperable manner in
a single administrative domain. This framework will be applied
to network QoS.

3. A general information model, derived from the CIM/DEN
policy model, will be produced. This is intended to serve as a
generic means for representing policies and policy information.
In addition, a mapping of this information model to a form that
can be implemented in a directory that uses LDAPv3 as its
access protocol will also be done.

4. Refinements to the above, for representing signaled and
provisioned QoS, will be done. That is, both the information
model as well as the schema will be extended to focus on
network QoS. This will also be used to prove the general
extensibility of the model.

5. A key part of demonstrating that this model can provide
end-to-end translation of high-level policy specifications to
device configurations is to ensure that the information
model and schemata are compatible with and can use the
information contained in the PIB(s) and MIB(s) being
developed in the Differentiated Services WG. To this end,
the Policy Framework WG will supply input to the
development of the PIBs, and include all applicable PIBs
and MIBs in its development considerations for the
framework, information model, and schemata.

6. Policy information may be communicated using several
protocols. The COPS protocol, being developed in the RAP
WG, is an example of one such protocol. The Policy
Framework WG will work with the RAP WG to define usage
directives for use of the COPS base protocol to support
policy information exchange transactions within the
framework being standardized in the Policy Framework WG.

7. The Policy Framework WG will work closely with the
IPSP WG to ensure that the IPsec data model fits and can
be supported within the general framework defined by the
Policy Framework WG.

8. The Policy Framework WG will work with other WGs as
needed to ensure that the framework, information model,
and specific schemata produced meet the needs of these
WGs.

9. To ensure that security issues are addressed, Russ Mundy
will serve as this working group's security adviser.

10. The charter specifically excludes:
  -protocol definition
  -schema attributes or classes that are vendor-specific
   (although the schema defined in this group will be defined
   in a way that is extensible by specific vendors)


Goals and Milestones:



1. Policy terminology draft (Informational Track)
   - Existing draft already submitted
   - Working Group Last Call by August 1999
2. Use case definition (Informational Track)
   - First draft by end of August 1999
   - Working Group Last Call by November 1999
3. Core information model draft (Standards Track)
   - Existing draft already submitted
   - Working group Last Call September 1999
4. Core LDAP schema draft (Standards Track)
   - Existing draft already submitted
   - Working group Last Call October 1999
5. Framework draft (Informational Track)
   - Existing draft already submitted
   - Working group Last Call, December 19996
6. QoS Schema draft(s) (Standards Track)
   - Existing draft already submitted
   - Working group Last Call, March 2000



From majordomo@raleigh.ibm.com  Sun Jun 20 02:37:23 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id CAA11144
	for <policy-archive@odin.ietf.org>; Sun, 20 Jun 1999 02:37:23 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id CAA07788;
	Sun, 20 Jun 1999 02:36:53 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id CAA22232;
	Sun, 20 Jun 1999 02:36:44 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA42438; Sun, 20 Jun 1999 02:15:19 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA51846; Sun, 20 Jun 1999 02:15:14 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id CAA38918;
	Sun, 20 Jun 1999 02:15:15 -0400
Received: from warp.lannet.com (at.lannet.com [194.90.94.231])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id CAA24654;
	Sun, 20 Jun 1999 02:15:08 -0400
Received: from itc-eml1.ta.israel.madge.com (itceml1.ta.israel.madge.com [149.49.50.90])
	by warp.lannet.com (8.8.8+Sun/8.8.8) with ESMTP id JAA29992;
	Sun, 20 Jun 1999 09:05:04 +0300 (IDT)
Received: by itc-eml1 with Internet Mail Service (5.5.2448.0)
	id <LF26GNWV>; Sun, 20 Jun 1999 09:13:55 +0200
Message-Id: <8C9B70BAC7C9D2119A0A0090273FADFB2AC838@itc-eml1>
From: Dan Romascanu <dromasca@lucent.com>
To: "'John Strassner'" <jstrassn@cisco.com>, policy@raleigh.ibm.com
Cc: johns@cisco.com, ellesson@raleigh.ibm.com, wijnen@vnet.ibm.com,
        randy@psg.com, fred@cisco.com
Subject: RE: NEWLY REVISED Policy Framework Working Group Charter
Date: Sun, 20 Jun 1999 09:13:55 +0200
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Dan Romascanu <dromasca@lucent.com>

	John,

> Below is a new version of our proposed new charter. This version
> takes into account the new IESG position on the PIB, plus talks
> with both Fred and Bert and other working group chairs and ADs.
> 
	[Dan]  Can you update the WG about 'the new IESG position on the
PIB', or provide us a pointer about where this information is available?

	Is there a commonly agreed reference for the PIB concept?

	Thanks and Regards,

	Dan




From majordomo@raleigh.ibm.com  Mon Jun 21 23:59:55 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id XAA16394
	for <policy-archive@odin.ietf.org>; Mon, 21 Jun 1999 23:59:54 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id XAA19822;
	Mon, 21 Jun 1999 23:59:01 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id XAA27118;
	Mon, 21 Jun 1999 23:58:54 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA45570; Mon, 21 Jun 1999 23:48:37 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA32762; Mon, 21 Jun 1999 23:48:35 -0400
Received: from nlvm1.benelux.ibm.com ([9.166.35.168])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id XAA04190
	for <POLICY@raleigh.ibm.com>; Mon, 21 Jun 1999 23:48:40 -0400
Message-Id: <199906220348.XAA04190@rtpmail03.raleigh.ibm.com>
Received: from UITVM1 by nlvm1.benelux.ibm.com (IBM VM SMTP V2R4)
   with BSMTP id 1553; Tue, 22 Jun 99 05:46:19 DST
Date: Tue, 22 Jun 99 05:46:19 DST
From: "Bert Wijnen" <WIJNEN@vnet.ibm.com>
To: dromasca@lucent.com, jstrassn@cisco.com, POLICY@raleigh.ibm.com
Cc: johns@cisco.com, ELLESSON@raleigh.ibm.com, randy@psg.com, fred@cisco.com
Subject: NEWLY REVISED Policy Framework Working Group Charter
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Bert Wijnen" <WIJNEN@vnet.ibm.com>

Ref:  Your note of Sun, 20 Jun 1999 09:13:55 +0200

Subject: Re:   NEWLY REVISED Policy Framework Working Group Charter

Regarding the "IESG position on PIB work" let me try and phrase
it this way:

What the various involved ADs discussed and agreed upon was that
the PIB work basically belongs in the WG where the specific
protocol is defined. But we understand that it needs to be in sync
with the Schema defined in the Policy FW WG, so we have "encouraged"
that the WGs interact with each other on this topic.

Hope this helps,
Bert


From majordomo@raleigh.ibm.com  Tue Jun 22 11:22:24 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA06197
	for <policy-archive@odin.ietf.org>; Tue, 22 Jun 1999 11:22:24 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA23148;
	Tue, 22 Jun 1999 11:21:02 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id LAA28866;
	Tue, 22 Jun 1999 11:20:18 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA31014; Tue, 22 Jun 1999 11:10:07 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA24862; Tue, 22 Jun 1999 11:10:02 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id LAA13174
	for <policy@raleigh.ibm.com>; Tue, 22 Jun 1999 11:10:03 -0400
Received: from dirty.research.bell-labs.com (dirty.research.bell-labs.com [204.178.16.6])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with SMTP id LAA29058
	for <policy@raleigh.ibm.com>; Tue, 22 Jun 1999 11:10:01 -0400
Received: from grubby.research.bell-labs.com ([135.104.2.9]) by dirty; Tue Jun 22 11:09:09 EDT 1999
Received: from zydeco.research.bell-labs.com ([135.104.120.150]) by grubby; Tue Jun 22 11:09:07 EDT 1999
Received: from research.bell-labs.com (summer118-02.research.bell-labs.com [135.104.121.21])
	by zydeco.research.bell-labs.com (8.9.1/8.9.1) with ESMTP id LAA22344
	for <policy@raleigh.ibm.com>; Tue, 22 Jun 1999 11:08:19 -0400 (EDT)
Message-Id: <376FB542.8888AA8A@research.bell-labs.com>
Date: Tue, 22 Jun 1999 09:09:38 -0700
From: Pradosh Kumar Mohapatra <mohap@research.bell-labs.com>
X-Mailer: Mozilla 4.5 [en] (WinNT; U)
X-Accept-Language: en
Mime-Version: 1.0
To: policy@raleigh.ibm.com
Subject: Policy questions
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Pradosh Kumar Mohapatra <mohap@research.bell-labs.com>
Content-Transfer-Encoding: 7bit

Hi,
I have a few questions regarding the overall operation of the policy
framwork and
architecture:

1) Is there any specification for the policy server architecture? What
does it
mean by 'policy server maintaining a state?.' How do you define the
states of the
overall architecture? Is there any DFA or something of that stuff that
is
implemented in the policy server?
2) How does the actual enforcement of policy in PEP work? Once the
administrator
defines some policies in the server, how does it get stored in the PEP
so that
when such a condition arises, the PEP will require to send a policy
notification
to the PDP?

I did not get any proper answers to the above questions after going
through all
the documents.

