From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct  6 11:36:41 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA11721;
	Wed, 6 Oct 1999 11:36:38 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id LAA28483
	for nasreq-outgoing; Wed, 6 Oct 1999 11:08:55 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from smtpgw02.wanmail.net (smtpgw02.wanmail.net [209.154.194.61])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id LAA28479
	for <nasreq@tdmx.rutgers.edu>; Wed, 6 Oct 1999 11:08:53 -0400 (EDT)
Received: from wcomexch1.wcom.net (wcomexch1.wcom.net [209.154.236.12])
	by smtpgw02.wanmail.net (8.9.1/8.9.1) with ESMTP id LAA13338
	for <nasreq@tdmx.rutgers.edu>; Wed, 6 Oct 1999 11:08:17 -0400 (EDT)
Received: by wcomexch1.wcom.net with Internet Mail Service (5.5.2448.0)
	id <SAX9FB7R>; Wed, 6 Oct 1999 11:07:19 -0400
Message-ID: <9A74C89D6696D2118E500008C7BA7C560C03B16A@wcomexch1.wcom.net>
From: "Beadles, Mark A." <MBeadles@wcom.net>
To: "Nasreqng List (E-mail)" <nasreq@tdmx.rutgers.edu>
Subject: nasreq New Criteria Draft
Date: Wed, 6 Oct 1999 11:07:13 -0400 
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: multipart/mixed;
	boundary="----_=_NextPart_000_01BF100C.7B99BA6C"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.

------_=_NextPart_000_01BF100C.7B99BA6C
Content-Type: text/plain;
	charset="iso-8859-1"

I have posted draft-ietf-nasreq-criteria-02.txt, attached.  This is the
latest version of the NAS Requirements document.  There are some significant
changes from the last version, so please read it carefully.

--------------------

Change log:

-Incorporated items from Glen Zorn's YAAP document.
-Incorporated items from Dave Mitton's RADIUS Practices document.
-Incorporated various comments from the WG.
-Organizational changes: 
 * more granularity in section numbering
 * separation of protocol requirements, attribute requirements, and security
requirements

--------------------

Here is a condensed outline of the AAA requirements from this document:

     8.  AAA Protocol Requirements
     8.1.  General protocol characteristics
     8.1.1.  Transport Requirements
     8.1.2.  Attribute-Value Protocol Model
     8.1.3.  Security Requirements
     8.2.  Authentication and User Security Requirements
     8.2.1.  Authentication protocol requirements
     8.2.2.  Authentication Attribute Requirements
     8.2.3.  Authentication Protocol Security Requirements
     8.3.  Authorization, Policy, and Resource management
     8.3.1.  Authorization Protocol Requirements
     8.3.2.  Authorization Attribute Requirements
     8.3.3.  Authorization Protocol Security Requirements
     8.4.  Accounting and Auditing Requirements
     8.4.1.  Accounting Protocol Requirements
     8.4.2.  Accounting attribute requirements
     8.4.3.  Accounting Protocol Security Requirements

And for the masochists among you, here is a complete outline of the
document:

     1.  Status of this Memo
     2.  Copyright Statement
     3.  Abstract
     4.  Requirements language
     5.  Introduction
     6.  Access Protocol Requirements
     7.  Network Protocol Requirements
     8.  AAA Protocol Requirements
     8.1.  General protocol characteristics
     8.1.1.  Transport requirements
     8.1.1.1.  Transport independence
     8.1.1.2.  Long-term exchange of small packets
     8.1.1.3.  Support for multiple AAA servers
     8.1.1.4.  Support for Multiple Administrative Domains
     8.1.2.  Attribute-Value Protocol Model
     8.1.2.1.  Attribute Data Types
     8.1.2.2.  Minimum Set of Attributes
     8.1.2.3.  Attribute Extensibility
     8.1.3.  Security Requirements
     8.1.3.1.  Mutual Authentication
     8.1.3.2.  Shared Secrets
     8.1.3.3.  Public Key Security
     8.1.3.4.  Encryption of Attributes
     8.2.  Authentication and User Security Requirements
     8.2.1.  Authentication protocol requirements
     8.2.1.1.  Bi-directional Authentication
     8.2.1.2.  Dynamic Authentication
     8.2.1.3.  Multi-phase Authentication
     8.2.1.4.  Extensible Authentication Types
     8.2.2.  Authentication Attribute Requirements
     8.2.2.1.  PPP Authentication protocols
     8.2.2.2.  User Identification
     8.2.2.3.  Authentication Credentials
     8.2.3.  Authentication Protocol Security Requirements
     8.2.3.1.  End-to-End Hiding of Credentials
     8.3.  Authorization, Policy, and Resource management
     8.3.1.  Authorization Protocol Requirements
     8.3.1.1.  Dynamic Authorization
     8.3.1.2.  Resource Management
     8.3.2.  Authorization Attribute Requirements
     8.3.2.1.  Authorization Attribute Requirements - Access Restrictions
     8.3.2.2.  Authorization Attribute Requirements - Authorization Profiles
     8.3.2.3.  Resource Management Requirements
     8.3.3.  Authorization Protocol Security Requirements
     8.3.3.1.  Security of Compulsory Tunnel Credentials
     8.4.  Accounting and Auditing Requirements
     8.4.1.  Accounting Protocol Requirements
     8.4.1.1.  Guaranteed Delivery
     8.4.1.2.  Real Time Accounting
     8.4.1.3.  Batch Accounting
     8.4.1.4.  Accounting Time Stamps
     8.4.1.5.  Accounting Events
     8.4.1.6.  On-demand Accounting
     8.4.2.  Accounting attribute requirements
     8.4.3.  Accounting Protocol Security Requirements
     8.4.3.1.  Integrity and Confidentiality
     8.4.3.2.  Non-repudiation
     9.  Device Management Protocols
     10.  Acknowledgments
     11.  Security considerations
     12.  IANA Considerations
     13.  References
     14.  Author's Address
     15.  Full Copyright Statement
     16.  Expiration Date

 <<draft-ietf-nasreq-criteria-02.txt>> 

+ Mark Anthony Beadles + Co-chair, IETF NASREQ WG   + 
+  mbeadles@wcom.net   + http://nasreq.rutgers.edu/ +
+  Voice 614.723.1941  +       Fax 614.723.8407     +




------_=_NextPart_000_01BF100C.7B99BA6C
Content-Type: text/plain;
	name="draft-ietf-nasreq-criteria-02.txt"
Content-Disposition: attachment;
	filename="draft-ietf-nasreq-criteria-02.txt"
Content-Transfer-Encoding: quoted-printable

     NASREQ Working Group                              M. Beadles
     INTERNET-DRAFT                UUNET, an MCI WorldCom Company
     Category: Informational
     <draft-ietf-nasreq-criteria-02.txt>
     6 October 1999


            Criteria for Evaluating Network Access Server Protocols


     1.  Status of this Memo


     This document is an Internet-Draft and is in full conformance with =
all
     provisions of Section 10 of RFC2026.  Internet-Drafts are working =
doc-
     uments  of  the Internet Engineering Task Force (IETF), its areas, =
and
     its working groups.  Note that other groups may also distribute  =
work-
     ing documents as Internet-Drafts.

     Internet-Drafts  are draft documents valid for a maximum of six =
months
     and may be updated, replaced, or obsoleted by other documents  at  =
any
     time.   It is inappropriate to use Internet- Drafts as reference =
mate-
     rial or to cite them other than as "work in progress."

     The   list   of   current   Internet-Drafts   can   be   accessed  =
 at
     http://www.ietf.org/ietf/1id-abstracts.txt

     The  list  of  Internet-Draft  Shadow  Directories  can be =
accessed at
     http://www.ietf.org/shadow.html.

     The  distribution  of  this  draft  is  unlimited.   It  is  filed =
 as
     <draft-ietf-nasreq-criteria-02.txt>   and   expires  April  06,  =
2000.
     Please send comments to the author.


     2.  Copyright Statement


     Copyright   (C) The Internet Society 1999.  All Rights Reserved.


     3.  Abstract


     This document defines  requirements  for  protocols  used  by  =
Network
     Access  Servers  (NAS).   Protocols  used by NAS's may be divided =
into
     four spaces:  Access protocols, Network protocols, AAA protocols,  =
and
     Management  protocols.  Primary attention is given to setting =
require-
     ments for AAA protocols, since that space is currently the least  =
well
     defined.







     Beadles                 Category: Informational               =
[Page 1]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     4.  Requirements language


     In  this document, the key words "MAY", "MUST, "MUST NOT", =
"optional",
     "recommended", "SHOULD", and "SHOULD NOT", are to  be  interpreted =
 as
     described in [KEYWORDS].


     5.  Introduction


     This  document  defines  requirements  for  protocols  used by =
Network
     Access Servers (NAS).Protocols used by NAS's may be divided into  =
four
     spaces:   Access  protocols,  Network  protocols,  AAA  protocols, =
and
     Device Management protocols.   The primary focus of this  document =
 is
     on AAA protocols.  The reference model of a NAS used by this =
document,
     and the analysis of the functions of a NAS which led to  the  =
develop-
     ment of these requirements, may be found in [NAS-MODEL].


     6.  Access Protocol Requirements


     There  are three basic types of access protocols used by NAS's.  =
First
     are the traditional telephony-based access protocols, which  =
interface
     to  the  NAS  via a modem or terminal adapter or similar device. =
These
     protocols typically support asynchronous or synchronous PPP [PPP] =
car-
     ried over a telephony protocol.  Second are broadband =
pseudo-telephony
     access protocols, which are carried over xDSL  or  cable  modems,  =
for
     example.   These  protocols  typically support an encapsulation =
method
     such as PPP over Ethernet [PPPOE].  Finally  are  the  virtual  =
access
     protocols  used  by NAS's that terminate tunnels.  One example of =
this
     type of protocol is L2TP [L2TP].

     It is a central assumption of the NAS  model  used  here  that  a  =
NAS
     accepts multiple point-to-point links via one of the above access =
pro-
     tocols.  Therefore, at a minimum, any NAS access protocol MUST be =
able
     to  carry  PPP.   The  exception to this requirement is for NAS's =
that
     support legacy text login methods such as telnet [TELNET], rlogin, =
 or
     LAT.   Only  these access protocols are exempt from the =
requirement to
     support PPP.


     7.  Network Protocol Requirements


     The network protocols supported by a NAS depend entirely on  the  =
kind
     of network to which a NAS is providing access.  This document does =
not
     impose any additional requirements on  network  protocols  beyond  =
the
     protocol specifications themselves.  For example, if a NAS that =
serves
     a routed network includes internet routing  functionality,  then  =
that
     NAS must adhere to [ROUTING-REQUIREMENTS], but there are no =
additional
     protocol requirements imposed by virtue of the  device  being  a  =
NAS.




     Beadles                 Category: Informational               =
[Page 2]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     8.  AAA Protocol Requirements



     8.1.  General protocol characteristics


     There  are  certain general characteristics that any AAA protocol =
used
     by NAS's must meet.  Note that the transport requirements for  =
authen-
     tication/authorization  are  not  necessarily  the  same  as those =
for
     accounting/auditing.  An AAA protocol suite MAY use the same =
transport
     and protocol for both functions, but this is not strictly =
required.


     8.1.1.  Transport requirements



     8.1.1.1.  Transport independence

     The AAA protocol MUST be transport independent, and MUST be =
capable of
     using both TCP and UDP as a transport.   Additionally, the AAA  =
proto-
     col  SHOULD  be designed to easily support any new transport =
protocols
     developed by the Internet standards community.


     8.1.1.2.  Long-term exchange of small packets


     Very large scale NAS's that serve up to thousands of simultaneous =
ses-
     sions  are now being deployed.  This means that, in the extreme, =
there
     may be an almost constant exchange of many small packets  between  =
the
     NAS  and  the  AAA  server.   An AAA protocol transport SHOULD =
support
     being optimized for a long-term exchange of small packets in a  =
stream
     between a pair of hosts.


     8.1.1.3.  Support for multiple AAA servers


     In  order to operationally support large loads, load balancing to =
mul-
     tiple AAA servers will be required.  The AAA protocol MUST provide =
for
     NAS's  to  balance  AAA sessions between two or more AAA servers.  =
The
     load balancing mechanism SHOULD  be  built  in  to  the  AAA  =
protocol
     itself.   The AAA protocol design MUST NOT introduce a single =
point of
     failure during the AAA process.  The AAA protocol MUST allow any  =
ses-
     sions between a NAS and a given AAA server to fail over to a =
secondary
     server without loss of state information.   This  fail-over  =
mechanism
     SHOULD be built in to the AAA protocol itself.