Thanks and Regards,
Pradosh






From majordomo@raleigh.ibm.com  Tue Jun 22 11:45:21 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA06841
	for <policy-archive@odin.ietf.org>; Tue, 22 Jun 1999 11:45:21 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA07878;
	Tue, 22 Jun 1999 11:45:00 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id LAA20626;
	Tue, 22 Jun 1999 11:44:58 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA38994; Tue, 22 Jun 1999 11:37:40 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA29002; Tue, 22 Jun 1999 11:37:37 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id LAA31368;
	Tue, 22 Jun 1999 11:37:39 -0400
Received: from mail-blue.research.att.com (mail-blue.research.att.com [135.207.30.102])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA12580;
	Tue, 22 Jun 1999 11:37:36 -0400
Received: from bigmail.research.att.com (bigmail.research.att.com [135.207.30.101])
	by mail-blue.research.att.com (Postfix) with ESMTP
	id 0F8844CE04; Tue, 22 Jun 1999 11:37:32 -0400 (EDT)
Received: from pcraju (pc-raju.research.att.com [135.207.27.140])
	by bigmail.research.att.com (8.8.8/8.8.8) with SMTP id LAA10910;
	Tue, 22 Jun 1999 11:37:31 -0400 (EDT)
Message-Id: <003201bebcc6$291dd500$8c1bcf87@research.att.com>
From: "Raju Rajan" <rajan@research.att.com>
To: "Bert Wijnen" <WIJNEN@vnet.ibm.com>, <dromasca@lucent.com>,
        <jstrassn@cisco.com>, <POLICY@raleigh.ibm.com>
Cc: <johns@cisco.com>, <ELLESSON@raleigh.ibm.com>, <randy@psg.com>,
        <fred@cisco.com>
References: <199906220348.XAA04190@rtpmail03.raleigh.ibm.com>
Subject: Re: NEWLY REVISED Policy Framework Working Group Charter
Date: Tue, 22 Jun 1999 11:44:48 -0400
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Raju Rajan" <rajan@research.att.com>
Content-Transfer-Encoding: 7bit

Hi Bert, John

What Bert says below helps a little. From expanding the acronym I also
figured out that it has something to do with "policy" and "information". But
I am in the dark about what type of policy, where this information is
intended to reside, how it is expected to be transported, and most
importantly, what the *purpose* of this activity is. Is the LDAP schema a
PIB? Why or why not? It would be better if we could scope the activity and
come up with the right acronym rather than proceed in the opposite
direction.

raju

----- Original Message -----
From: Bert Wijnen <WIJNEN@vnet.ibm.com>
To: <dromasca@lucent.com>; <jstrassn@cisco.com>; <POLICY@raleigh.ibm.com>
Cc: <johns@cisco.com>; <ELLESSON@raleigh.ibm.com>; <randy@psg.com>;
<fred@cisco.com>
Sent: Tuesday, June 22, 1999 1:46 AM
Subject: NEWLY REVISED Policy Framework Working Group Charter


> Ref:  Your note of Sun, 20 Jun 1999 09:13:55 +0200
>
> Subject: Re:   NEWLY REVISED Policy Framework Working Group Charter
>
> Regarding the "IESG position on PIB work" let me try and phrase
> it this way:
>
> What the various involved ADs discussed and agreed upon was that
> the PIB work basically belongs in the WG where the specific
> protocol is defined. But we understand that it needs to be in sync
> with the Schema defined in the Policy FW WG, so we have "encouraged"
> that the WGs interact with each other on this topic.
>
> Hope this helps,
> Bert
>



From majordomo@raleigh.ibm.com  Tue Jun 22 12:47:23 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA08321
	for <policy-archive@odin.ietf.org>; Tue, 22 Jun 1999 12:47:23 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id MAA24062;
	Tue, 22 Jun 1999 12:47:06 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id MAA34156;
	Tue, 22 Jun 1999 12:47:03 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA32904; Tue, 22 Jun 1999 12:38:02 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA33404; Tue, 22 Jun 1999 12:37:58 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id MAA07028;
	Tue, 22 Jun 1999 12:37:58 -0400
Received: from nsm-mail2.cisco.com (nsm-mail2.cisco.com [171.71.236.25])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id MAA28942;
	Tue, 22 Jun 1999 12:37:54 -0400
Received: from jstrassnpc.cisco.com ([171.69.108.130])
	by nsm-mail2.cisco.com (8.8.8-Cisco List Logging/8.8.8) with SMTP id JAA13192;
	Tue, 22 Jun 1999 09:30:25 -0700 (PDT)
Message-Id: <03a101bebccc$9f342ea0$826c45ab@cisco.com>
From: "John Strassner" <jstrassn@cisco.com>
To: "Raju Rajan" <rajan@research.att.com>, "Bert Wijnen" <WIJNEN@vnet.ibm.com>,
        <dromasca@lucent.com>, <jstrassn@cisco.com>, <POLICY@raleigh.ibm.com>
Cc: <johns@cisco.com>, <ELLESSON@raleigh.ibm.com>, <randy@psg.com>,
        <fred@cisco.com>
References: <199906220348.XAA04190@rtpmail03.raleigh.ibm.com> <003201bebcc6$291dd500$8c1bcf87@research.att.com>
Subject: Re: NEWLY REVISED Policy Framework Working Group Charter
Date: Tue, 22 Jun 1999 09:31:03 -0700
Organization: Cisco Systems
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2314.1300
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2314.1300
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "John Strassner" <jstrassn@cisco.com>
Content-Transfer-Encoding: 7bit

I suggest you look at the minutes for Minneapolis. Keith gave a
presentation on the PIB, which stands for Policy Information
Base.

John
----- Original Message -----
From: Raju Rajan <rajan@research.att.com>
To: Bert Wijnen <WIJNEN@vnet.ibm.com>; <dromasca@lucent.com>;
<jstrassn@cisco.com>; <POLICY@raleigh.ibm.com>
Cc: <johns@cisco.com>; <ELLESSON@raleigh.ibm.com>;
<randy@psg.com>; <fred@cisco.com>
Sent: Tuesday, June 22, 1999 8:44 AM
Subject: Re: NEWLY REVISED Policy Framework Working Group Charter


> Hi Bert, John
>
> What Bert says below helps a little. From expanding the acronym
I also
> figured out that it has something to do with "policy" and
"information". But
> I am in the dark about what type of policy, where this
information is
> intended to reside, how it is expected to be transported, and
most
> importantly, what the *purpose* of this activity is. Is the
LDAP schema a
> PIB? Why or why not? It would be better if we could scope the
activity and
> come up with the right acronym rather than proceed in the
opposite
> direction.
>
> raju
>
> ----- Original Message -----
> From: Bert Wijnen <WIJNEN@vnet.ibm.com>
> To: <dromasca@lucent.com>; <jstrassn@cisco.com>;
<POLICY@raleigh.ibm.com>
> Cc: <johns@cisco.com>; <ELLESSON@raleigh.ibm.com>;
<randy@psg.com>;
> <fred@cisco.com>
> Sent: Tuesday, June 22, 1999 1:46 AM
> Subject: NEWLY REVISED Policy Framework Working Group Charter
>
>
> > Ref:  Your note of Sun, 20 Jun 1999 09:13:55 +0200
> >
> > Subject: Re:   NEWLY REVISED Policy Framework Working Group
Charter
> >
> > Regarding the "IESG position on PIB work" let me try and
phrase
> > it this way:
> >
> > What the various involved ADs discussed and agreed upon was
that
> > the PIB work basically belongs in the WG where the specific
> > protocol is defined. But we understand that it needs to be in
sync
> > with the Schema defined in the Policy FW WG, so we have
"encouraged"
> > that the WGs interact with each other on this topic.
> >
> > Hope this helps,
> > Bert
> >
>
>



From majordomo@raleigh.ibm.com  Tue Jun 22 15:21:49 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA14129
	for <policy-archive@odin.ietf.org>; Tue, 22 Jun 1999 15:21:48 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA31978;
	Tue, 22 Jun 1999 15:21:01 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id PAA31052;
	Tue, 22 Jun 1999 15:20:58 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA33220; Tue, 22 Jun 1999 15:15:04 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA34234; Tue, 22 Jun 1999 15:15:02 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id PAA24476
	for <policy@raleigh.ibm.com>; Tue, 22 Jun 1999 15:15:05 -0400
Received: from sol.extremenetworks.com (sol.extremenetworks.com [216.52.8.2])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA22042
	for <policy@raleigh.ibm.com>; Tue, 22 Jun 1999 15:15:01 -0400
Received: by ftp.extremenetworks.com with Internet Mail Service (5.5.2448.0)
	id <NJQF3X62>; Tue, 22 Jun 1999 12:14:32 -0700
Message-Id: <D0805D3B448BD211A7990008C7B18130142F01@ftp.extremenetworks.com>
From: Andrew Smith <andrew@extremenetworks.com>
To: "'Pradosh Kumar Mohapatra'" <mohap@research.bell-labs.com>,
        policy@raleigh.ibm.com
Subject: RE: Policy questions
Date: Tue, 22 Jun 1999 12:14:28 -0700
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Andrew Smith <andrew@extremenetworks.com>

I'll let others answer the first part of your questions. 

Enforcement of policy decisions in PEP devices for RSVP admission policy is
described in the RAG WG documents e.g. in
http://www.ietf.org/internet-drafts/draft-ietf-rap-framework-03.txt and the
succeeding references. Enforcement of other types of policy decisions have
not really been well described in the standards' literature so far - I think
it is left as an exercise for the implementor to work out how to use MIBs
and other configuration mechanisms (e.g. command line/telnet) to make
devices apply the desired rules. IETF is just starting up standards' work in
some of these areas e.g. configuration of diffserv QoS rules (there will be
a number of drafts submitted this week to the diffserv WG on this topic). I
believe that there is work in progress in the AAA WG on how to enforce
policy rules for some other application areas e.g. security policy. 

Right now in the QoS policy area (with the exception of RSVP policy), I
believe that proprietary solutions (e.g. vendor/enterprise MIBs,
command-lines, proprietary uses of COPS) are the only viable mechanisms.

Andrew Smith
RAP WG co-chair


> -----Original Message-----
> From: Pradosh Kumar Mohapatra [mailto:mohap@research.bell-labs.com]
> Sent: Tuesday, June 22, 1999 9:10 AM
> To: policy@raleigh.ibm.com
> Subject: Policy questions
> 
> 
> Hi,
> I have a few questions regarding the overall operation of the policy
> framwork and
> architecture:
> 
> 1) Is there any specification for the policy server architecture? What
> does it
> mean by 'policy server maintaining a state?.' How do you define the
> states of the
> overall architecture? Is there any DFA or something of that stuff that
> is
> implemented in the policy server?
> 2) How does the actual enforcement of policy in PEP work? Once the
> administrator
> defines some policies in the server, how does it get stored in the PEP
> so that
> when such a condition arises, the PEP will require to send a policy
> notification
> to the PDP?
> 
> I did not get any proper answers to the above questions after going
> through all
> the documents.
> 
> Thanks and Regards,
> Pradosh
> 
> 
> 
> 


From majordomo@raleigh.ibm.com  Tue Jun 22 17:44:37 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA18293
	for <policy-archive@odin.ietf.org>; Tue, 22 Jun 1999 17:44:37 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA32506;
	Tue, 22 Jun 1999 17:44:04 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id RAA23136;
	Tue, 22 Jun 1999 17:43:56 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA34274; Tue, 22 Jun 1999 17:37:08 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA23770; Tue, 22 Jun 1999 17:37:06 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id RAA35122
	for <policy@raleigh.ibm.com>; Tue, 22 Jun 1999 17:37:05 -0400
Received: from mail.webstream.net (mail.webstream.net [209.203.197.7])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA19828
	for <policy@raleigh.ibm.com>; Tue, 22 Jun 1999 17:36:54 -0400
Received: from deterministic ([205.179.45.178]) by mail.webstream.net
          (Post.Office MTA v3.1 release PO205e ID# 0-12345L500S10000V35)
          with SMTP id AAA202; Tue, 22 Jun 1999 17:38:42 -0400
Message-Id: <003601bebcf7$9fab9a00$b22db3cd@DeterministicNetworks.com>
From: stevej@DeterministicNetworks.com (SteveJ)
To: "Andrew Smith" <andrew@extremenetworks.com>,
        "'Pradosh Kumar Mohapatra'" <mohap@research.bell-labs.com>,
        <policy@raleigh.ibm.com>
References: <D0805D3B448BD211A7990008C7B18130142F01@ftp.extremenetworks.com>
Subject: Re: Policy questions
Date: Tue, 22 Jun 1999 14:38:52 -0700
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-Msmail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.00.2014.211
X-Mimeole: Produced By Microsoft MimeOLE V5.00.2014.211
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: stevej@DeterministicNetworks.com (SteveJ)
Content-Transfer-Encoding: 7bit

Hi Pradosh,

> > 2) How does the actual enforcement of policy in PEP work? Once the
> > administrator
> > defines some policies in the server, how does it get stored in the PEP
> > so that
> > when such a condition arises, the PEP will require to send a policy
> > notification
> > to the PDP?
> >
I might be able to help you out here by describing what we do in our PEPs. I
think Andrew did a good job of explaining how the information is passed to
the PEP today by most vendors - several of the policy server vendors use
mechanisms like CLI to 'configure' the PEPs.   We take a slightly different
approach in the PEPs.  First, our PEPs are host-based.  That is, they reside
on clients and servers.  This allows us to change the model slightly because
we have better classification of flows and more resident storage for policy
information. To some degree, this is similar to what the folks at IBM are
doing with their IBM Common Policy Engine (CPE) - a combined PDP-PEP.

We're not convinced that a model where the PEP 'sends a policy notification
to the PDP' is scalable.  After all, even a simple browser page can involve
dozens of flows.  Add dozens of pages and potentially dozens or hundreds of
clients, and you see the problem. Keep in mind that the original
architecture of COPS was inspired by RSVP where the policies would be
associated with per-flow reservations.  While we do support this, we prefer
to have the PDP/policy server set policy objects into our PEP.  We then
watch for flows either in promiscuous mode (if we're a non-intrusive
monitor), as they pass through us (if we're a gateway), or when they're
initiated by the applications (when we're running on the hosts) - we don't
notify the PDP on a per-flow basis.  Our classification engine identifies
the flow start and notifies the local policy agent (or tells the PDP that a
flow has started).  The policy agent looks at our PIB to see if this flow
needs to have one or more policies applied.  If so, it 'configures'  the
enforcement agents to apply the specific policies to the flow.  Depending on
the duration (one of our policy objects) of the policy, the policy may
terminate when the flow does or it may continue until it expires.

So, although this is a bit different than the current RAP framework might
describe, with the updates to their charter, and less reliance on RSVP, we
should see (actually we already are) more scalable approaches in the
relationships of Policy Server, PDP, and PEP.


Steve




----------------------------------------------------------------
Steve Jackowski
President/CEO
Deterministic Networks, Inc.
245M Mt. Hermon Rd. #140
Scotts Valley, Ca 95066

Office: (831) 421-0388 x18
Home office: (831) 426-4001
Mobile: (408) 813-6294
Fax: (831) 421-0394




From majordomo@raleigh.ibm.com  Wed Jun 23 00:29:48 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id AAA07090
	for <policy-archive@odin.ietf.org>; Wed, 23 Jun 1999 00:29:48 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id AAA36542;
	Wed, 23 Jun 1999 00:29:30 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id AAA35192;
	Wed, 23 Jun 1999 00:27:31 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA45864; Wed, 23 Jun 1999 00:19:06 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA50720; Wed, 23 Jun 1999 00:19:04 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id AAA31694
	for <policy@raleigh.ibm.com>; Wed, 23 Jun 1999 00:19:05 -0400
Received: from gw.ipdevices.com (gw.ipdevices.com [207.18.183.19])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id AAA17546
	for <policy@raleigh.ibm.com>; Wed, 23 Jun 1999 00:19:02 -0400
Received: (from smtp@localhost) by gw.ipdevices.com (8.8.5/8.7.3) id VAA10339; Tue, 22 Jun 1999 21:19:00 -0700 (PDT)
Received: from paradise.ipdevices.com (paradise.ipdevices.com [172.18.0.99]) by gw.ipdevices.com id rfVAA10331; Tue Jun 22 21:18:46 1999
Received: from montalvo.ipdevices.com (montalvo.ipdevices.com [172.18.0.27])
	by paradise.ipdevices.com (8.9.3/8.9.1) with ESMTP id VAA02273;
	Tue, 22 Jun 1999 21:18:45 -0700 (PDT)
	(envelope-from mikel@ipdevices.com)
Received: from ipdevices.com (localhost [127.0.0.1])
	by montalvo.ipdevices.com (8.9.3/8.8.5) with ESMTP id VAA61728;
	Tue, 22 Jun 1999 21:19:08 -0700 (PDT)
Message-Id: <3770603C.A729A8D0@ipdevices.com>
Date: Tue, 22 Jun 1999 21:19:08 -0700
From: Mikel Lechner <mikel@ipdevices.com>
X-Mailer: Mozilla 4.51 [en] (X11; I; FreeBSD 3.2-RELEASE i386)
X-Accept-Language: en
Mime-Version: 1.0
To: policy@raleigh.ibm.com
Cc: KIDO Masahiro <kido@ccs.mt.nec.co.jp>
Subject: Proposed additions to vpn policy schema
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Mikel Lechner <mikel@ipdevices.com>
Content-Transfer-Encoding: 7bit

I've been having discussions with Kido Masahiro at NEC
about using policy definitions to limit modifications
of various parameter settings in VPN devices.
The intention is to prevent "dumb" mistakes
and to insure that some basic minimum policy
is always applied.

As an example you might use a policy rule to specify
that DES encryption should not be allowed for IPSEC,
since it is no longer generally regarded as secure.

Another example would be to disallow modification
of parameters via less secure methods (eg. a telnet session).