     Beadles                 Category: Informational               =
[Page 3]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     8.1.1.4.  Support for Multiple Administrative Domains


     NAS's  operated  by  one authority provide network access services =
for
     clients operated by another authority, to network  destinations  =
oper-
     ated by yet another authority.  This type of arrangement is of =
growing
     importance; for example, dial roaming is now a nearly ubiquitous  =
ser-
     vice.   Therefore,  the  AAA  protocol  MUST support AAA services =
that
     travel between multiple domains of authority.  The AAA  protocol  =
MUST
     NOT  use  a  model that assumes a single domain of authority.  The =
AAA
     protocol MUST NOT dictate particular business models for the =
relation-
     ship  between  the administrative domains.  The AAA protocol MUST =
sup-
     port proxy, and in additiona SHOULD support other  multi-domain  =
rela-
     tionships such as brokering and referral.

     The AAA protocol MUST also meet the protocol requirements =
specified in
     [ROAMING-REQUIREMENTS].


     8.1.2.  Attribute-Value Protocol Model


     Years of operational experience  with  AAA  protocols  and  NAS's  =
has
     proven that the Attribute-Value protocol model is an optimal =
represen-
     tation of AAA data.  The protocol SHOULD use an Attribute-Value =
repre-
     sentation for AAA data.  This document will assume such a model.  =
Even
     if the AAA protocol does not use this as an on-the-wire data =
represen-
     tation,  Attribute-Value  can  serve as abstraction for discussing =
AAA
     information.

     Experience has also shown  that  attribute  space  tends  to  run  =
out
     quickly.  In  order  to  provide  room  for expansion in the =
attribute
     space, the AAA protocol MUST support a minimum of 64K  Attributes  =
(16
     bits), each with a minimum length of 64K (16 bits).


     8.1.2.1.  Attribute Data Types


     The  AAA  protocol MUST support simple attribute data types, =
including
     integer, enumeration, text string, IP address, and date/time. The  =
AAA
     protocol  MUST  also  provide some support for complex structured =
data
     types. Wherever IP addresses are carried within the AAA protocol,  =
the
     protocol  MUST  support  both IPv4 and IPv6 [IPV6] addresses. =
Wherever
     text information is carried within the AAA protocol, the protocol =
MUST
     comply  with  the  IETF  Policy  on  Character Sets and Languages =
[RFC
     2277].


     8.1.2.2.  Minimum Set of Attributes


     At a minimum, the AAA protocol MUST support, or be easily extended =
 to
     support, the set of attributes supported by RADIUS [RADIUS] and =
RADIUS



     Beadles                 Category: Informational               =
[Page 4]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     Accounting [RADIUS-ACCOUNTING].  If the base  AAA  protocol  does  =
not
     support  this  complete  set  of attributes, then an extension to =
that
     protocol MUST be defined which supports this set.


     8.1.2.3.  Attribute Extensibility


     NAS and AAA development is always progressing.  In  order  to  =
prevent
     the  AAA  protocol  from being a limiting factor in NAS and AAA =
Server
     development, the AAA protocol MUST provide  a  built-in  =
extensibility
     mechanism,  which  MUST  include  a  means  for  adding  new  =
standard
     attribute extensions.  This MUST include a method for  registering =
 or
     requesting  extensions  through  IANA, so that long-term working =
group
     involvement is not required to create new attribute  types.   =
Ideally,
     the  AAA  protocol SHOULD separate specification of the transport =
from
     specificatoin of the attributes.

     The AAA protocol MUST include a means for individual  vendors  to  =
add
     value  through  vendor-specific  attributes and SHOULD include =
support
     for vendor-specific data types.


     8.1.3.  Security Requirements



     8.1.3.1.  Mutual Authentication


     It is poor security practice for a NAS  to  communicate  with  an  =
AAA
     server  that  is  not  trusted, and vice versa.  The AAA protocol =
MUST
     provide mutual authentication between AAA server and NAS.


     8.1.3.2.  Shared Secrets


     At a minimum, the AAA protocol SHOULD support use of a  secret  =
shared
     pairwise  between each NAS and AAA server to mutually verify =
identity.
     This is intended for small-scale deployments.


     8.1.3.3.  Public Key Security


     AAA server/NAS identity verification based solely  on  shared  =
secrets
     can  be  difficult  to  deploy  properly at large scale, and it =
can be
     tempting for NAS operators to use a single shared secret (that  =
rarely
     changes)  across  all  NAS's.  This can lead to easy compromise of =
the
     secret.  Therefore, the AAA protocol MUST also support mutual  =
verifi-
     cation  of  identity  using  a public-key infrastructure that =
supports
     expiration and revocation of keys.




     Beadles                 Category: Informational               =
[Page 5]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     8.1.3.4.  Encryption of Attributes


     Some attributes are more operationally sensitive than  others.   =
Also,
     in a multi-domain scenario, attributes may be inserted by servers =
from
     different administrative domains. Therefore,  the  AAA  protocol  =
MUST
     support   selective  encryption  of  attributes  on  an  =
attribute-by-
     attribute basis,  even  within  the  same  message.  This  =
requirement
     applies equally to Authentication, Authorization, and Accounting =
data.


     8.2.  Authentication and User Security Requirements



     8.2.1.  Authentication protocol requirements


     End users who are requesting network access through a NAS will =
present
     various  types  of credentials.  It is the purpose of the AAA =
protocol
     to transport these credentials between the NAS and the AAA server.


     8.2.1.1.  Bi-directional Authentication


     The AAA protocol MUST support transport of credentials  from  the  =
AAA
     server to the NAS for the purpose of mutual (bi-directional) =
authenti-
     cation between the User and the NAS, and between the NAS and  the  =
AAA
     server.


     8.2.1.2.  Dynamic Authentication


     The AAA protocol MUST support re-authentication at any time during =
the
     course of a session, initiated from either end of the user =
session.


     8.2.1.3.  Multi-phase Authentication


     The AAA protocol MUST be able to  support  multi-phase  =
authentication
     methods, including but not limited to support for:
          -Text prompting from the NAS to the user

          -A series of binary challenges and responses of arbitrary =
length

          -An  authentication failure reason to be transmitted from the =
NAS
          to the user

          -Callback to a pre-determined phone number





     Beadles                 Category: Informational               =
[Page 6]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     8.2.1.4.  Extensible Authentication Types


     Security protocol development is going on constantly  as  new  =
threats
     are  identified  and  better  cracking methods are developed.  =
Today's
     secure authentication methods may be proven  insecure  tomorrow.   =
The
     AAA protocol MUST provide some support for addition of new =
authentica-
     tion credential types.


     8.2.2.  Authentication Attribute Requirements


     In addition to the minimum attribute set, the AAA protocol  must  =
sup-
     port and define attributes that provide the following functions:


     8.2.2.1.  PPP Authentication protocols


     Many authentication protocols are defined within the framework of =
PPP.
     The AAA protocol MUST be able  to  act  as  an  intermediary  =
protocol
     between  the  authenticatee  and  the  authenticator for the =
following
     authentication protocols:

          -PPP Password Authentication Protocol [PPP]

          -PPP Challenge Handshake Authentication Protocol [CHAP]

          -PPP Extensible Authentication Protocol [EAP]


     8.2.2.2.  User Identification


     The following are common types of credentials used for user  =
identifi-
     cation.  The AAA protocol MUST be able to carry the following =
types of
     identity credentials:
          -A user name in the form of a Network Access Identifier =
[NAI].

          -An Extensible Authentication  Protocol  [EAP]  Identity  =
Request
          Type packet.

          -Telephony  dialing information such as Dialed Number =
Identifica-
          tion Service (DNIS) and Caller ID.
     If a particular type of identity credential is not needed for  a  =
par-
     ticular  user  session,  the  AAA protocol MUST NOT require that =
dummy
     credentials be filled in.  That is,  the  AAA  protocol  MUST  =
support
     identification without authentication.








     Beadles                 Category: Informational               =
[Page 7]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     8.2.2.3.  Authentication Credentials


     The following are common types of credentials used for =
authentication.
     The AAA protocol MUST be able to carry the following types of  =
authen-
     ticating credentials at a minimum:
          -A secret or password.

          -A response to a challenge presented by the NAS to the user

          -A one-time password

          -An X.509 digital certificate [X.509]

          -A Kerberos v5 ticket [KERBEROS]


     8.2.3.  Authentication Protocol Security Requirements



     8.2.3.1.  End-to-End Hiding of Credentials


     Where passwords are used as authentication credentials, the AAA =
proto-
     col MUST provide a secure means of hiding the password from end to =
end
     of  the  AAA  conversation.   Where  challenge/response mechanisms =
are
     used, the AAA protocol MUST also prevent against replay attacks.


     8.3.  Authorization, Policy, and Resource management



     8.3.1.  Authorization Protocol Requirements


     In all cases, the protocol MUST specify that authorization  data  =
sent
     from  the  NAS  to  the AAA server is to be regarded as =
information or
     "hints", and not directives.   The AAA protocol MUST  be  designed =
 so
     that  the  AAA server makes all final authorization decisions and =
does
     not depend on a certain state being expected by the NAS.


     8.3.1.1.  Dynamic Authorization


     The AAA protocol MUST support dynamic  re-authorization  at  any  =
time
     during  a  user  session.   This  re-authorization may be =
initiated in
     either  direction.   This  dynamic  re-authorization  capability  =
MUST
     include  the  capability  to  request  a  NAS  to disconnect a =
user on
     demand.





     Beadles                 Category: Informational               =
[Page 8]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     8.3.1.2.  Resource Management


     Resource management MUST be supported on demand  by  the  NAS  or  =
AAA
     Server at any time during the course of a user session.


     8.3.2.  Authorization Attribute Requirements



     8.3.2.1.  Authorization Attribute Requirements - Access =
Restrictions


     The  AAA protocol serves as a primary means of gathering data used =
for
     making Policy decisions for network access.  Therefore, the AAA =
proto-
     col MUST allow network operators to make policy decisions based on =
the
     following parameters:

          -Time/day restrictions.  The AAA protocol MUST be able to =
provide
          an  unambiguous  time  stamp,  NAS time zone indication, and =
date
          indication to the AAA server in the Authorization =
information.

          -Location restrictions:  The AAA protocol MUST be able to =
provide
          an  unambiguous  location code that reflects the geographic =
loca-
          tion of the NAS.  Note that this is not the same type of =
thing as
          either the dialing or dialed station.

          -Dialing  restrictions:  The AAA protocol MUST be able to =
provide
          accurate dialed and dialing station indications.

          -Concurrent login limitations:  The AAA protocol  MUST  allow =
 an
          AAA  Server  to  limit  concurrent logins by a particular =
user or
          group of users.  This mechanism does not need  to  be  =
explicitly
          built  into  the  AAA protocol, but the AAA protocol must =
provide
          sufficient authorization information for an AAA  server  to  =
make
          that determination through an out-of-band mechanism.


     8.3.2.2.   Authorization  Attribute  Requirements - Authorization =
Pro-
     files


     The AAA protocol is used to enforce policy at the  NAS.   =
Essentially,
     on  granting  of access, a particular access profile is applied to =
the
     user's session.  The AAA protocol MUST at a minimum provide a =
means of
     applying profiles containing the following types of information:

          -IP  Address assignment: The AAA protocol MUST provide a =
means of
          assigning an IPv4 or IPv6 address to an incoming user.

          -Protocol Filter application:  The AAA protocol  MUST  =
provide  a
          means of applying IP protocol filters to user sessions.  Two =
dif-
          ferent methods MUST be supported.



     Beadles                 Category: Informational               =
[Page 9]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


          First, the AAA protocol MUST provide a means of selecting a  =
pro-
          tocol  filter  by reference to an identifier, with the =
details of
          the filter action being specified out of band.  The AAA  =
protocol
          SHOULD define this out-of-band reference mechanism.

          Second, the AAA protocol MUST provide a means of passing a =
proto-
          col filter by value.  This means explicit passing  of  =
pass/block
          information  by address range, TCP/UDP port number, and IP =
proto-
          col number at a minimum.

          -Compulsory Tunneling:  The AAA protocol MUST provide a means =
 of
          directing  a NAS to build a tunnel or tunnels to a specified =
end-
          point.  It MUST support creation of multiple simultaneous =
tunnels
          in  a  specified  order.   The protocol MUST allow, at a =
minimum,
          specification of the tunnel endpoints, tunneling  protocol  =
type,
          underlying  tunnel  media type, and tunnel authentication =
creden-
          tials (if required by the tunnel type).  The  AAA  protocol  =
MUST
          support  at  least the creation of tunnels using the L2TP =
[L2TP],
          ESP [ESP], and AH [AH]  protocols.   The  protocol  MUST  =
provide
          means of adding new tunnel types as they are standardized.