In line with this ideas we've sketched out the following
possible policy classes for use in the schema.
The policy classes described here are more general than
just VPNs, but since our focus is VPNs, we thought we would
start with this.

Below I've included the definition of new classes,
which would be used to implement this new functionality.

The question is do others in the WG think such policies are
useful for VPNs, or for that matter for more general policy rules?


-- Mikel


======


class DevParamSetPolicyCondition

NAME               DevParamSetPolicyCondition
TYPE               Structural
DERIVED FROM       PolicyCondition
AUXILIARY CLASS    None
MAY                TargetParameter,
                   SettingMethod,
                   SettingRange


NAME      TargetParameter
DESC      Lists the names of device parameters for the condition to be applied
EQUALITY  caseIgnoreMatch
SYNTAX    DirectoryString
MULTI-VALUED
FORMAT    The following values are allowed:
              xxxParam
SEMANTICS xxxParam denotes a particular parameter xxx of a VPN device


NAME      SettingMethod
DESC      Denotes the method of setting for the condition to be applied
EQUALITY  caseIgnoreMatch
SYNTAX    DirectoryString
MULTI-VALUED
FORMAT    The following values are allowed:
              TELNET
              SNMP
SEMANTICS TELNET denotes the telnet protocol.
          SNMP denotes the snmp protocol.


NAME      SettingRange
DESC      Denotes the range of setting for the condition to be applied
EQUALITY  caseExactIA5Match
SYNTAX    IA5String
MULTI-VALUED
FORMAT    String consisting of two colon separated positive
          integers, the second no smaller than the first, or one
          positive integer.  (* Might need to allow a list of values
          or a list of ranges [eg. 1:3,5:7 to denote 1 through 3 and 5
          through 7] *)


NAME             DevParamSetControlAction
TYPE             Structural
DERIVED FROM     PolicyAction
AUXILIARY CLASS  None
MUST             ControlAction
MAY              ReportMethod


NAME      ControlAction
DESC      Control action for the condition
EQUALITY  caseIngoreStringMatch
SYNTAX    DirectoryString
SINGLE-VALUED
FORMAT    The following values are allowed
            Permit
            Deny
SEMANTICS Deny means that the setting must be rejected.
          Permit means that the setting may be accepted.


NAME      ReportMethod
DESC      Report method for the condition
EQUALITY  caseIngoreStringMatch
SYNTAX    DirectoryString
MULTI-VALUED
FORMAT    The following values are allowed
            SendingEMail
            Logging
SEMANTICS SendingEMail means that the report must be sent by E-Mail.
          Logging means that the report must be logged.


From majordomo@raleigh.ibm.com  Wed Jun 23 14:42:59 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA09141
	for <policy-archive@odin.ietf.org>; Wed, 23 Jun 1999 14:42:58 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id OAA28280;
	Wed, 23 Jun 1999 14:42:40 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id OAA22266;
	Wed, 23 Jun 1999 14:42:29 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA35342; Wed, 23 Jun 1999 14:34:05 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA50694; Wed, 23 Jun 1999 14:34:02 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id OAA33590
	for <policy@raleigh.ibm.com>; Wed, 23 Jun 1999 14:34:05 -0400
Received: from dirty.research.bell-labs.com (dirty.research.bell-labs.com [204.178.16.6])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with SMTP id OAA33526
	for <policy@raleigh.ibm.com>; Wed, 23 Jun 1999 14:34:00 -0400
Received: from scummy.research.bell-labs.com ([135.104.2.10]) by dirty; Wed Jun 23 14:33:38 EDT 1999
Received: from zydeco.research.bell-labs.com ([135.104.120.150]) by scummy; Wed Jun 23 14:33:37 EDT 1999
Received: from research.bell-labs.com (summer118-02.research.bell-labs.com [135.104.121.21])
	by zydeco.research.bell-labs.com (8.9.1/8.9.1) with ESMTP id OAA26198
	for <policy@raleigh.ibm.com>; Wed, 23 Jun 1999 14:33:37 -0400 (EDT)
Message-Id: <377136E3.5E15F213@research.bell-labs.com>
Date: Wed, 23 Jun 1999 12:34:59 -0700
From: Pradosh Kumar Mohapatra <mohap@research.bell-labs.com>
X-Mailer: Mozilla 4.5 [en] (WinNT; U)
X-Accept-Language: en
Mime-Version: 1.0
To: policy@raleigh.ibm.com
Subject: multiple clients involved in a policy
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Pradosh Kumar Mohapatra <mohap@research.bell-labs.com>
Content-Transfer-Encoding: 7bit

Hi,
Thanks for the answers to my previous questions regarding PEP and PDPs.
From the documents that I read, I could come to the conclusion that the
policy execution, right now is a single client-to-server communication
where in, if some policy condition is triggered on the PEP, the PEP
sends a COPS message to the PDP, the PDP evaluates it and sends the
decision (some action to be taken based on the condition) to the PEP.
I am not sure how the current framework will provide space to do the
following:
say a policy is entered by the administrator which states if event e1
occurs in device d1, event e2 occurs in device d2, ...., then do some
action a1 in d1, a3 in d3 ... etc.

e1, e2, .... can be events (transitions in the document by Silvano)
occuring at the same time in d1, d2, ... or they occur in different time
frames.

Does the current architecture support this?
If so, then how?

Regards,
Pradosh




From majordomo@raleigh.ibm.com  Thu Jun 24 15:01:43 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA19878
	for <policy-archive@odin.ietf.org>; Thu, 24 Jun 1999 15:01:42 -0400 (EDT)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id PAA36514;
	Thu, 24 Jun 1999 15:01:25 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id PAA27016;
	Thu, 24 Jun 1999 15:01:00 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA43088; Thu, 24 Jun 1999 14:44:37 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA42824; Thu, 24 Jun 1999 14:44:35 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id OAA33068
	for <policy@raleigh.ibm.com>; Thu, 24 Jun 1999 14:44:35 -0400
Received: from atlrel1.hp.com (atlrel1.hp.com [156.153.255.210])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id OAA32414
	for <policy@raleigh.ibm.com>; Thu, 24 Jun 1999 14:44:26 -0400
Received: from xpeditio.cnd.hp.com (xpeditio.cnd.hp.com [15.2.113.211])
	by atlrel1.hp.com (8.8.6 (PHNE_17135)/8.8.5tis) with ESMTP id OAA07125
	for <policy@raleigh.ibm.com>; Thu, 24 Jun 1999 14:43:57 -0400 (EDT)
Received: (from mhugh@localhost) by xpeditio.cnd.hp.com (8.7.1/8.7.3 TIS 5.0) id MAA03073 for policy@raleigh.ibm.com; Thu, 24 Jun 1999 12:44:21 -0600 (MDT)
From: "Hugh F. Mahon" <mhugh@xpeditio.cnd.hp.com>
Message-Id: <199906241844.MAA03073@xpeditio.cnd.hp.com>
Subject: Policy framework requirements
To: policy@raleigh.ibm.com
Date: Thu, 24 Jun 1999 12:44:21 MDT
Organization:  HP Network & System Management Division
X-Mailer: Elm [revision: 212.2]
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Hugh F. Mahon" <mhugh@xpeditio.cnd.hp.com>

Hi Folks,

Within this working group there have been a lot of discussions of what 
kind of policy information is needed, and there's been a lot of work on 
schema for policy information.  There has also been a lot of discussion 
of how that information is to be used, and how it is shared.

Each of the participants has had in mind a model for how things work, 
but we rarely discuss this aspect of policy management.  If we are to 
indeed develop standards around which we can have interoperable policy 
management tools we need to have discussions of how these mechanisms 
will work together.

I am specifically not talking about how a UI looks or how policy is 
verified to be usable.  

I am talking about how policy gets from the administrator to where it 
is actually used, what the administrator (customer) expects from the 
policy management system, and the minimum pieces necessary to provide 
a functional system.  Without this understanding I do not see how we 
can define schema and how it is to be used that allows true 
interoperability (or value to customers).

To begin the discussion I've put together a simple description of what 
an administrator would do and expect.  The purpose is to drive toward 
common understanding and help in determining the solutions (or even 
allow us to choose which things to solve and what not to solve).

The Scenario
------------

An administrator must author policy.  This can be done either from 
scratch or using existing policies (either editing existing policies 
or copying them and using them for creating new policies).  Once new 
policies have been authored, the administrator must associate them 
with the managed entities the policies are to affect.  The 
administrator can either 'just know' to what policies can be 
associated, or a list can be presented.  There are other reasons that 
such a list would be valuable in the management system (see below).

Because policy can be complex, or have information the administrator 
may wish to track over time, it would be valuable to allow versioning 
of policy information, that is, earlier versions of policy information 
would be stored.  This versioning allows administrators to make 
changes without fear that valuable information will be irretrievably 
lost.  One example of this value is being able to quickly roll-back to 
a previous version of policy if the new policy is found to be 
incorrect. 

Once the policy has been created or edited by the administrator, it 
should be stored in a repository.  (Note that it is this repository 
that has been the focus of much of the group's discussions and is 
viewed as the central interoperability point.)  The managed entity 
should then receive the policy in a timely fashion and implement the 
policy.  Once the policy has been implemented on the managed entity 
the administrator should be notified of the status of the policy 
deployment: success, success with warnings, or failure (with the 
reason(s) for the failure).  This is an area where the list of managed 
entities would provide value: a place to associate status information 
about policy on those managed entities. 

Administrator Expectations
--------------------------

For example, if a network is experiencing poor performance, the 
administrator may choose to change policy.  Say a new tool for 
receiving news has been introduced, and it is using significant 
amounts of bandwidth.  We'll call it NewTool.  The administrator 
cannot stop all users from using NewTool, but can manage the QoS for 
the traffic it generates.  To improve the performance of other 
applications the administrator chooses to reduce the priority of 
NewTool.  Customers are clamoring for improved performance, so the 
administrator would like the changes to take effect immediately.  It 
is not desirable to wait an hour for all of the affected devices to 
query a central repository and learn that new policy exists.   Neither 
is it desirable to have each of the managed entities generating 
extraneous traffic to check for policy changes, adding a further 
burden to the already overworked network. 

In addition, the administrator wants to be notified as soon as the 
policy is in place on each of the affected managed entities.  Without 
such feedback, the administrator is not in control of the environment 
but at the mercy of the policy management system.

The Requirements
----------------

Based on the simple scenario described above, the requirements for a 
policy management system should include:

          - storage of policy information        
          - storage of information about managed entities 
          - an association between policies and what they are to manage
          - timely asynchronous notification of new/changed policy 
            information
          - ability to provide feedback to administrator of policy 
            status at managed entity
          - transactional qualities (the administrator can determine 
            that the changes were the result of a specific 
            administrator action)
          - definition of how policy information is to be used on 
            managed entities (e.g., how many actions are performed for 
            any event, etc.)
          - ensure that policies put in directory are what is read by 
            the PDP/interpreter

LDAP (and directories accessed by LDAP) can't do all of these things, 
at least not as it exists today (or in the near future).  Other 
mechanisms need to be in place to supplement LDAP's capabilities.

I agree with the Area Directors' concerns about adding more to the 
Working Group's tasks before the WG achieves its initial goals (i.e., 
a schema), but I can't help but feel that we will miss important 
things if we don't discuss how all of this is supposed to work.  We 
don't have to define all of it before we get some pieces firmed up, 
but we do need to discuss these other aspects more before we can call 
anything done or ready for use.


        Hugh Mahon                              |__|     
        Hewlett-Packard Company                 |  |     
        e-mail: hugh_mahon@hp.com                  |\  /|
        Phone: (970) 898-BITS (898-2487)           | \/ |     



From majordomo@raleigh.ibm.com  Fri Jun 25 14:57:44 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA00761
	for <policy-archive@odin.ietf.org>; Fri, 25 Jun 1999 14:57:43 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id OAA28368;
	Fri, 25 Jun 1999 14:57:28 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id OAA37796;
	Fri, 25 Jun 1999 14:57:16 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA41440; Fri, 25 Jun 1999 14:36:48 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA41428; Fri, 25 Jun 1999 14:36:45 -0400
Received: from southrelay02.raleigh.ibm.com (southrelay02.raleigh.ibm.com [9.37.3.209])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id OAA31016
	for <policy@raleigh.ibm.com>; Fri, 25 Jun 1999 14:36:46 -0400
From: remoore@us.ibm.com
Received: from d54mta01.raleigh.ibm.com (d54mta01.raleigh.ibm.com [9.67.228.33])
	by southrelay02.raleigh.ibm.com (8.8.8m2/NCO v2.03) with SMTP id OAB42982;
	Fri, 25 Jun 1999 14:36:44 -0400
Received: by d54mta01.raleigh.ibm.com(Lotus SMTP MTA v4.6.4  (830.2 3-23-1999))  id 8525679B.00663BF6 ; Fri, 25 Jun 1999 14:36:40 -0400
X-Lotus-Fromdomain: IBMUS
To: policy@raleigh.ibm.com, wg-service@dmtf.org
Message-Id: <8525679B.00663944.00@d54mta01.raleigh.ibm.com>
Date: Fri, 25 Jun 1999 14:28:08 -0400
Subject: Two New Core Policy Internet-Drafts
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: remoore@us.ibm.com



I've submitted the two Core Policy Internet-Drafts (i.e., the
information model and the LDAP representation of it) to the
Internet-Drafts repository.  The drafts are also available on
our AIW FTP site, at the following URLs:

ftp://ftp.networking.ibm.com/pub/standards/aiw/misc
     /draft-ietf-policy-core-info-model-00.txt

ftp://ftp.networking.ibm.com/pub/standards/aiw/misc
     /draft-ietf-policy-core-schema-04.txt

Send me a note if you have any trouble retrieving these.

Regards,
Bob

Bob Moore
IBM Networking Software
+1-919-254-4436
remoore@us.ibm.com




From majordomo@raleigh.ibm.com  Fri Jun 25 18:08:50 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA06733
	for <policy-archive@odin.ietf.org>; Fri, 25 Jun 1999 18:08:49 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id SAA07680;
	Fri, 25 Jun 1999 18:08:22 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id SAA29746;
	Fri, 25 Jun 1999 18:08:18 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA33514; Fri, 25 Jun 1999 17:54:18 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA53218; Fri, 25 Jun 1999 17:54:16 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id RAA31500
	for <policy@raleigh.ibm.com>; Fri, 25 Jun 1999 17:54:17 -0400
Received: from atlrel2.hp.com (atlrel2.hp.com [156.153.255.202])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA32010
	for <policy@raleigh.ibm.com>; Fri, 25 Jun 1999 17:54:08 -0400
Received: from xpeditio.cnd.hp.com (xpeditio.cnd.hp.com [15.2.113.211])
	by atlrel2.hp.com (8.8.6 (PHNE_17135)/8.8.5tis) with ESMTP id RAA24897;
	Fri, 25 Jun 1999 17:53:35 -0400 (EDT)
Received: (from mhugh@localhost) by xpeditio.cnd.hp.com (8.7.1/8.7.3 TIS 5.0) id PAA21019; Fri, 25 Jun 1999 15:53:46 -0600 (MDT)
From: "Hugh F. Mahon" <mhugh@xpeditio.cnd.hp.com>
Message-Id: <199906252153.PAA21019@xpeditio.cnd.hp.com>
Subject: Re: multiple clients involved in a policy
To: mohap@research.bell-labs.com (Pradosh Kumar Mohapatra)
Date: Fri, 25 Jun 1999 15:53:46 MDT
Cc: policy@raleigh.ibm.com
In-Reply-To: <377136E3.5E15F213@research.bell-labs.com>; from "Pradosh Kumar Mohapatra" at Jun 23, 99 12:34 (noon)
Organization:  HP Network & System Management Division
X-Mailer: Elm [revision: 212.2]
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: "Hugh F. Mahon" <mhugh@xpeditio.cnd.hp.com>

Hi Pradosh,

> Hi,
> Thanks for the answers to my previous questions regarding PEP and PDPs.
> From the documents that I read, I could come to the conclusion that the
> policy execution, right now is a single client-to-server communication
> where in, if some policy condition is triggered on the PEP, the PEP
> sends a COPS message to the PDP, the PDP evaluates it and sends the
> decision (some action to be taken based on the condition) to the PEP.
> I am not sure how the current framework will provide space to do the
> following:
> say a policy is entered by the administrator which states if event e1
> occurs in device d1, event e2 occurs in device d2, ...., then do some
> action a1 in d1, a3 in d3 ... etc.
> 
> e1, e2, .... can be events (transitions in the document by Silvano)
> occuring at the same time in d1, d2, ... or they occur in different time
> frames.
> 
> Does the current architecture support this?
> If so, then how?
> 
> Regards,
> Pradosh

As Andrew and Steve have pointed out, the PEP is on the device 
enforcing the policies.  The PDP can be on the device with the PEP or 
remote.  I agree with Steve's comments about scalability of PEPs and 
PDPs separated across the network. 

The question of how the PEP and PDP work and their roles in policy 
operations comes up quite often.  I think many people have ideas of how 
these things relate, but you are right that for general policy these 
components have not been well documented.  Here is my attempt at 
describing the components of a policy system :-).

The functional components can be broken out as follows:

        - policy UI (policy authoring/editing)
        - policy repository
        - policy server
        - policy interpreter
        - Policy Decision Point (PDP)
        - Policy Enforcement Point (PEP)

As far as implementation of a policy system, these functional 
components could all stand alone, or be grouped in various ways.  Much 
of the discussion has centered around the data stored in the 
repository.  A directory service using LDAP as its interface is the 
current focus for the role of repository.  The other components are 
largely undiscussed to date.  Of course there could be other 
components as part of a policy management system, those listed above 
are the minimal set.  An architecture which defined the interfaces 
between these components would allow extensibility in an interoperable 
way. 

The terms PDP and PEP derive from the work of the RAP working group 
[1].  The terms PEP and PDP are described in the document "A Framework 
For Policy-based Admission Control"[2].  The usage models described 
for these terms is strongly tied to a signaled QoS environment, which 
after all is what RSVP, the 'R' in RAP, is all about.  