          -Routing:   The  AAA protocol MUST provide a means of =
assigning a
          particular static route to an incoming user session.

          -Expirations/timeouts:  The AAA protocol MUST provide a means =
 of
          communication  session expiration information to a NAS.  =
Types of
          expirations that MUST be supported are:  total session time, =
idle
          time, total bytes transmitted, and total bytes received.

          -Quality  of  Service:   The AAA protocol MUST provide a =
means of
          applying Quality of Service parameters to  individual  user  =
ses-
          sions.


     8.3.2.3.  Resource Management Requirements


     The  AAA  protocol is a means for network operators to perform =
manage-
     ment of network resources.  The AAA protocol MUST provide a  means =
 of
     collecting  resource state information, and controlling resource =
allo-
     cation for the following types of network resources.

          -Network bandwidth usage per session,  including  multilink  =
ses-
          sions.

          -Access port usage, including concurrent usage and usage =
pools.

          -Connect time.

          -IP Addresses and pools.

          -Compulsory tunnel limits.





     Beadles                 Category: Informational              [Page =
10]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     8.3.3.  Authorization Protocol Security Requirements



     8.3.3.1.  Security of Compulsory Tunnel Credentials


     When an AAA protocol passes credentials that will be used to =
authenti-
     cate compulsory tunnels, the AAA protocol  MUST  provide  a  means =
 of
     securing the credentials from end to end of the AAA conversation.  =
The
     AAA protocol MUST also provide protection against  replay  attacks =
 in
     this situation.


     8.4.  Accounting and Auditing Requirements



     8.4.1.  Accounting Protocol Requirements



     8.4.1.1.  Guaranteed Delivery


     The accounting and auditing functions of the AAA protocol are used =
for
     network planning, resource management,  policy  decisions,  and  =
other
     functions  that  require  accurate  knowledge of the state of the =
NAS.
     NAS operators need to be able to engineer their network usage =
measure-
     ment  systems  to  a predictable level of accuracy.  Therefore, an =
AAA
     protocol MUST provide a means of  guaranteed  delivery  of  =
accounting
     information between the NAS and the AAA Server(s).


     8.4.1.2.  Real Time Accounting


     NAS  operators  often require a real time view onto the status of =
ses-
     sions served by a NAS.  Therefore, the AAA protocol MUST support =
real-
     time  delivery  of  accounting and auditing information.  In this =
con-
     text, real time is defined as accounting information  delivery  =
begin-
     ning within one second of the triggering event.


     8.4.1.3.  Batch Accounting


     The AAA protocol SHOULD also support delivery of stored accounting =
and
     auditing information in batches (non-real time).








     Beadles                 Category: Informational              [Page =
11]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     8.4.1.4.  Accounting Time Stamps


     There may be delays associated with the delivery of accounting  =
infor-
     mation.   The NAS operator will desire to know the time an event =
actu-
     ally occurred, rather than simply the time when  notification  of  =
the
     event  was  received.  Therefore, the AAA protocol MUST carry an =
unam-
     biguous time stamp associated with each accounting event.   This  =
time
     stamp  MUST  be  unambiguous with regard to time zone.  Note that =
this
     assumes  that  the  NAS  has  access  to  a   correct   time   =
source.


     8.4.1.5.  Accounting Events


     At a minimum, the AAA protocol MUST  support  delivery  of  =
accounting
     information triggered by the following events:
          -Start of a user session

          -End of a user session

          -Expiration  of  a predetermined repeating time interval =
during a
          user session.  The AAA protocol MUST provide a means for the  =
AAA
          server  to  request  that a NAS use a certain interval =
accounting
          time.

          -Dynamic  re-authorization  during  a  user  session  (e.g.,  =
new
          resources being delivered to the user)

          -Dynamic re-authentication during a user session


     8.4.1.6.  On-demand Accounting


     NAS  operators  need  to  maintain an accurate view onto the =
status of
     sessions served by a NAS, even  through  failure  of  an  AAA  =
server.
     Therefore, the AAA protocol MUST support a means of requesting =
current
     session state and accounting from the NAS on demand.


     8.4.2.  Accounting attribute requirements


     At a minimum, the AAA protocol MUST support delivery of the  =
following
     types of accounting/auditing data:
          -All parameters used to authenticate a session.

          -Details  of  the  authorization  profile that was applied to =
the
          session.

          -The duration of the session.

          -The cumulative number of bytes  sent  by  the  user  during  =
the



     Beadles                 Category: Informational              [Page =
12]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


          session.

          -The  cumulative  number of bytes received by the user during =
the
          session.

          -The cumulative number of packets sent by  the  user  during  =
the
          session.

          -The cumulative number of packets received by the user during =
the
          session.

          -Details of the access protocol used  during  the  session  =
(port
          type, connect speeds, etc.)


     8.4.3.  Accounting Protocol Security Requirements



     8.4.3.1.  Integrity and Confidentiality


     Note  that  accounting  and  auditing data are operationally =
sensitive
     information.  The AAA protocol MUST provide a means to assure  =
end-to-
     end  integrity  of this data.  The AAA protocol SHOULD provide a =
means
     of assuring the end-to-end confidentiality of this data.


     8.4.3.2.  Non-repudiation


     Network operators use accounting data for network  planning,  =
resource
     management,  and other business-critical functions that require =
confi-
     dence in the correctness of this data. The AAA protocol SHOULD =
provide
     a  mechanism to ensure that the source of Accounting data cannot =
repu-
     diate this data after transmission.


     9.  Device Management Protocols


     This document does not specify any requirements for device  =
management
     protocols.


     10.  Acknowledgments

     Many  of  the  requirements  in  this document first took form in =
Glen
     Zorn's "Yet Another  Authentication  Protocol  (YAAP)"  document,  =
for
     which grateful acknowledgment is made.







     Beadles                 Category: Informational              [Page =
13]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     11.  Security considerations


     See above for security requirements for the NAS AAA protocol.

     Where  an  AAA  architecture  spans multiple domains of authority, =
AAA
     information may need to cross trust boundaries.  In this =
situation,  a
     NAS  might  operate as a shared device that services multiple =
adminis-
     trative domains.  Network operators are advised take this into =
consid-
     eration when deploying NAS's and AAA Servers.


     12.  IANA Considerations


     This document does not directly specify any IANA considerations.  =
How-
     ever, the following recommendations are made:

     Future development and extension of an AAA protocol will be made  =
much
     easier  if  new  attributes  and values can be requested or =
registered
     directly through IANA, rather than  through  an  IETF  =
Standardization
     process.

     The  AAA  protocol  might  use  enumerated values for some =
attributes,
     which enumerate already-defined IANA types (such as protocol  =
number).
     In  these cases, the AAA protocol SHOULD use the IANA assigned =
numbers
     as the enumerated values.



     13.  References


     [KEYWORDS] S. Bradner.   "Key  words  for  use  in  RFCs  to  =
Indicate
     Requirement Levels."  RFC 2119, Harvard University, March 1997.

     [NAS-MODEL]  D.  Mitton,  M. Beadles.  "Network Access Server =
Require-
     ments Next Generation (NASREQNG) NAS Model."  Work in progress.

     [PPPOE] L. Mamakos et al.  "A Method for Transmitting PPP Over  =
Ether-
     net (PPPoE)."  RFC 2516, UUNET Technologies, Inc., February 1999.

     [L2TP] W. M. Townsley, et al.  "Layer Two Tunneling Protocol  =
(L2TP)."
     RFC 2661, IBM, Cisco, Ascend, Microsoft, August 1999.

     [PPP] W.  Simpson.  "The Point-to-Point Protocol  (PPP)."   RFC  =
1661,
     Daydreamer, July 1994.

     [TELNET]  J.  Postel,  J.  Reynolds.  "Telnet Protocol =
Specification."
     STD 8, RFC 854, ISI, May 1983.

     [ROUTING-REQUIREMENTS] F.  Baker.   "Requirements  for  IP  =
Version  4
     Routers."  RFC 1812, Cisco Systems, June 1995.




     Beadles                 Category: Informational              [Page =
14]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     [IPV6]  S.  Deering,  R. Hinden.  "Internet Protocol, Version 6 =
(IPv6)
     Specification."  RFC 2460, Cisco, Nokia, December 1998.

     [RFC 2277] H. Alvestrand.  "IETF Policy on  Character  Sets  and  =
Lan-
     guages."  RFC 2277, UNINETT, January 1998.

     [CHAP]  W.  Simpson.  "PPP Challenge Handshake Authentication =
Protocol
     (CHAP)."  RFC 1994, Daydreamer, August 1996.

     [EAP] L. Blunk, J. Vollbrecht.  "PPP Extensible Authentication  =
Proto-
     col (EAP)."  RFC 2284, Merit Network, Inc., March 1998.

     [NAI]  B.   Aboba,  M. Beadles.  "The Network Access Identifier."  =
RFC
     2486, Microsoft, WorldCom Advanced Networks, January 1999.

     [X.509] ITU-T Recommendation X.509 (1997 E): Information =
Technology  -
     Open  Systems  Interconnection  - The Directory: Authentication =
Frame-
     work, June 1997.

     [KERBEROS] J. Kohl, C. Neuman.  "The Kerberos  Network  =
Authentication
     Service  (V5)."  RFC 1510, Digital Equipment Corporation, ISI, =
Septem-
     ber 1993.

     [ESP] S.  Kent,  R.  Atkinson.   "IP  Encapsulating  Security  =
Payload
     (ESP)."  RFC 2406, BBN Corp, @Home Network, November 1998.

     [AH]  S.  Kent,  R.  Atkinson.   "IP Authentication Header (AH)."  =
RFC
     2402, BBN Corp, @Home Network, November 1998.

     [ROAMING-REQUIREMENTS] B. Aboba, G. Zorn.   "Criteria  for  =
Evaluating
     Roaming Protocols."  RFC 2477, Microsoft, January 1999.

     [RADIUS]

     [RADIUS-ACCOUNTING]



     14.  Author's Address



     Mark Anthony Beadles
     UUNET, an MCI WorldCom Company
     5000 Britton Rd.
     Hilliard, OH 43026

     Phone: 614-723-1941
     EMail: mbeadles@wcom.net








     Beadles                 Category: Informational              [Page =
15]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October =
1999


     15.  Full Copyright Statement


     Copyright (C) The Internet Society (1999).  All Rights Reserved.

     This  document  and  translations of it may be copied and =
furnished to
     others, and derivative works that comment on or otherwise  explain =
 it
     or  assist in its implmentation may be prepared, copied, published =
and
     distributed, in whole or in part, without  restriction  of  any  =
kind,
     provided  that  the  above  copyright  notice  and  this paragraph =
are
     included on all such copies and derivative works.  However, this =
docu-
     ment  itself  may  not be modified in any way, such as by removing =
the
     copyright notice or references to the Internet Society or other =
Inter-
     net  organizations,  except  as  needed  for the purpose of =
developing
     Internet standards in which case the procedures for copyrights =
defined
     in  the Internet Standards process must be followed, or as =
required to
     translate it into languages other than   English.  The limited =
permis-
     sions  granted  above  are  perpetual  and  will not be revoked by =
the
     Internet Society or its successors or assigns.  This document and  =
the
     information  contained  herein is provided on an "AS IS" basis and =
THE
     INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS =
ALL
     WARRANTIES,  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY =
WAR-
     RANTY THAT THE USE OF THE INFORMATION HEREIN  WILL  NOT  INFRINGE  =
ANY
     RIGHTS  OR  ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS =
FOR A
     PARTICULAR PURPOSE."


     16.  Expiration Date


     This document is  filed  as  <draft-ietf-nasreq-criteria-02.txt>,  =
and
     expires April 06, 2000.

