These terms still make sense for other policy usage since decisions 
must be made and enforced (or acted on).  

One way to group components is to have the PDP and PEP on the same 
device, so that when a decision is required the activities are all 
local.  This allows for efficient operation if the device can support 
processing and storage of decision data, and all of the effects are 
local to the device.  Note that this description does not specify 
where the policy interpretation is located.  Policy interpretation may 
reside on a separate system and convert policy into a 'native' 
configuration format for a given device.  In this way, although a 
device would contain the ability to make decisions, it does not need 
to have the capability to deal with general policy descriptions.  The 
policy would be translated into the more easily consumed 
device-specific form before reaching the device.  Policy 
interpretation (or translation) may reside with the PDP or separately.

As noted in Steve's message, it is important to understand the cost to 
the network of sending decision requests and responses across the 
network.  PDPs which are remote from PEPs would likely be used for 
highly-granular activities.  Examples include the establishment of a 
long-term connection, such as for a teleconference.  Following Steve's 
example, if signaling were used to establish QoS for a user browsing 
the Web, it would make more sense for the PEP to make a request to the 
PDP about the first connection to the Web server, and the PEP to cache 
the answer and reuse the PDP's response for successive connections to 
the same Web server. 

A PEP is also not limited to using one PDP.  For example, a PEP may 
use a remote PDP for handling RSVP requests, but may use a local PDP 
for handling DiffServ-like operations. 

Of course, if the result of a policy decision is to affect more than 
one device (or a device other than the one where the decision request 
was initiated) it may make sense for the PDP to be remote from the 
PEPs.  How such a relationship would be described has not been defined.

Another interesting document to read is "Policy Framework Architecture" 
[3].

The Policy server is another component which has not been discussed in 
detail relative to the work of the Policy Framework WG.  A central 
repository is important, but there are other aspects of information 
management other than storage and retrieval which must occur between 
the policy authoring tool and the policy interpreter.  The policy 
server, and possibly other components (for such tasks as policy 
conflict detection, etc.), would fit in this space.

References:

[1] IETF RSVP Admission Protocol (RAP) Working Group, 
    http://www.ietf.org/html.charters/rap-charter.html 
[2] "A Framework For Policy-based Admission Control", Raj Yavatkar, 
    Dimitrios Pendarakis, Roch Guerin, 
    http://www.ietf.org/internet-drafts/draft-ietf-rap-framework-03.txt
[3] "Policy Framework Architecture", G. Waters, J. Wheeler, A. 
    Westerinen, L. Rafalow, R. Moore
    http://www.ietf.org/internet-drafts/draft-waters-policy-arch-00.txt


I realize that this leaves room for more questions, but I hope it also 
answers some questions.

Thanks,

Hugh Mahon              |__|
hugh_mahon@hp.com       |  |
                           |\  /|
                           | \/ |

 "We can lick gravity, but sometimes the paperwork is overwhelming."
                         -- Werner Von Braun



From majordomo@raleigh.ibm.com  Fri Jun 25 21:05:49 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA09152
	for <policy-archive@odin.ietf.org>; Fri, 25 Jun 1999 21:05:48 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id VAA33286;
	Fri, 25 Jun 1999 21:05:35 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id VAA33138;
	Fri, 25 Jun 1999 21:05:32 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA58012; Fri, 25 Jun 1999 20:54:55 -0400
Received: from rtpmail03.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA54164; Fri, 25 Jun 1999 20:54:52 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id UAA20634
	for <policy@raleigh.ibm.com>; Fri, 25 Jun 1999 20:54:53 -0400
Received: from sol.extremenetworks.com (sol.extremenetworks.com [216.52.8.2])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id UAA07874
	for <policy@raleigh.ibm.com>; Fri, 25 Jun 1999 20:54:51 -0400
Received: by ftp.extremenetworks.com with Internet Mail Service (5.5.2448.0)
	id <NJQFP1R5>; Fri, 25 Jun 1999 17:54:21 -0700
Message-Id: <D0805D3B448BD211A7990008C7B18130142F81@ftp.extremenetworks.com>
From: Andrew Smith <andrew@extremenetworks.com>
To: "'Hugh F. Mahon'" <mhugh@xpeditio.cnd.hp.com>
Cc: policy@raleigh.ibm.com
Subject: RE: multiple clients involved in a policy
Date: Fri, 25 Jun 1999 17:54:20 -0700
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Andrew Smith <andrew@extremenetworks.com>

And, just to confuse things, the "R" in RAP now stands for "Resource" not
"RSVP". We just re-chartered the WG to start looking at ways to communicate
different types of info between PDPs and PEPs - it's now the "Resource
Assignment Protocol" WG - we thought that backwards compatibility in TLAs
was a Good Thing even though the name could have been better.

BTW, the charter that is currently on the RAP web page referenced by Hugh
[1] needs to be fixed since right now it says we're going to some things
that the diffserv WG is actually going to do.

Andrew Smith
RAP WG co-chair

> -----Original Message-----
> From: Hugh F. Mahon [mailto:mhugh@xpeditio.cnd.hp.com]
> Sent: Friday, June 25, 1999 2:54 PM
> To: mohap@research.bell-labs.com
> Cc: policy@raleigh.ibm.com
> Subject: Re: multiple clients involved in a policy
> 
> 
> Hi Pradosh,
> 
> > Hi,
> > Thanks for the answers to my previous questions regarding 
> PEP and PDPs.
> > From the documents that I read, I could come to the 
> conclusion that the
> > policy execution, right now is a single client-to-server 
> communication
> > where in, if some policy condition is triggered on the PEP, the PEP
> > sends a COPS message to the PDP, the PDP evaluates it and sends the
> > decision (some action to be taken based on the condition) 
> to the PEP.
> > I am not sure how the current framework will provide space to do the
> > following:
> > say a policy is entered by the administrator which states 
> if event e1
> > occurs in device d1, event e2 occurs in device d2, ...., 
> then do some
> > action a1 in d1, a3 in d3 ... etc.
> > 
> > e1, e2, .... can be events (transitions in the document by Silvano)
> > occuring at the same time in d1, d2, ... or they occur in 
> different time
> > frames.
> > 
> > Does the current architecture support this?
> > If so, then how?
> > 
> > Regards,
> > Pradosh
> 
> As Andrew and Steve have pointed out, the PEP is on the device 
> enforcing the policies.  The PDP can be on the device with the PEP or 
> remote.  I agree with Steve's comments about scalability of PEPs and 
> PDPs separated across the network. 
> 
> The question of how the PEP and PDP work and their roles in policy 
> operations comes up quite often.  I think many people have 
> ideas of how 
> these things relate, but you are right that for general policy these 
> components have not been well documented.  Here is my attempt at 
> describing the components of a policy system :-).
> 
> The functional components can be broken out as follows:
> 
>         - policy UI (policy authoring/editing)
>         - policy repository
>         - policy server
>         - policy interpreter
>         - Policy Decision Point (PDP)
>         - Policy Enforcement Point (PEP)
> 
> As far as implementation of a policy system, these functional 
> components could all stand alone, or be grouped in various 
> ways.  Much 
> of the discussion has centered around the data stored in the 
> repository.  A directory service using LDAP as its interface is the 
> current focus for the role of repository.  The other components are 
> largely undiscussed to date.  Of course there could be other 
> components as part of a policy management system, those listed above 
> are the minimal set.  An architecture which defined the interfaces 
> between these components would allow extensibility in an 
> interoperable 
> way. 
> 
> The terms PDP and PEP derive from the work of the RAP working group 
> [1].  The terms PEP and PDP are described in the document "A 
> Framework 
> For Policy-based Admission Control"[2].  The usage models described 
> for these terms is strongly tied to a signaled QoS environment, which 
> after all is what RSVP, the 'R' in RAP, is all about.  
> 
> These terms still make sense for other policy usage since decisions 
> must be made and enforced (or acted on).  
> 
> One way to group components is to have the PDP and PEP on the same 
> device, so that when a decision is required the activities are all 
> local.  This allows for efficient operation if the device can support 
> processing and storage of decision data, and all of the effects are 
> local to the device.  Note that this description does not specify 
> where the policy interpretation is located.  Policy 
> interpretation may 
> reside on a separate system and convert policy into a 'native' 
> configuration format for a given device.  In this way, although a 
> device would contain the ability to make decisions, it does not need 
> to have the capability to deal with general policy descriptions.  The 
> policy would be translated into the more easily consumed 
> device-specific form before reaching the device.  Policy 
> interpretation (or translation) may reside with the PDP or separately.
> 
> As noted in Steve's message, it is important to understand 
> the cost to 
> the network of sending decision requests and responses across the 
> network.  PDPs which are remote from PEPs would likely be used for 
> highly-granular activities.  Examples include the establishment of a 
> long-term connection, such as for a teleconference.  
> Following Steve's 
> example, if signaling were used to establish QoS for a user browsing 
> the Web, it would make more sense for the PEP to make a 
> request to the 
> PDP about the first connection to the Web server, and the PEP 
> to cache 
> the answer and reuse the PDP's response for successive connections to 
> the same Web server. 
> 
> A PEP is also not limited to using one PDP.  For example, a PEP may 
> use a remote PDP for handling RSVP requests, but may use a local PDP 
> for handling DiffServ-like operations. 
> 
> Of course, if the result of a policy decision is to affect more than 
> one device (or a device other than the one where the decision request 
> was initiated) it may make sense for the PDP to be remote from the 
> PEPs.  How such a relationship would be described has not 
> been defined.
> 
> Another interesting document to read is "Policy Framework 
> Architecture" 
> [3].
> 
> The Policy server is another component which has not been 
> discussed in 
> detail relative to the work of the Policy Framework WG.  A central 
> repository is important, but there are other aspects of information 
> management other than storage and retrieval which must occur between 
> the policy authoring tool and the policy interpreter.  The policy 
> server, and possibly other components (for such tasks as policy 
> conflict detection, etc.), would fit in this space.
> 
> References:
> 
> [1] IETF RSVP Admission Protocol (RAP) Working Group, 
>     http://www.ietf.org/html.charters/rap-charter.html 
> [2] "A Framework For Policy-based Admission Control", Raj Yavatkar, 
>     Dimitrios Pendarakis, Roch Guerin, 
>     
http://www.ietf.org/internet-drafts/draft-ietf-rap-framework-03.txt
[3] "Policy Framework Architecture", G. Waters, J. Wheeler, A. 
    Westerinen, L. Rafalow, R. Moore
    http://www.ietf.org/internet-drafts/draft-waters-policy-arch-00.txt