     Beadles                 Category: Informational              [Page =
16]



------_=_NextPart_000_01BF100C.7B99BA6C--


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct  6 13:07:17 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA14080;
	Wed, 6 Oct 1999 13:07:15 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id MAA29088
	for nasreq-outgoing; Wed, 6 Oct 1999 12:39:44 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id MAA29084
	for <nasreq@tdmx.rutgers.edu>; Wed, 6 Oct 1999 12:39:43 -0400 (EDT)
Received: from engmail1.Eng.Sun.COM ([129.146.1.13])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id JAA06167;
	Wed, 6 Oct 1999 09:39:31 -0700 (PDT)
Received: from hsmpka.eng.sun.com (phys-hsmpka.Eng.Sun.COM [129.146.105.37])
	by engmail1.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id JAA00024;
	Wed, 6 Oct 1999 09:39:27 -0700 (PDT)
Received: from hsmpka.Eng.Sun.COM by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id JAA22900; Wed, 6 Oct 1999 09:39:12 -0700
From: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Message-Id: <199910061639.JAA22900@hsmpka.eng.sun.com>
Date: Wed, 6 Oct 1999 09:39:17 -0700
To: "Beadles, Mark A." <MBeadles@wcom.net>,
        "Nasreqng List (E-mail)" <nasreq@tdmx.rutgers.edu>
Reply-To: <Pat.Calhoun@eng.sun.com>
Subject: Re: nasreq New Criteria Draft
X-Mailer: Sun NetMail 2.3
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

I'll be having a look today, but I would like to understand the intent of
the document. Is this supposed to be requirements for the AAA WG? If so,
is there any way that the AAA-specific requirements be split up into a separate
document?

PatC
>I have posted draft-ietf-nasreq-criteria-02.txt, attached.  This is the
>latest version of the NAS Requirements document.  There are some significant
>changes from the last version, so please read it carefully.
>
>--------------------
>
>Change log:
>
>-Incorporated items from Glen Zorn's YAAP document.
>-Incorporated items from Dave Mitton's RADIUS Practices document.
>-Incorporated various comments from the WG.
>-Organizational changes: 
> * more granularity in section numbering
> * separation of protocol requirements, attribute requirements, and security
>requirements
>
>--------------------
>
>Here is a condensed outline of the AAA requirements from this document:
>
>     8.  AAA Protocol Requirements
>     8.1.  General protocol characteristics
>     8.1.1.  Transport Requirements
>     8.1.2.  Attribute-Value Protocol Model
>     8.1.3.  Security Requirements
>     8.2.  Authentication and User Security Requirements
>     8.2.1.  Authentication protocol requirements
>     8.2.2.  Authentication Attribute Requirements
>     8.2.3.  Authentication Protocol Security Requirements
>     8.3.  Authorization, Policy, and Resource management
>     8.3.1.  Authorization Protocol Requirements
>     8.3.2.  Authorization Attribute Requirements
>     8.3.3.  Authorization Protocol Security Requirements
>     8.4.  Accounting and Auditing Requirements
>     8.4.1.  Accounting Protocol Requirements
>     8.4.2.  Accounting attribute requirements
>     8.4.3.  Accounting Protocol Security Requirements
>
>And for the masochists among you, here is a complete outline of the
>document:
>
>     1.  Status of this Memo
>     2.  Copyright Statement
>     3.  Abstract
>     4.  Requirements language
>     5.  Introduction
>     6.  Access Protocol Requirements
>     7.  Network Protocol Requirements
>     8.  AAA Protocol Requirements
>     8.1.  General protocol characteristics
>     8.1.1.  Transport requirements
>     8.1.1.1.  Transport independence
>     8.1.1.2.  Long-term exchange of small packets
>     8.1.1.3.  Support for multiple AAA servers
>     8.1.1.4.  Support for Multiple Administrative Domains
>     8.1.2.  Attribute-Value Protocol Model
>     8.1.2.1.  Attribute Data Types
>     8.1.2.2.  Minimum Set of Attributes
>     8.1.2.3.  Attribute Extensibility
>     8.1.3.  Security Requirements
>     8.1.3.1.  Mutual Authentication
>     8.1.3.2.  Shared Secrets
>     8.1.3.3.  Public Key Security
>     8.1.3.4.  Encryption of Attributes
>     8.2.  Authentication and User Security Requirements
>     8.2.1.  Authentication protocol requirements
>     8.2.1.1.  Bi-directional Authentication
>     8.2.1.2.  Dynamic Authentication
>     8.2.1.3.  Multi-phase Authentication
>     8.2.1.4.  Extensible Authentication Types
>     8.2.2.  Authentication Attribute Requirements
>     8.2.2.1.  PPP Authentication protocols
>     8.2.2.2.  User Identification
>     8.2.2.3.  Authentication Credentials
>     8.2.3.  Authentication Protocol Security Requirements
>     8.2.3.1.  End-to-End Hiding of Credentials
>     8.3.  Authorization, Policy, and Resource management
>     8.3.1.  Authorization Protocol Requirements
>     8.3.1.1.  Dynamic Authorization
>     8.3.1.2.  Resource Management
>     8.3.2.  Authorization Attribute Requirements
>     8.3.2.1.  Authorization Attribute Requirements - Access Restrictions
>     8.3.2.2.  Authorization Attribute Requirements - Authorization Profiles
>     8.3.2.3.  Resource Management Requirements
>     8.3.3.  Authorization Protocol Security Requirements
>     8.3.3.1.  Security of Compulsory Tunnel Credentials
>     8.4.  Accounting and Auditing Requirements
>     8.4.1.  Accounting Protocol Requirements
>     8.4.1.1.  Guaranteed Delivery
>     8.4.1.2.  Real Time Accounting
>     8.4.1.3.  Batch Accounting
>     8.4.1.4.  Accounting Time Stamps
>     8.4.1.5.  Accounting Events
>     8.4.1.6.  On-demand Accounting
>     8.4.2.  Accounting attribute requirements
>     8.4.3.  Accounting Protocol Security Requirements
>     8.4.3.1.  Integrity and Confidentiality
>     8.4.3.2.  Non-repudiation
>     9.  Device Management Protocols
>     10.  Acknowledgments
>     11.  Security considerations
>     12.  IANA Considerations
>     13.  References
>     14.  Author's Address
>     15.  Full Copyright Statement
>     16.  Expiration Date
>
> <<draft-ietf-nasreq-criteria-02.txt>> 
>
>+ Mark Anthony Beadles + Co-chair, IETF NASREQ WG   + 
>+  mbeadles@wcom.net   + http://nasreq.rutgers.edu/ +
>+  Voice 614.723.1941  +       Fax 614.723.8407     +
>
>
>




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct  6 13:07:54 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA14115;
	Wed, 6 Oct 1999 13:07:54 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id MAA29127
	for nasreq-outgoing; Wed, 6 Oct 1999 12:47:41 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id MAA29123
	for <nasreq@tdmx.rutgers.edu>; Wed, 6 Oct 1999 12:47:39 -0400 (EDT)
Received: from engmail4.Eng.Sun.COM ([129.144.134.6])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id JAA10750;
	Wed, 6 Oct 1999 09:47:36 -0700 (PDT)
Received: from hsmpka.eng.sun.com (hsmpka.Eng.Sun.COM [129.146.51.47])
	by engmail4.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id JAA09322;
	Wed, 6 Oct 1999 09:47:32 -0700 (PDT)
Received: from hsmpka.Eng.Sun.COM by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id JAA25492; Wed, 6 Oct 1999 09:47:20 -0700
From: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Message-Id: <199910061647.JAA25492@hsmpka.eng.sun.com>
Date: Wed, 6 Oct 1999 09:47:20 -0700
To: "Beadles, Mark A." <MBeadles@wcom.net>,
        "Nasreqng List (E-mail)" <nasreq@tdmx.rutgers.edu>
Reply-To: <Pat.Calhoun@eng.sun.com>
Subject: Re: nasreq New Criteria Draft
X-Mailer: Sun NetMail 2.3
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Ignore the previous comment. The document really only concentrates on AAA.

Sorry for the confusion. Should have looked at the doc before sending the
mail.

PatC
>I have posted draft-ietf-nasreq-criteria-02.txt, attached.  This is the
>latest version of the NAS Requirements document.  There are some significant
>changes from the last version, so please read it carefully.
>
>--------------------
>
>Change log:
>
>-Incorporated items from Glen Zorn's YAAP document.
>-Incorporated items from Dave Mitton's RADIUS Practices document.
>-Incorporated various comments from the WG.
>-Organizational changes: 
> * more granularity in section numbering
> * separation of protocol requirements, attribute requirements, and security
>requirements
>
>--------------------
>
>Here is a condensed outline of the AAA requirements from this document:
>
>     8.  AAA Protocol Requirements
>     8.1.  General protocol characteristics
>     8.1.1.  Transport Requirements
>     8.1.2.  Attribute-Value Protocol Model
>     8.1.3.  Security Requirements
>     8.2.  Authentication and User Security Requirements
>     8.2.1.  Authentication protocol requirements
>     8.2.2.  Authentication Attribute Requirements
>     8.2.3.  Authentication Protocol Security Requirements
>     8.3.  Authorization, Policy, and Resource management
>     8.3.1.  Authorization Protocol Requirements
>     8.3.2.  Authorization Attribute Requirements
>     8.3.3.  Authorization Protocol Security Requirements
>     8.4.  Accounting and Auditing Requirements
>     8.4.1.  Accounting Protocol Requirements
>     8.4.2.  Accounting attribute requirements
>     8.4.3.  Accounting Protocol Security Requirements
>
>And for the masochists among you, here is a complete outline of the
>document:
>
>     1.  Status of this Memo
>     2.  Copyright Statement
>     3.  Abstract
>     4.  Requirements language
>     5.  Introduction
>     6.  Access Protocol Requirements
>     7.  Network Protocol Requirements
>     8.  AAA Protocol Requirements
>     8.1.  General protocol characteristics
>     8.1.1.  Transport requirements
>     8.1.1.1.  Transport independence
>     8.1.1.2.  Long-term exchange of small packets
>     8.1.1.3.  Support for multiple AAA servers
>     8.1.1.4.  Support for Multiple Administrative Domains
>     8.1.2.  Attribute-Value Protocol Model
>     8.1.2.1.  Attribute Data Types
>     8.1.2.2.  Minimum Set of Attributes
>     8.1.2.3.  Attribute Extensibility
>     8.1.3.  Security Requirements
>     8.1.3.1.  Mutual Authentication
>     8.1.3.2.  Shared Secrets
>     8.1.3.3.  Public Key Security
>     8.1.3.4.  Encryption of Attributes
>     8.2.  Authentication and User Security Requirements
>     8.2.1.  Authentication protocol requirements
>     8.2.1.1.  Bi-directional Authentication
>     8.2.1.2.  Dynamic Authentication
>     8.2.1.3.  Multi-phase Authentication
>     8.2.1.4.  Extensible Authentication Types
>     8.2.2.  Authentication Attribute Requirements
>     8.2.2.1.  PPP Authentication protocols
>     8.2.2.2.  User Identification
>     8.2.2.3.  Authentication Credentials
>     8.2.3.  Authentication Protocol Security Requirements
>     8.2.3.1.  End-to-End Hiding of Credentials
>     8.3.  Authorization, Policy, and Resource management
>     8.3.1.  Authorization Protocol Requirements
>     8.3.1.1.  Dynamic Authorization
>     8.3.1.2.  Resource Management
>     8.3.2.  Authorization Attribute Requirements
>     8.3.2.1.  Authorization Attribute Requirements - Access Restrictions
>     8.3.2.2.  Authorization Attribute Requirements - Authorization Profiles
>     8.3.2.3.  Resource Management Requirements
>     8.3.3.  Authorization Protocol Security Requirements
>     8.3.3.1.  Security of Compulsory Tunnel Credentials
>     8.4.  Accounting and Auditing Requirements
>     8.4.1.  Accounting Protocol Requirements
>     8.4.1.1.  Guaranteed Delivery
>     8.4.1.2.  Real Time Accounting
>     8.4.1.3.  Batch Accounting
>     8.4.1.4.  Accounting Time Stamps
>     8.4.1.5.  Accounting Events
>     8.4.1.6.  On-demand Accounting
>     8.4.2.  Accounting attribute requirements
>     8.4.3.  Accounting Protocol Security Requirements
>     8.4.3.1.  Integrity and Confidentiality
>     8.4.3.2.  Non-repudiation
>     9.  Device Management Protocols
>     10.  Acknowledgments
>     11.  Security considerations
>     12.  IANA Considerations
>     13.  References
>     14.  Author's Address
>     15.  Full Copyright Statement
>     16.  Expiration Date
>
> <<draft-ietf-nasreq-criteria-02.txt>> 
>
>+ Mark Anthony Beadles + Co-chair, IETF NASREQ WG   + 
>+  mbeadles@wcom.net   + http://nasreq.rutgers.edu/ +
>+  Voice 614.723.1941  +       Fax 614.723.8407     +
>
>
>




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct  6 21:02:58 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA23144;
	Wed, 6 Oct 1999 21:02:57 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id UAA01332
	for nasreq-outgoing; Wed, 6 Oct 1999 20:28:23 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id UAA01328
	for <nasreq@tdmx.rutgers.edu>; Wed, 6 Oct 1999 20:28:20 -0400 (EDT)
From: Pat.Calhoun@eng.sun.com
Received: from engmail4.Eng.Sun.COM ([129.144.134.6])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id RAA18918
	for <nasreq@tdmx.rutgers.edu>; Wed, 6 Oct 1999 17:28:20 -0700 (PDT)
Received: from hsmpka.eng.sun.com (hsmpka.Eng.Sun.COM [129.146.51.47])
	by engmail4.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id RAA11590
	for <nasreq@tdmx.rutgers.edu>; Wed, 6 Oct 1999 17:28:19 -0700 (PDT)
Received: from westapp by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id RAA04691; Wed, 6 Oct 1999 17:28:15 -0700
Message-ID: <525299046.939256094705.JavaMail.pcalhoun@hsmpka.Eng.Sun.COM>
Date: Wed, 6 Oct 1999 17:28:14 -0700 (PDT)
To: nasreq@tdmx.rutgers.edu
Subject: Comments on draft-ietf-nasreq-criteria-02.txt
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
X-Mailer: Sun(TM) Web Access 1.0
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit


Mark,

Here are my comments on the draft. Sorry they are so extensive that it
looks like most of the I-D is included in this e-mail. I did it this way
because in most cases some context is needed to understand the comment.