I realize that this leaves room for more questions, but I hope it also 
answers some questions.

Thanks,

Hugh Mahon              |__|
hugh_mahon@hp.com       |  |
                           |\  /|
                           | \/ |

 "We can lick gravity, but sometimes the paperwork is overwhelming."
                         -- Werner Von Braun


From majordomo@raleigh.ibm.com  Fri Jun 25 21:25:46 1999
Received: from fwns1.raleigh.ibm.com (fwns1d.raleigh.ibm.com [204.146.167.235])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA09447
	for <policy-archive@odin.ietf.org>; Fri, 25 Jun 1999 21:25:46 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id VAA20986;
	Fri, 25 Jun 1999 21:25:33 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id VAA20172;
	Fri, 25 Jun 1999 21:25:28 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA54474; Fri, 25 Jun 1999 21:15:26 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA53186; Fri, 25 Jun 1999 21:15:24 -0400
Received: from fwns2.raleigh.ibm.com (fwns2.raleigh.ibm.com [9.37.0.4])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id VAA35230
	for <policy@raleigh.ibm.com>; Fri, 25 Jun 1999 21:15:26 -0400
Received: from sol.extremenetworks.com (sol.extremenetworks.com [216.52.8.2])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id VAA26782
	for <policy@raleigh.ibm.com>; Fri, 25 Jun 1999 21:15:24 -0400
Received: by ftp.extremenetworks.com with Internet Mail Service (5.5.2448.0)
	id <NJQFP1S6>; Fri, 25 Jun 1999 18:14:54 -0700
Message-Id: <D0805D3B448BD211A7990008C7B18130142F83@ftp.extremenetworks.com>
From: Andrew Smith <andrew@extremenetworks.com>
To: policy@raleigh.ibm.com
Subject: RE: Policy framework requirements
Date: Fri, 25 Jun 1999 18:14:53 -0700
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Andrew Smith <andrew@extremenetworks.com>

I think that's a good start to defining the problem at a high level - I'd
recommend that we try to merge Hugh's description in to the policy framework
draft somewhere.

Andrew

> -----Original Message-----
> From: Hugh F. Mahon [mailto:mhugh@xpeditio.cnd.hp.com]
> Sent: Thursday, June 24, 1999 11:44 AM
> To: policy@raleigh.ibm.com
> Subject: Policy framework requirements
> 
> 
> Hi Folks,
> 
> Within this working group there have been a lot of 
> discussions of what 
> kind of policy information is needed, and there's been a lot 
> of work on 
> schema for policy information.  There has also been a lot of 
> discussion 
> of how that information is to be used, and how it is shared.
> 
> Each of the participants has had in mind a model for how things work, 
> but we rarely discuss this aspect of policy management.  If we are to 
> indeed develop standards around which we can have 
> interoperable policy 
> management tools we need to have discussions of how these mechanisms 
> will work together.
> 
> I am specifically not talking about how a UI looks or how policy is 
> verified to be usable.  
> 
> I am talking about how policy gets from the administrator to where it 
> is actually used, what the administrator (customer) expects from the 
> policy management system, and the minimum pieces necessary to provide 
> a functional system.  Without this understanding I do not see how we 
> can define schema and how it is to be used that allows true 
> interoperability (or value to customers).
> 
> To begin the discussion I've put together a simple 
> description of what 
> an administrator would do and expect.  The purpose is to drive toward 
> common understanding and help in determining the solutions (or even 
> allow us to choose which things to solve and what not to solve).
> 
> The Scenario
> ------------
> 
> An administrator must author policy.  This can be done either from 
> scratch or using existing policies (either editing existing policies 
> or copying them and using them for creating new policies).  Once new 
> policies have been authored, the administrator must associate them 
> with the managed entities the policies are to affect.  The 
> administrator can either 'just know' to what policies can be 
> associated, or a list can be presented.  There are other reasons that 
> such a list would be valuable in the management system (see below).
> 
> Because policy can be complex, or have information the administrator 
> may wish to track over time, it would be valuable to allow versioning 
> of policy information, that is, earlier versions of policy 
> information 
> would be stored.  This versioning allows administrators to make 
> changes without fear that valuable information will be irretrievably 
> lost.  One example of this value is being able to quickly 
> roll-back to 
> a previous version of policy if the new policy is found to be 
> incorrect. 
> 
> Once the policy has been created or edited by the administrator, it 
> should be stored in a repository.  (Note that it is this repository 
> that has been the focus of much of the group's discussions and is 
> viewed as the central interoperability point.)  The managed entity 
> should then receive the policy in a timely fashion and implement the 
> policy.  Once the policy has been implemented on the managed entity 
> the administrator should be notified of the status of the policy 
> deployment: success, success with warnings, or failure (with the 
> reason(s) for the failure).  This is an area where the list 
> of managed 
> entities would provide value: a place to associate status information 
> about policy on those managed entities. 
> 
> Administrator Expectations
> --------------------------
> 
> For example, if a network is experiencing poor performance, the 
> administrator may choose to change policy.  Say a new tool for 
> receiving news has been introduced, and it is using significant 
> amounts of bandwidth.  We'll call it NewTool.  The administrator 
> cannot stop all users from using NewTool, but can manage the QoS for 
> the traffic it generates.  To improve the performance of other 
> applications the administrator chooses to reduce the priority of 
> NewTool.  Customers are clamoring for improved performance, so the 
> administrator would like the changes to take effect immediately.  It 
> is not desirable to wait an hour for all of the affected devices to 
> query a central repository and learn that new policy exists.  
>  Neither 
> is it desirable to have each of the managed entities generating 
> extraneous traffic to check for policy changes, adding a further 
> burden to the already overworked network. 
> 
> In addition, the administrator wants to be notified as soon as the 
> policy is in place on each of the affected managed entities.  Without 
> such feedback, the administrator is not in control of the environment 
> but at the mercy of the policy management system.
> 
> The Requirements
> ----------------
> 
> Based on the simple scenario described above, the requirements for a 
> policy management system should include:
> 
>           - storage of policy information        
>           - storage of information about managed entities 
>           - an association between policies and what they are 
> to manage
>           - timely asynchronous notification of new/changed policy 
>             information
>           - ability to provide feedback to administrator of policy 
>             status at managed entity
>           - transactional qualities (the administrator can determine 
>             that the changes were the result of a specific 
>             administrator action)
>           - definition of how policy information is to be used on 
>             managed entities (e.g., how many actions are 
> performed for 
>             any event, etc.)
>           - ensure that policies put in directory are what is read by 
>             the PDP/interpreter
> 
> LDAP (and directories accessed by LDAP) can't do all of these things, 
> at least not as it exists today (or in the near future).  Other 
> mechanisms need to be in place to supplement LDAP's capabilities.
> 
> I agree with the Area Directors' concerns about adding more to the 
> Working Group's tasks before the WG achieves its initial goals (i.e., 
> a schema), but I can't help but feel that we will miss important 
> things if we don't discuss how all of this is supposed to work.  We 
> don't have to define all of it before we get some pieces firmed up, 
> but we do need to discuss these other aspects more before we can call 
> anything done or ready for use.
> 
> 
>         Hugh Mahon                              |__|     
>         Hewlett-Packard Company                 |  |     
>         e-mail: hugh_mahon@hp.com                  |\  /|
>         Phone: (970) 898-BITS (898-2487)           | \/ |     
> 