Good work, btw.

PatC
----

     5.  Introduction

     This  document  defines  requirements  for  protocols  used by Network
     Access Servers (NAS).Protocols used by NAS's may be divided into  four
     spaces:   Access  protocols,  Network  protocols,  AAA  protocols, and
     Device Management protocols.   The primary focus of this  document  is
     on AAA protocols.  The reference model of a NAS used by this document,
     and the analysis of the functions of a NAS which led to  the  develop-
     ment of these requirements, may be found in [NAS-MODEL].
PRC: Nit. Space is missing before Protocols in second sentence.

     6.  Access Protocol Requirements


     There  are three basic types of access protocols used by NAS's.  First
     are the traditional telephony-based access protocols, which  interface
     to  the  NAS  via a modem or terminal adapter or similar device. These
     protocols typically support asynchronous or synchronous PPP [PPP] car-
     ried over a telephony protocol.
PRC: What is a telephony protocol? Q.931?
     Second are broadband pseudo-telephony
     access protocols, which are carried over xDSL  or  cable  modems,  for
     example.   These  protocols  typically support an encapsulation method
     such as PPP over Ethernet [PPPOE].  Finally  are  the  virtual  access
     protocols  used  by NAS's that terminate tunnels.  One example of this
     type of protocol is L2TP [L2TP].
PRC: What about cellular devices, such as in the 3G model. Initially the
devices will connect via PPP, but there is talk to also support HDLC in
order to get rid of PPP (since we that is needed is the framing).

     It is a central assumption of the NAS  model  used  here  that  a  NAS
     accepts multiple point-to-point links via one of the above access pro-
     tocols.  Therefore, at a minimum, any NAS access protocol MUST be able
     to  carry  PPP.   The  exception to this requirement is for NAS's that
     support legacy text login methods such as telnet [TELNET], rlogin,  or
     LAT.   Only  these access protocols are exempt from the requirement to
     support PPP.
PRC: So have we abandoned SLIP? I am not stating that we need to support it.
I am just curious.

     8.1.1.1.  Transport independence

     The AAA protocol MUST be transport independent, and MUST be capable of
     using both TCP and UDP as a transport.   Additionally, the AAA  proto-
     col  SHOULD  be designed to easily support any new transport protocols
     developed by the Internet standards community.
PRC:Having an AAA protocol support both TCP and UDP has brought *lots* of criticism from the likes of Fred Baker. In fact, due to such pressures one protocol that comes to mind now only supports UDP. This specific protocol
had support for both because the applications people did not need fast
fail-over, but I think that nasreq is a good candidate for fast fail-over.

    8.1.1.2.  Long-term exchange of small packets


     Very large scale NAS's that serve up to thousands of simultaneous ses-
     sions  are now being deployed.  This means that, in the extreme, there
     may be an almost constant exchange of many small packets  between  the
     NAS  and  the  AAA  server.   An AAA protocol transport SHOULD support
     being optimized for a long-term exchange of small packets in a  stream
     between a pair of hosts.
PRC: So are we talking about multiple sub-requests within a single request?
If so, this adds some interesting problems in roaming environments where part
of a request goes to A, and the other part to B.

     8.1.1.3.  Support for multiple AAA servers


     In  order to operationally support large loads, load balancing to mul-
     tiple AAA servers will be required.  The AAA protocol MUST provide for
     NAS's  to  balance  AAA sessions between two or more AAA servers.  The
     load balancing mechanism SHOULD  be  built  in  to  the  AAA  protocol
     itself.   The AAA protocol design MUST NOT introduce a single point of
     failure during the AAA process.  The AAA protocol MUST allow any  ses-
     sions between a NAS and a given AAA server to fail over to a secondary
     server without loss of state information.   This  fail-over  mechanism
     SHOULD be built in to the AAA protocol itself.
PRC: I think that is *also* needed here is a method to flow the traffic between
the NAS and the server. Otherwise, flodding a server that is already
busy is pointless. This would allow for a more effective load balancing
mechanism.

     8.1.1.4.  Support for Multiple Administrative Domains
[...]
     port proxy, and in additiona SHOULD support other  multi-domain  rela-
     tionships such as brokering and referral.
PRC: Should read "addition"

     8.1.2.  Attribute-Value Protocol Model

[...]
     Experience has also shown  that  attribute  space  tends  to  run  out
     quickly.  In  order  to  provide  room  for expansion in the attribute
     space, the AAA protocol MUST support a minimum of 64K  Attributes  (16
     bits), each with a minimum length of 64K (16 bits).
PRC: Of course, my preference here is 32 bit since if the protocol is properly
aligned on a 32-bit boundary, it will be more CPU friendly. This will allow
for a higher number of transactions/sec.

     8.1.2.1.  Attribute Data Types


     The  AAA  protocol MUST support simple attribute data types, including
     integer, enumeration, text string, IP address, and date/time. The  AAA
     protocol  MUST  also  provide some support for complex structured data
     types. Wherever IP addresses are carried within the AAA protocol,  the
     protocol  MUST  support  both IPv4 and IPv6 [IPV6] addresses. Wherever
     text information is carried within the AAA protocol, the protocol MUST
     comply  with  the  IETF  Policy  on  Character Sets and Languages [RFC
     2277].
PRC: The requirement for IP V6 is interesting. Note that one data type is just
a series of octets, not necessarily a string (e.g. challenge).

     8.1.2.3.  Attribute Extensibility

[...]
     The AAA protocol MUST include a means for individual  vendors  to  add
     value  through  vendor-specific  attributes and SHOULD include support
     for vendor-specific data types.
PRC: No mention for vendor specific commands? Is this also a requirement?

     8.1.3.1.  Mutual Authentication


     It is poor security practice for a NAS  to  communicate  with  an  AAA
     server  that  is  not  trusted, and vice versa.  The AAA protocol MUST
     provide mutual authentication between AAA server and NAS.
PRC: Although it should be possible for the protocol to make use of an underlying security protocol, such as AH/EAP. However, in practice this is more difficult since the application cannot guarantee that the policy is correctly setup and could in fact be sending sensitive information unprotected in the clear. However, if a network operator has deployed IP Security, the protocol should allow it.

     8.1.3.3.  Public Key Security


     AAA server/NAS identity verification based solely  on  shared  secrets
     can  be  difficult  to  deploy  properly at large scale, and it can be
     tempting for NAS operators to use a single shared secret (that  rarely
     changes)  across  all  NAS's.  This can lead to easy compromise of the
     secret.  Therefore, the AAA protocol MUST also support mutual  verifi-
     cation  of  identity  using  a public-key infrastructure that supports
     expiration and revocation of keys.
PRC: I think that for hop-by-hop security, IP Security would provide this
and would allow for re-keying. Should the AAA protocol also provide this?
I believe that it may in inter-domain cases since it MAY be a requirement
to include non-repudiation within an accounting record, and this wouldn't
be possible with IP Security.

     8.2.1.4.  Extensible Authentication Types


     Security protocol development is going on constantly  as  new  threats
     are  identified  and  better  cracking methods are developed.  Today's
     secure authentication methods may be proven  insecure  tomorrow.   The
     AAA protocol MUST provide some support for addition of new authentica-
     tion credential types.
PRC: Should we simply call for EAP here? Or are we talking about AAA-transport
security?

    8.2.2.2.  User Identification


[...]
     If a particular type of identity credential is not needed for  a  par-
     ticular  user  session,  the  AAA protocol MUST NOT require that dummy
     credentials be filled in.  That is,  the  AAA  protocol  MUST  support
     identification without authentication.
PRC: Or perhaps the other way around?

     8.2.2.3.  Authentication Credentials

[...]
          -A secret or password.

          -A response to a challenge presented by the NAS to the user

          -A one-time password

          -An X.509 digital certificate [X.509]

          -A Kerberos v5 ticket [KERBEROS]
PRC: I am not sure that I get it. I think that each authentication scheme
has it's own type of credential, so what is this section intended to state?

     8.2.3.1.  End-to-End Hiding of Credentials


     Where passwords are used as authentication credentials, the AAA proto-
     col MUST provide a secure means of hiding the password from end to end
     of  the  AAA  conversation.   Where  challenge/response mechanisms are
     used, the AAA protocol MUST also prevent against replay attacks.

PRC: I would prefer to see end-to-end authentication. Glen and I have a draft that
talks about this, and this effectively eliminates the authentication replay problem
that occurs with CHAP/RADIUS today.

     8.3.1.2.  Resource Management


     Resource management MUST be supported on demand  by  the  NAS  or  AAA
     Server at any time during the course of a user session.
PRC: Inter-Domain resource management can be *very* difficult. However the
I-D that Glen and I put together makes it scale across the Internet by
requiring periodical re-authentication.

     8.3.2.1.  Authorization Attribute Requirements - Access Restrictions


[...]
          -Time/day restrictions.  The AAA protocol MUST be able to provide
          an  unambiguous  time  stamp,  NAS time zone indication, and date
          indication to the AAA server in the Authorization information.
PRC: This is good, and the work in Secure NTP will certainly help. This is
especially true in inter-domain cases where assuming that all hosts are time
synched is difficult.

          -Location restrictions:  The AAA protocol MUST be able to provide
          an  unambiguous  location code that reflects the geographic loca-
          tion of the NAS.  Note that this is not the same type of thing as
          either the dialing or dialed station.
PRC: I hear privacy issues here, but perhaps it *could* be a requirement. I
am not sure that I'd want anyone to know that I am in a cyber-coffee shop
somewhere in the Netherlands.

[...]
          -Concurrent login limitations:  The AAA protocol  MUST  allow  an
          AAA  Server  to  limit  concurrent logins by a particular user or
          group of users.  This mechanism does not need  to  be  explicitly
          built  into  the  AAA protocol, but the AAA protocol must provide
          sufficient authorization information for an AAA  server  to  make
          that determination through an out-of-band mechanism.
PRC: Working on the draft with Glen and previous exercises with Merit to come
up with such a scheme for inter-domain has lead me to believe that this is
a very difficult problem. However, one can get similar functionality if you
use the scheme described in the draft that Glen and I put together.

     8.3.2.2.   Authorization  Attribute  Requirements - Authorization Pro-
     files
[...]
          -Protocol Filter application:  The AAA protocol  MUST  provide  a
          means of applying IP protocol filters to user sessions.  Two dif-
          ferent methods MUST be supported.

          First, the AAA protocol MUST provide a means of selecting a  pro-
          tocol  filter  by reference to an identifier, with the details of
          the filter action being specified out of band.  The AAA  protocol
          SHOULD define this out-of-band reference mechanism.
PRC: This smells of scaling problems.

     8.3.3.1.  Security of Compulsory Tunnel Credentials


     When an AAA protocol passes credentials that will be used to authenti-
     cate compulsory tunnels, the AAA protocol  MUST  provide  a  means  of
     securing the credentials from end to end of the AAA conversation.  The
     AAA protocol MUST also provide protection against  replay  attacks  in
     this situation.
PRC: Is this the authentication blob, or the username?

     8.4.1.1.  Guaranteed Delivery


     The accounting and auditing functions of the AAA protocol are used for
     network planning, resource management,  policy  decisions,  and  other
     functions  that  require  accurate  knowledge of the state of the NAS.
     NAS operators need to be able to engineer their network usage measure-
     ment  systems  to  a predictable level of accuracy.  Therefore, an AAA
     protocol MUST provide a means of  guaranteed  delivery  of  accounting
     information between the NAS and the AAA Server(s).
PRC: My experience is that hop-by-hop delivery acknowledgement is the best way
since it reduces the traffic, and the end-to-end retransmissions known to
exist in RADIUS.

     8.4.1.4.  Accounting Time Stamps


     There may be delays associated with the delivery of accounting  infor-
     mation.   The NAS operator will desire to know the time an event actu-
     ally occurred, rather than simply the time when  notification  of  the
     event  was  received.  Therefore, the AAA protocol MUST carry an unam-
     biguous time stamp associated with each accounting event.   This  time
     stamp  MUST  be  unambiguous with regard to time zone.  Note that this
     assumes  that  the  NAS  has  access  to  a   correct   time   source.
PRC: Does it matter if an intermediate node changes the time? Or does the time
have the be protected end-to-end?

     8.4.3.2.  Non-repudiation

     Network operators use accounting data for network  planning,  resource
     management,  and other business-critical functions that require confi-
     dence in the correctness of this data. The AAA protocol SHOULD provide
     a  mechanism to ensure that the source of Accounting data cannot repu-
     diate this data after transmission.
PRC: non-repudiation of both the source and the destination means that sufficient
data must be present in the accounting response. This allows the sender to
prove that the peer did in fact receive, and acknowledge receipt of the
accounting data.



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct  6 21:04:44 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA23181;
	Wed, 6 Oct 1999 21:04:43 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id UAA01326
	for nasreq-outgoing; Wed, 6 Oct 1999 20:28:08 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id UAA01322
	for <nasreq@tdmx.rutgers.edu>; Wed, 6 Oct 1999 20:28:06 -0400 (EDT)
From: Pat.Calhoun@eng.sun.com
Received: from engmail2.Eng.Sun.COM ([129.146.1.25])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id RAA18851
	for <nasreq@tdmx.rutgers.edu>; Wed, 6 Oct 1999 17:28:05 -0700 (PDT)
Received: from hsmpka.eng.sun.com (hsmpka.Eng.Sun.COM [129.146.51.47])
	by engmail2.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id RAA09164
	for <nasreq@tdmx.rutgers.edu>; Wed, 6 Oct 1999 17:28:04 -0700 (PDT)
Received: from westapp by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id RAA04639; Wed, 6 Oct 1999 17:28:02 -0700
Message-ID: <525314688.939256082195.JavaMail.pcalhoun@hsmpka.Eng.Sun.COM>
Date: Wed, 6 Oct 1999 17:28:01 -0700 (PDT)
To: nasreq@tdmx.rutgers.edu
Subject: Comments on draft-ietf-nasreq-criteria-02.txt
Mime-Version: 1.0
Content-Type: multipart/mixed; boundary="525316457.939256081980.JavaMail.root@westapp"
X-Mailer: Sun(TM) Web Access 1.0
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

--525316457.939256081980.JavaMail.root@westapp
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit


Mark,

Here are my comments on the draft. Sorry they are so extensive that it
looks like most of the I-D is included in this e-mail. I did it this way
because in most cases some context is needed to understand the comment.

Good work, btw.

PatC
----

     5.  Introduction

     This  document  defines  requirements  for  protocols  used by Network
     Access Servers (NAS).Protocols used by NAS's may be divided into  four
     spaces:   Access  protocols,  Network  protocols,  AAA  protocols, and
     Device Management protocols.   The primary focus of this  document  is
     on AAA protocols.  The reference model of a NAS used by this document,
     and the analysis of the functions of a NAS which led to  the  develop-
     ment of these requirements, may be found in [NAS-MODEL].
PRC: Nit. Space is missing before Protocols in second sentence.

     6.  Access Protocol Requirements


     There  are three basic types of access protocols used by NAS's.  First
     are the traditional telephony-based access protocols, which  interface
     to  the  NAS  via a modem or terminal adapter or similar device. These
     protocols typically support asynchronous or synchronous PPP [PPP] car-
     ried over a telephony protocol.
PRC: What is a telephony protocol? Q.931?
     Second are broadband pseudo-telephony
     access protocols, which are carried over xDSL  or  cable  modems,  for
     example.   These  protocols  typically support an encapsulation method
     such as PPP over Ethernet [PPPOE].  Finally  are  the  virtual  access
     protocols  used  by NAS's that terminate tunnels.  One example of this
     type of protocol is L2TP [L2TP].
PRC: What about cellular devices, such as in the 3G model. Initially the
devices will connect via PPP, but there is talk to also support HDLC in
order to get rid of PPP (since we that is needed is the framing).

     It is a central assumption of the NAS  model  used  here  that  a  NAS
     accepts multiple point-to-point links via one of the above access pro-
     tocols.  Therefore, at a minimum, any NAS access protocol MUST be able
     to  carry  PPP.   The  exception to this requirement is for NAS's that
     support legacy text login methods such as telnet [TELNET], rlogin,  or
     LAT.   Only  these access protocols are exempt from the requirement to
     support PPP.
PRC: So have we abandoned SLIP? I am not stating that we need to support it.
I am just curious.

     8.1.1.1.  Transport independence

     The AAA protocol MUST be transport independent, and MUST be capable of
     using both TCP and UDP as a transport.   Additionally, the AAA  proto-
     col  SHOULD  be designed to easily support any new transport protocols
     developed by the Internet standards community.
PRC:Having an AAA protocol support both TCP and UDP has brought *lots* of criticism from the likes of Fred Baker. In fact, due to such pressures one protocol that comes to mind now only supports UDP. This specific protocol
had support for both because the applications people did not need fast
fail-over, but I think that nasreq is a good candidate for fast fail-over.

    8.1.1.2.  Long-term exchange of small packets


     Very large scale NAS's that serve up to thousands of simultaneous ses-
     sions  are now being deployed.  This means that, in the extreme, there
     may be an almost constant exchange of many small packets  between  the
     NAS  and  the  AAA  server.   An AAA protocol transport SHOULD support
     being optimized for a long-term exchange of small packets in a  stream
     between a pair of hosts.
PRC: So are we talking about multiple sub-requests within a single request?
If so, this adds some interesting problems in roaming environments where part
of a request goes to A, and the other part to B.

     8.1.1.3.  Support for multiple AAA servers


     In  order to operationally support large loads, load balancing to mul-
     tiple AAA servers will be required.  The AAA protocol MUST provide for
     NAS's  to  balance  AAA sessions between two or more AAA servers.  The
     load balancing mechanism SHOULD  be  built  in  to  the  AAA  protocol
     itself.   The AAA protocol design MUST NOT introduce a single point of
     failure during the AAA process.  The AAA protocol MUST allow any  ses-
     sions between a NAS and a given AAA server to fail over to a secondary
     server without loss of state information.   This  fail-over  mechanism
     SHOULD be built in to the AAA protocol itself.
PRC: I think that is *also* needed here is a method to flow the traffic between
the NAS and the server. Otherwise, flodding a server that is already
busy is pointless. This would allow for a more effective load balancing
mechanism.

     8.1.1.4.  Support for Multiple Administrative Domains
[...]
     port proxy, and in additiona SHOULD support other  multi-domain  rela-
     tionships such as brokering and referral.
PRC: Should read "addition"

     8.1.2.  Attribute-Value Protocol Model

[...]
     Experience has also shown  that  attribute  space  tends  to  run  out
     quickly.  In  order  to  provide  room  for expansion in the attribute
     space, the AAA protocol MUST support a minimum of 64K  Attributes  (16
     bits), each with a minimum length of 64K (16 bits).
PRC: Of course, my preference here is 32 bit since if the protocol is properly
aligned on a 32-bit boundary, it will be more CPU friendly. This will allow
for a higher number of transactions/sec.

     8.1.2.1.  Attribute Data Types


     The  AAA  protocol MUST support simple attribute data types, including
     integer, enumeration, text string, IP address, and date/time. The  AAA
     protocol  MUST  also  provide some support for complex structured data
     types. Wherever IP addresses are carried within the AAA protocol,  the
     protocol  MUST  support  both IPv4 and IPv6 [IPV6] addresses. Wherever
     text information is carried within the AAA protocol, the protocol MUST
     comply  with  the  IETF  Policy  on  Character Sets and Languages [RFC
     2277].
PRC: The requirement for IP V6 is interesting. Note that one data type is just
a series of octets, not necessarily a string (e.g. challenge).

     8.1.2.3.  Attribute Extensibility

[...]
     The AAA protocol MUST include a means for individual  vendors  to  add
     value  through  vendor-specific  attributes and SHOULD include support
     for vendor-specific data types.
PRC: No mention for vendor specific commands? Is this also a requirement?

     8.1.3.1.  Mutual Authentication


     It is poor security practice for a NAS  to  communicate  with  an  AAA
     server  that  is  not  trusted, and vice versa.  The AAA protocol MUST
     provide mutual authentication between AAA server and NAS.
PRC: Although it should be possible for the protocol to make use of an underlying security protocol, such as AH/EAP. However, in practice this is more difficult since the application cannot guarantee that the policy is correctly setup and could in fact be sending sensitive information unprotected in the clear. However, if a network operator has deployed IP Security, the protocol should allow it.

     8.1.3.3.  Public Key Security


     AAA server/NAS identity verification based solely  on  shared  secrets
     can  be  difficult  to  deploy  properly at large scale, and it can be
     tempting for NAS operators to use a single shared secret (that  rarely
     changes)  across  all  NAS's.  This can lead to easy compromise of the
     secret.  Therefore, the AAA protocol MUST also support mutual  verifi-
     cation  of  identity  using  a public-key infrastructure that supports
     expiration and revocation of keys.
PRC: I think that for hop-by-hop security, IP Security would provide this
and would allow for re-keying. Should the AAA protocol also provide this?
I believe that it may in inter-domain cases since it MAY be a requirement
to include non-repudiation within an accounting record, and this wouldn't
be possible with IP Security.

     8.2.1.4.  Extensible Authentication Types


     Security protocol development is going on constantly  as  new  threats
     are  identified  and  better  cracking methods are developed.  Today's
     secure authentication methods may be proven  insecure  tomorrow.   The
     AAA protocol MUST provide some support for addition of new authentica-
     tion credential types.
PRC: Should we simply call for EAP here? Or are we talking about AAA-transport
security?

    8.2.2.2.  User Identification


[...]
     If a particular type of identity credential is not needed for  a  par-
     ticular  user  session,  the  AAA protocol MUST NOT require that dummy
     credentials be filled in.  That is,  the  AAA  protocol  MUST  support
     identification without authentication.
PRC: Or perhaps the other way around?

     8.2.2.3.  Authentication Credentials

[...]
          -A secret or password.

          -A response to a challenge presented by the NAS to the user

          -A one-time password

          -An X.509 digital certificate [X.509]

          -A Kerberos v5 ticket [KERBEROS]
PRC: I am not sure that I get it. I think that each authentication scheme
has it's own type of credential, so what is this section intended to state?

     8.2.3.1.  End-to-End Hiding of Credentials


     Where passwords are used as authentication credentials, the AAA proto-
     col MUST provide a secure means of hiding the password from end to end
     of  the  AAA  conversation.   Where  challenge/response mechanisms are
     used, the AAA protocol MUST also prevent against replay attacks.

PRC: I would prefer to see end-to-end authentication. Glen and I have a draft that
talks about this, and this effectively eliminates the authentication replay problem
that occurs with CHAP/RADIUS today.

     8.3.1.2.  Resource Management


     Resource management MUST be supported on demand  by  the  NAS  or  AAA
     Server at any time during the course of a user session.
PRC: Inter-Domain resource management can be *very* difficult. However the
I-D that Glen and I put together makes it scale across the Internet by
requiring periodical re-authentication.

     8.3.2.1.  Authorization Attribute Requirements - Access Restrictions


[...]
          -Time/day restrictions.  The AAA protocol MUST be able to provide
          an  unambiguous  time  stamp,  NAS time zone indication, and date
          indication to the AAA server in the Authorization information.
PRC: This is good, and the work in Secure NTP will certainly help. This is
especially true in inter-domain cases where assuming that all hosts are time
synched is difficult.

          -Location restrictions:  The AAA protocol MUST be able to provide
          an  unambiguous  location code that reflects the geographic loca-
          tion of the NAS.  Note that this is not the same type of thing as
          either the dialing or dialed station.
PRC: I hear privacy issues here, but perhaps it *could* be a requirement. I
am not sure that I'd want anyone to know that I am in a cyber-coffee shop
somewhere in the Netherlands.