From majordomo@raleigh.ibm.com  Fri Jun 25 23:13:00 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id XAA12861
	for <policy-archive@odin.ietf.org>; Fri, 25 Jun 1999 23:13:00 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id XAA07882;
	Fri, 25 Jun 1999 23:12:48 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id XAA28848;
	Fri, 25 Jun 1999 23:12:31 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA53362; Fri, 25 Jun 1999 23:02:50 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA46418; Fri, 25 Jun 1999 23:02:46 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id XAA36198
	for <policy@raleigh.ibm.com>; Fri, 25 Jun 1999 23:02:49 -0400
Received: from bmailnj.iphighway.com ([209.3.6.76])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id XAA27314
	for <policy@raleigh.ibm.com>; Fri, 25 Jun 1999 23:02:46 -0400
Received: by BMAILNJ with Internet Mail Service (5.5.2448.0)
	id <KN70SVW5>; Fri, 25 Jun 1999 22:56:45 -0400
Message-Id: <6399122981E1D211AB490090271E0AA313338A@BMAILNJ>
From: Francis Reichmeyer -NJ <FranR@iphighway.com>
To: "'Andrew Smith'" <andrew@extremenetworks.com>, policy@raleigh.ibm.com
Subject: RE: Policy framework requirements
Date: Fri, 25 Jun 1999 22:56:43 -0400
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Francis Reichmeyer -NJ <FranR@iphighway.com>


-----Original Message-----
From: Andrew Smith [mailto:andrew@extremenetworks.com]
Sent: Friday, June 25, 1999 9:15 PM
To: policy@raleigh.ibm.com
Subject: RE: Policy framework requirements


I think that's a good start to defining the problem at a high level - I'd
recommend that we try to merge Hugh's description in to the policy framework
draft somewhere.

	[franr] And which policy framework draft would that be? :^)

Andrew

-Fran


From majordomo@raleigh.ibm.com  Mon Jun 28 16:33:10 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA12679
	for <policy-archive@odin.ietf.org>; Mon, 28 Jun 1999 16:33:10 -0400 (EDT)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id QAA31022;
	Mon, 28 Jun 1999 16:32:29 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id QAA29868;
	Mon, 28 Jun 1999 16:30:51 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA254244; Mon, 28 Jun 1999 16:20:55 -0400
Received: from rtpmail01.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA246174; Mon, 28 Jun 1999 16:19:12 -0400
Received: from ellesson.raleigh.ibm.com (ss02.raleigh.ibm.com [9.37.3.88])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id QAA35690;
	Mon, 28 Jun 1999 16:16:29 -0400
Message-Id: <199906282016.QAA35690@rtpmail01.raleigh.ibm.com>
X-Sender: ellesson@dnjake2.raleigh.ibm.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0
Date: Mon, 28 Jun 1999 16:16:55 -0400
To: policy@raleigh.ibm.com
From: Ed Ellesson <ellesson@raleigh.ibm.com>
Subject: Posting of Inappropriate Draft to Policy Frameworks Home Page
Cc: internet-drafts@ietf.org, stevenju@microsoft.com
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Ed Ellesson <ellesson@raleigh.ibm.com>

Dear Policy Enthusiasts:  

Please take note that the document recently posted to the IETF Policy
Frameworks home page, titled "Designing Directory Schemata for General
Use", was posted there by mistake, and the file name should not have
included the word "policy", which is the shortform name of our working
group.  This document is an individual submission.  To quote from the
appropriate IETF guidelines: "you must have the consent of your working
group chair, and the working charter needs to have a corresponding work
item, to consider your document a working group document."  We, the chairs,
consider this document to be out of scope as a work item for the policy
frameworks working group, since it attempts to address general schema
design principles, and not just those relevant to policy. 

John and Ed
Co-chairs, Policy Framework Working Group  

 

  




From majordomo@raleigh.ibm.com  Mon Jun 28 17:03:42 1999
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA13274
	for <policy-archive@odin.ietf.org>; Mon, 28 Jun 1999 17:03:41 -0400 (EDT)
Received: from rtpmail01.raleigh.ibm.com (rtpmail01.raleigh.ibm.com [9.37.172.24])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id RAA31342;
	Mon, 28 Jun 1999 17:02:52 -0400
Received: from rtpaix11.raleigh.ibm.com (rtpaix11.raleigh.ibm.com [9.37.172.4])
	by rtpmail01.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with SMTP id RAA33490;
	Mon, 28 Jun 1999 17:02:49 -0400
Received: by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA151554; Mon, 28 Jun 1999 16:54:11 -0400
Received: from rtpmail02.raleigh.ibm.com by rtpaix11.raleigh.ibm.com (AIX 4.1/UCB 5.64/4.03-RAL)
          id AA198900; Mon, 28 Jun 1999 16:54:08 -0400
Received: from fwns1.raleigh.ibm.com (fwns1.raleigh.ibm.com [9.37.0.3])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id QAA36520
	for <policy@raleigh.ibm.com>; Mon, 28 Jun 1999 16:54:11 -0400
Received: from mail5.microsoft.com (mail5.microsoft.com [131.107.3.121])
	by fwns1.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with SMTP id QAA28336
	for <policy@raleigh.ibm.com>; Mon, 28 Jun 1999 16:54:07 -0400
Received: from 157.54.9.108 by mail5.microsoft.com (InterScan E-Mail VirusWall NT); Mon, 28 Jun 1999 13:48:33 -0700 (Pacific Daylight Time)
Received: by INET-IMC-05 with Internet Mail Service (5.5.2524.0)
	id <N5JFGHQ6>; Mon, 28 Jun 1999 13:48:32 -0700
Message-Id: <B9D1827FDF66D111925800805F3102E30B381BFB@RED-MSG-57>
From: Steven Judd <stevenju@microsoft.com>
To: "'Ed Ellesson'" <ellesson@raleigh.ibm.com>
Cc: policy@raleigh.ibm.com, internet-drafts@ietf.org
Subject: RE: Posting of Inappropriate Draft to Policy Frameworks Home Page
Date: Mon, 28 Jun 1999 13:48:27 -0700
X-Mailer: Internet Mail Service (5.5.2524.0)
Sender: policy-owner@raleigh.ibm.com
Precedence: bulk
Reply-To: Steven Judd <stevenju@microsoft.com>

Ed -

It was certainly not my intent to misrepresent this draft as a WG document.
To quote from the IESG guidelines on document titles
(http://www.ietf.org/ID-nits.html). flagged with ->>>:

==========================================
Notes on the IESG review procedure
Naming your Internet Draft
The IETF generally uses this convention for Internet Draft file names. NN is
a sequence number, starting from 00 and monotonically increasing. 
  
Independent submissions: 
Draft-<author name>-<descriptive text>-NN.txt 
Example: draft-baker-spam-what-am-00.txt
 
->>> Independent submissions to a working group 
->>> Draft-<author name>-<working group>-<descriptive text>-NN.txt 
->>> Example: draft-baker-spamwg-spam-what-am-not-00.txt 

Working Group documents 
Draft-ietf-<working group>-descriptive text>-NN.txt 
Example: draft-ietf-spamwg-how-to-send-spam-00.txt 

You must have the consent of your working group chair, and the working
charter needs to have a corresponding work item, to consider your document a
working group document. Individual submissions (whether general or to a
working group) may be made without prior notice, but unless you ask someone
to do something with them, nothing further will be done.
============================================

As you know, I did ask you and John about considering this draft, and you
declined since you felt it was out of scope.  However, since I thought it
would be of interest to the Policy Framework WG, I submitted it as an
individual submission to the WG, per the IESG guidelines, and made this
clear in my message to the RFC editor (included below).  You will note that
the name requested is in the form required for individual submissions.

tx,

-s


+++++++++++++++++++++++++++++++++++++++++++++++++++
The attached Informational Internet-draft,
draft-judd-policy-ldap-schema-design-00.txt, is an individual submission to
the Policy Framework working group.

Abstract 

   This document is intended to aid in the design of directory schemata 
   for implementation on diverse directory services that support the 
   LDAP (RFC 2251 et al.) protocol as their primary access mechanism.  
   It is not specific to any directory service implementation. 

Thanks,

Steven G. Judd
Microsoft Corporation
++++++++++++++++++++++++++++++++++++++++++++++++++++


-----Original Message-----
From: Ed Ellesson [mailto:ellesson@raleigh.ibm.com]
Sent: Monday, June 28, 1999 1:17 PM
To: policy@raleigh.ibm.com
Cc: internet-drafts@ietf.org; Steven Judd
Subject: Posting of Inappropriate Draft to Policy Frameworks Home Page


Dear Policy Enthusiasts:  

Please take note that the document recently posted to the IETF Policy
Frameworks home page, titled "Designing Directory Schemata for General
Use", was posted there by mistake, and the file name should not have
included the word "policy", which is the shortform name of our working
group.  This document is an individual submission.  To quote from the
appropriate IETF guidelines: "you must have the consent of your working
group chair, and the working charter needs to have a corresponding work
item, to consider your document a working group document."  We, the chairs,
consider this document to be out of scope as a work item for the policy
frameworks working group, since it attempts to address general schema
design principles, and not just those relevant to policy. 

John and Ed
Co-chairs, Policy Framework Working Group  

 

  