[...]
          -Concurrent login limitations:  The AAA protocol  MUST  allow  an
          AAA  Server  to  limit  concurrent logins by a particular user or
          group of users.  This mechanism does not need  to  be  explicitly
          built  into  the  AAA protocol, but the AAA protocol must provide
          sufficient authorization information for an AAA  server  to  make
          that determination through an out-of-band mechanism.
PRC: Working on the draft with Glen and previous exercises with Merit to come
up with such a scheme for inter-domain has lead me to believe that this is
a very difficult problem. However, one can get similar functionality if you
use the scheme described in the draft that Glen and I put together.

     8.3.2.2.   Authorization  Attribute  Requirements - Authorization Pro-
     files
[...]
          -Protocol Filter application:  The AAA protocol  MUST  provide  a
          means of applying IP protocol filters to user sessions.  Two dif-
          ferent methods MUST be supported.

          First, the AAA protocol MUST provide a means of selecting a  pro-
          tocol  filter  by reference to an identifier, with the details of
          the filter action being specified out of band.  The AAA  protocol
          SHOULD define this out-of-band reference mechanism.
PRC: This smells of scaling problems.

     8.3.3.1.  Security of Compulsory Tunnel Credentials


     When an AAA protocol passes credentials that will be used to authenti-
     cate compulsory tunnels, the AAA protocol  MUST  provide  a  means  of
     securing the credentials from end to end of the AAA conversation.  The
     AAA protocol MUST also provide protection against  replay  attacks  in
     this situation.
PRC: Is this the authentication blob, or the username?

     8.4.1.1.  Guaranteed Delivery


     The accounting and auditing functions of the AAA protocol are used for
     network planning, resource management,  policy  decisions,  and  other
     functions  that  require  accurate  knowledge of the state of the NAS.
     NAS operators need to be able to engineer their network usage measure-
     ment  systems  to  a predictable level of accuracy.  Therefore, an AAA
     protocol MUST provide a means of  guaranteed  delivery  of  accounting
     information between the NAS and the AAA Server(s).
PRC: My experience is that hop-by-hop delivery acknowledgement is the best way
since it reduces the traffic, and the end-to-end retransmissions known to
exist in RADIUS.

     8.4.1.4.  Accounting Time Stamps


     There may be delays associated with the delivery of accounting  infor-
     mation.   The NAS operator will desire to know the time an event actu-
     ally occurred, rather than simply the time when  notification  of  the
     event  was  received.  Therefore, the AAA protocol MUST carry an unam-
     biguous time stamp associated with each accounting event.   This  time
     stamp  MUST  be  unambiguous with regard to time zone.  Note that this
     assumes  that  the  NAS  has  access  to  a   correct   time   source.
PRC: Does it matter if an intermediate node changes the time? Or does the time
have the be protected end-to-end?

     8.4.3.2.  Non-repudiation

     Network operators use accounting data for network  planning,  resource
     management,  and other business-critical functions that require confi-
     dence in the correctness of this data. The AAA protocol SHOULD provide
     a  mechanism to ensure that the source of Accounting data cannot repu-
     diate this data after transmission.
PRC: non-repudiation of both the source and the destination means that sufficient
data must be present in the accounting response. This allows the sender to
prove that the peer did in fact receive, and acknowledge receipt of the
accounting data.

--525316457.939256081980.JavaMail.root@westapp
Content-Type: text/plain; name=draft-ietf-nasreq-criteria-02.txt
Content-Disposition: attachment; filename=draft-ietf-nasreq-criteria-02.txt
Content-Transfer-Encoding: 7bit

     NASREQ Working Group                              M. Beadles
     INTERNET-DRAFT                UUNET, an MCI WorldCom Company
     Category: Informational
     <draft-ietf-nasreq-criteria-02.txt>
     6 October 1999


            Criteria for Evaluating Network Access Server Protocols


     1.  Status of this Memo


     This document is an Internet-Draft and is in full conformance with all
     provisions of Section 10 of RFC2026.  Internet-Drafts are working doc-
     uments  of  the Internet Engineering Task Force (IETF), its areas, and
     its working groups.  Note that other groups may also distribute  work-
     ing documents as Internet-Drafts.

     Internet-Drafts  are draft documents valid for a maximum of six months
     and may be updated, replaced, or obsoleted by other documents  at  any
     time.   It is inappropriate to use Internet- Drafts as reference mate-
     rial or to cite them other than as "work in progress."

     The   list   of   current   Internet-Drafts   can   be   accessed   at
     http://www.ietf.org/ietf/1id-abstracts.txt

     The  list  of  Internet-Draft  Shadow  Directories  can be accessed at
     http://www.ietf.org/shadow.html.

     The  distribution  of  this  draft  is  unlimited.   It  is  filed  as
     <draft-ietf-nasreq-criteria-02.txt>   and   expires  April  06,  2000.
     Please send comments to the author.


     2.  Copyright Statement


     Copyright   (C) The Internet Society 1999.  All Rights Reserved.


     3.  Abstract


     This document defines  requirements  for  protocols  used  by  Network
     Access  Servers  (NAS).   Protocols  used by NAS's may be divided into
     four spaces:  Access protocols, Network protocols, AAA protocols,  and
     Management  protocols.  Primary attention is given to setting require-
     ments for AAA protocols, since that space is currently the least  well
     defined.







     Beadles                 Category: Informational               [Page 1]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     4.  Requirements language


     In  this document, the key words "MAY", "MUST, "MUST NOT", "optional",
     "recommended", "SHOULD", and "SHOULD NOT", are to  be  interpreted  as
     described in [KEYWORDS].


     5.  Introduction


     This  document  defines  requirements  for  protocols  used by Network
     Access Servers (NAS).Protocols used by NAS's may be divided into  four
     spaces:   Access  protocols,  Network  protocols,  AAA  protocols, and
     Device Management protocols.   The primary focus of this  document  is
     on AAA protocols.  The reference model of a NAS used by this document,
     and the analysis of the functions of a NAS which led to  the  develop-
     ment of these requirements, may be found in [NAS-MODEL].
PRC: Nit. Space is missing before Protocols in second sentence.

     6.  Access Protocol Requirements


     There  are three basic types of access protocols used by NAS's.  First
     are the traditional telephony-based access protocols, which  interface
     to  the  NAS  via a modem or terminal adapter or similar device. These
     protocols typically support asynchronous or synchronous PPP [PPP] car-
     ried over a telephony protocol.
PRC: What is a telephony protocol? Q.931?
     Second are broadband pseudo-telephony
     access protocols, which are carried over xDSL  or  cable  modems,  for
     example.   These  protocols  typically support an encapsulation method
     such as PPP over Ethernet [PPPOE].  Finally  are  the  virtual  access
     protocols  used  by NAS's that terminate tunnels.  One example of this
     type of protocol is L2TP [L2TP].
PRC: What about cellular devices, such as in the 3G model. Initially the
devices will connect via PPP, but there is talk to also support HDLC in
order to get rid of PPP (since we that is needed is the framing).

     It is a central assumption of the NAS  model  used  here  that  a  NAS
     accepts multiple point-to-point links via one of the above access pro-
     tocols.  Therefore, at a minimum, any NAS access protocol MUST be able
     to  carry  PPP.   The  exception to this requirement is for NAS's that
     support legacy text login methods such as telnet [TELNET], rlogin,  or
     LAT.   Only  these access protocols are exempt from the requirement to
     support PPP.
PRC: So have we abandoned SLIP? I am not stating that we need to support it.
I am just curious.


     7.  Network Protocol Requirements


     The network protocols supported by a NAS depend entirely on  the  kind
     of network to which a NAS is providing access.  This document does not
     impose any additional requirements on  network  protocols  beyond  the
     protocol specifications themselves.  For example, if a NAS that serves
     a routed network includes internet routing  functionality,  then  that
     NAS must adhere to [ROUTING-REQUIREMENTS], but there are no additional
     protocol requirements imposed by virtue of the  device  being  a  NAS.




     Beadles                 Category: Informational               [Page 2]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     8.  AAA Protocol Requirements



     8.1.  General protocol characteristics


     There  are  certain general characteristics that any AAA protocol used
     by NAS's must meet.  Note that the transport requirements for  authen-
     tication/authorization  are  not  necessarily  the  same  as those for
     accounting/auditing.  An AAA protocol suite MAY use the same transport
     and protocol for both functions, but this is not strictly required.


     8.1.1.  Transport requirements



     8.1.1.1.  Transport independence

     The AAA protocol MUST be transport independent, and MUST be capable of
     using both TCP and UDP as a transport.   Additionally, the AAA  proto-
     col  SHOULD  be designed to easily support any new transport protocols
     developed by the Internet standards community.
Having an AAA protocol support both TCP and UDP has brought *lots* of criticism
from the likes of Fred Baker. In fact, due to such pressures one protocol that
comes to mind now only supports UDP.

     8.1.1.2.  Long-term exchange of small packets


     Very large scale NAS's that serve up to thousands of simultaneous ses-
     sions  are now being deployed.  This means that, in the extreme, there
     may be an almost constant exchange of many small packets  between  the
     NAS  and  the  AAA  server.   An AAA protocol transport SHOULD support
     being optimized for a long-term exchange of small packets in a  stream
     between a pair of hosts.
PRC: So are we talking about multiple sub-requests within a single request?
If so, this adds some interesting problems in roaming environments where part
of a request goes to A, and the other part to B.


     8.1.1.3.  Support for multiple AAA servers


     In  order to operationally support large loads, load balancing to mul-
     tiple AAA servers will be required.  The AAA protocol MUST provide for
     NAS's  to  balance  AAA sessions between two or more AAA servers.  The
     load balancing mechanism SHOULD  be  built  in  to  the  AAA  protocol
     itself.   The AAA protocol design MUST NOT introduce a single point of
     failure during the AAA process.  The AAA protocol MUST allow any  ses-
     sions between a NAS and a given AAA server to fail over to a secondary
     server without loss of state information.   This  fail-over  mechanism
     SHOULD be built in to the AAA protocol itself.
PRC: I think that is *also* needed here is a method to flow the traffic between
the NAS and the server. Otherwise, flodding a server that is already
busy is pointless. This would allow for a more effective load balancing
mechanism.









     Beadles                 Category: Informational               [Page 3]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     8.1.1.4.  Support for Multiple Administrative Domains


     NAS's  operated  by  one authority provide network access services for
     clients operated by another authority, to network  destinations  oper-
     ated by yet another authority.  This type of arrangement is of growing
     importance; for example, dial roaming is now a nearly ubiquitous  ser-
     vice.   Therefore,  the  AAA  protocol  MUST support AAA services that
     travel between multiple domains of authority.  The AAA  protocol  MUST
     NOT  use  a  model that assumes a single domain of authority.  The AAA
     protocol MUST NOT dictate particular business models for the relation-
     ship  between  the administrative domains.  The AAA protocol MUST sup-
     port proxy, and in additiona SHOULD support other  multi-domain  rela-
     tionships such as brokering and referral.
PRC: Should read "addition"

     The AAA protocol MUST also meet the protocol requirements specified in
     [ROAMING-REQUIREMENTS].


     8.1.2.  Attribute-Value Protocol Model


     Years of operational experience  with  AAA  protocols  and  NAS's  has
     proven that the Attribute-Value protocol model is an optimal represen-
     tation of AAA data.  The protocol SHOULD use an Attribute-Value repre-
     sentation for AAA data.  This document will assume such a model.  Even
     if the AAA protocol does not use this as an on-the-wire data represen-
     tation,  Attribute-Value  can  serve as abstraction for discussing AAA
     information.

     Experience has also shown  that  attribute  space  tends  to  run  out
     quickly.  In  order  to  provide  room  for expansion in the attribute
     space, the AAA protocol MUST support a minimum of 64K  Attributes  (16
     bits), each with a minimum length of 64K (16 bits).
PRC: Of course, my preference here is 32 bit since if the protocol is properly
aligned on a 32-bit boundary, it will be more CPU friendly. This will allow

for a higher number of transactions/sec.

     8.1.2.1.  Attribute Data Types


     The  AAA  protocol MUST support simple attribute data types, including
     integer, enumeration, text string, IP address, and date/time. The  AAA
     protocol  MUST  also  provide some support for complex structured data
     types. Wherever IP addresses are carried within the AAA protocol,  the
     protocol  MUST  support  both IPv4 and IPv6 [IPV6] addresses. Wherever
     text information is carried within the AAA protocol, the protocol MUST
     comply  with  the  IETF  Policy  on  Character Sets and Languages [RFC
     2277].
PRC: The requirement for IP V6 is interesting. Note that one data type is just
a series of octets, not necessarily a string (e.g. challenge).


     8.1.2.2.  Minimum Set of Attributes


     At a minimum, the AAA protocol MUST support, or be easily extended  to
     support, the set of attributes supported by RADIUS [RADIUS] and RADIUS



     Beadles                 Category: Informational               [Page 4]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     Accounting [RADIUS-ACCOUNTING].  If the base  AAA  protocol  does  not
     support  this  complete  set  of attributes, then an extension to that
     protocol MUST be defined which supports this set.


     8.1.2.3.  Attribute Extensibility


     NAS and AAA development is always progressing.  In  order  to  prevent
     the  AAA  protocol  from being a limiting factor in NAS and AAA Server
     development, the AAA protocol MUST provide  a  built-in  extensibility
     mechanism,  which  MUST  include  a  means  for  adding  new  standard
     attribute extensions.  This MUST include a method for  registering  or
     requesting  extensions  through  IANA, so that long-term working group
     involvement is not required to create new attribute  types.   Ideally,
     the  AAA  protocol SHOULD separate specification of the transport from
     specificatoin of the attributes.

     The AAA protocol MUST include a means for individual  vendors  to  add
     value  through  vendor-specific  attributes and SHOULD include support
     for vendor-specific data types.
PRC: No mention for vendor specific commands? Is this also a requirement?


     8.1.3.  Security Requirements



     8.1.3.1.  Mutual Authentication


     It is poor security practice for a NAS  to  communicate  with  an  AAA
     server  that  is  not  trusted, and vice versa.  The AAA protocol MUST
     provide mutual authentication between AAA server and NAS.
PRC: Although it should be possible for the protocol to make use of an underlying
security protocol, such as AH/EAP. However, in practice this is more difficult
since the application cannot guarantee that the policy is correctly setup and
could in fact be sending sensitive information unprotected in the clear. However,
if a network operator has deployed IP Security, the protocol should allow it.

     8.1.3.2.  Shared Secrets


     At a minimum, the AAA protocol SHOULD support use of a  secret  shared
     pairwise  between each NAS and AAA server to mutually verify identity.
     This is intended for small-scale deployments.


     8.1.3.3.  Public Key Security


     AAA server/NAS identity verification based solely  on  shared  secrets
     can  be  difficult  to  deploy  properly at large scale, and it can be
     tempting for NAS operators to use a single shared secret (that  rarely
     changes)  across  all  NAS's.  This can lead to easy compromise of the
     secret.  Therefore, the AAA protocol MUST also support mutual  verifi-
     cation  of  identity  using  a public-key infrastructure that supports
     expiration and revocation of keys.
PRC: I think that for hop-by-hop security, IP Security would provide this
and would allow for re-keying. Should the AAA protocol also provide this?
I believe that it may in inter-domain cases since it MAY be a requirement
to include non-repudiation within an accounting record, and this wouldn't
be possible with IP Security.




     Beadles                 Category: Informational               [Page 5]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     8.1.3.4.  Encryption of Attributes


     Some attributes are more operationally sensitive than  others.   Also,
     in a multi-domain scenario, attributes may be inserted by servers from
     different administrative domains. Therefore,  the  AAA  protocol  MUST
     support   selective  encryption  of  attributes  on  an  attribute-by-
     attribute basis,  even  within  the  same  message.  This  requirement
     applies equally to Authentication, Authorization, and Accounting data.


     8.2.  Authentication and User Security Requirements



     8.2.1.  Authentication protocol requirements


     End users who are requesting network access through a NAS will present
     various  types  of credentials.  It is the purpose of the AAA protocol
     to transport these credentials between the NAS and the AAA server.


     8.2.1.1.  Bi-directional Authentication


     The AAA protocol MUST support transport of credentials  from  the  AAA
     server to the NAS for the purpose of mutual (bi-directional) authenti-
     cation between the User and the NAS, and between the NAS and  the  AAA
     server.


     8.2.1.2.  Dynamic Authentication


     The AAA protocol MUST support re-authentication at any time during the
     course of a session, initiated from either end of the user session.


     8.2.1.3.  Multi-phase Authentication


     The AAA protocol MUST be able to  support  multi-phase  authentication
     methods, including but not limited to support for:
          -Text prompting from the NAS to the user

          -A series of binary challenges and responses of arbitrary length

          -An  authentication failure reason to be transmitted from the NAS
          to the user

          -Callback to a pre-determined phone number





     Beadles                 Category: Informational               [Page 6]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     8.2.1.4.  Extensible Authentication Types


     Security protocol development is going on constantly  as  new  threats
     are  identified  and  better  cracking methods are developed.  Today's
     secure authentication methods may be proven  insecure  tomorrow.   The
     AAA protocol MUST provide some support for addition of new authentica-
     tion credential types.
PRC: Should we simply call for EAP here? Or are we talking about AAA-transport
security?

     8.2.2.  Authentication Attribute Requirements


     In addition to the minimum attribute set, the AAA protocol  must  sup-
     port and define attributes that provide the following functions:


     8.2.2.1.  PPP Authentication protocols


     Many authentication protocols are defined within the framework of PPP.
     The AAA protocol MUST be able  to  act  as  an  intermediary  protocol
     between  the  authenticatee  and  the  authenticator for the following
     authentication protocols:

          -PPP Password Authentication Protocol [PPP]

          -PPP Challenge Handshake Authentication Protocol [CHAP]

          -PPP Extensible Authentication Protocol [EAP]


     8.2.2.2.  User Identification


     The following are common types of credentials used for user  identifi-
     cation.  The AAA protocol MUST be able to carry the following types of
     identity credentials:
          -A user name in the form of a Network Access Identifier [NAI].

          -An Extensible Authentication  Protocol  [EAP]  Identity  Request
          Type packet.

          -Telephony  dialing information such as Dialed Number Identifica-
          tion Service (DNIS) and Caller ID.
     If a particular type of identity credential is not needed for  a  par-
     ticular  user  session,  the  AAA protocol MUST NOT require that dummy
     credentials be filled in.  That is,  the  AAA  protocol  MUST  support
     identification without authentication.
PRC: Or perhaps the other way around?








     Beadles                 Category: Informational               [Page 7]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     8.2.2.3.  Authentication Credentials


     The following are common types of credentials used for authentication.
     The AAA protocol MUST be able to carry the following types of  authen-
     ticating credentials at a minimum:
          -A secret or password.

          -A response to a challenge presented by the NAS to the user

          -A one-time password

          -An X.509 digital certificate [X.509]

          -A Kerberos v5 ticket [KERBEROS]
PRC: I am not sure that I get it. I think that each authentication scheme
has it's own type of credential, so what is this section intended to state?

     8.2.3.  Authentication Protocol Security Requirements



     8.2.3.1.  End-to-End Hiding of Credentials


     Where passwords are used as authentication credentials, the AAA proto-
     col MUST provide a secure means of hiding the password from end to end
     of  the  AAA  conversation.   Where  challenge/response mechanisms are
     used, the AAA protocol MUST also prevent against replay attacks.

PRC: I would prefer to see end-to-end authentication. Glen and I have a draft that
talks about this, and this effectively eliminates the authentication replay problem
that occurs with CHAP/RADIUS today.

     8.3.  Authorization, Policy, and Resource management



     8.3.1.  Authorization Protocol Requirements


     In all cases, the protocol MUST specify that authorization  data  sent
     from  the  NAS  to  the AAA server is to be regarded as information or
     "hints", and not directives.   The AAA protocol MUST  be  designed  so
     that  the  AAA server makes all final authorization decisions and does
     not depend on a certain state being expected by the NAS.


     8.3.1.1.  Dynamic Authorization


     The AAA protocol MUST support dynamic  re-authorization  at  any  time
     during  a  user  session.   This  re-authorization may be initiated in
     either  direction.   This  dynamic  re-authorization  capability  MUST
     include  the  capability  to  request  a  NAS  to disconnect a user on
     demand.





     Beadles                 Category: Informational               [Page 8]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     8.3.1.2.  Resource Management


     Resource management MUST be supported on demand  by  the  NAS  or  AAA
     Server at any time during the course of a user session.
PRC: Inter-Domain resource management can be *very* difficult. However the
I-D that Glen and I put together makes it scale across the Internet by
requiring periodical re-authentication.

     8.3.2.  Authorization Attribute Requirements



     8.3.2.1.  Authorization Attribute Requirements - Access Restrictions


     The  AAA protocol serves as a primary means of gathering data used for
     making Policy decisions for network access.  Therefore, the AAA proto-
     col MUST allow network operators to make policy decisions based on the
     following parameters:

          -Time/day restrictions.  The AAA protocol MUST be able to provide
          an  unambiguous  time  stamp,  NAS time zone indication, and date
          indication to the AAA server in the Authorization information.
PRC: This is good, and the work in Secure NTP will certainly help. This is
especially true in inter-domain cases where assuming that all hosts are time
synched is difficult.

          -Location restrictions:  The AAA protocol MUST be able to provide
          an  unambiguous  location code that reflects the geographic loca-
          tion of the NAS.  Note that this is not the same type of thing as
          either the dialing or dialed station.
PRC: I hear privacy issues here, but perhaps it *could* be a requirement. I
am not sure that I'd want anyone to know that I am in a cyber-coffee shop
somewhere in the Netherlands.

          -Dialing  restrictions:  The AAA protocol MUST be able to provide
          accurate dialed and dialing station indications.

          -Concurrent login limitations:  The AAA protocol  MUST  allow  an
          AAA  Server  to  limit  concurrent logins by a particular user or
          group of users.  This mechanism does not need  to  be  explicitly
          built  into  the  AAA protocol, but the AAA protocol must provide
          sufficient authorization information for an AAA  server  to  make
          that determination through an out-of-band mechanism.
PRC: Working on the draft with Glen and previous exercises with Merit to come
up with such a scheme for inter-domain has lead me to believe that this is
a very difficult problem. However, one can get similar functionality if you
use the scheme described in the draft that Glen and I put together.

     8.3.2.2.   Authorization  Attribute  Requirements - Authorization Pro-
     files


     The AAA protocol is used to enforce policy at the  NAS.   Essentially,
     on  granting  of access, a particular access profile is applied to the
     user's session.  The AAA protocol MUST at a minimum provide a means of
     applying profiles containing the following types of information:

          -IP  Address assignment: The AAA protocol MUST provide a means of
          assigning an IPv4 or IPv6 address to an incoming user.

          -Protocol Filter application:  The AAA protocol  MUST  provide  a
          means of applying IP protocol filters to user sessions.  Two dif-
          ferent methods MUST be supported.



     Beadles                 Category: Informational               [Page 9]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


          First, the AAA protocol MUST provide a means of selecting a  pro-
          tocol  filter  by reference to an identifier, with the details of
          the filter action being specified out of band.  The AAA  protocol
          SHOULD define this out-of-band reference mechanism.
PRC: This smells of scaling problems.

          Second, the AAA protocol MUST provide a means of passing a proto-
          col filter by value.  This means explicit passing  of  pass/block
          information  by address range, TCP/UDP port number, and IP proto-
          col number at a minimum.

          -Compulsory Tunneling:  The AAA protocol MUST provide a means  of
          directing  a NAS to build a tunnel or tunnels to a specified end-
          point.  It MUST support creation of multiple simultaneous tunnels
          in  a  specified  order.   The protocol MUST allow, at a minimum,
          specification of the tunnel endpoints, tunneling  protocol  type,
          underlying  tunnel  media type, and tunnel authentication creden-
          tials (if required by the tunnel type).  The  AAA  protocol  MUST
          support  at  least the creation of tunnels using the L2TP [L2TP],
          ESP [ESP], and AH [AH]  protocols.   The  protocol  MUST  provide
          means of adding new tunnel types as they are standardized.

          -Routing:   The  AAA protocol MUST provide a means of assigning a
          particular static route to an incoming user session.

          -Expirations/timeouts:  The AAA protocol MUST provide a means  of
          communication  session expiration information to a NAS.  Types of
          expirations that MUST be supported are:  total session time, idle
          time, total bytes transmitted, and total bytes received.

          -Quality  of  Service:   The AAA protocol MUST provide a means of
          applying Quality of Service parameters to  individual  user  ses-
          sions.


     8.3.2.3.  Resource Management Requirements


     The  AAA  protocol is a means for network operators to perform manage-
     ment of network resources.  The AAA protocol MUST provide a  means  of
     collecting  resource state information, and controlling resource allo-
     cation for the following types of network resources.

          -Network bandwidth usage per session,  including  multilink  ses-
          sions.

          -Access port usage, including concurrent usage and usage pools.

          -Connect time.

          -IP Addresses and pools.

          -Compulsory tunnel limits.





     Beadles                 Category: Informational              [Page 10]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     8.3.3.  Authorization Protocol Security Requirements



     8.3.3.1.  Security of Compulsory Tunnel Credentials


     When an AAA protocol passes credentials that will be used to authenti-
     cate compulsory tunnels, the AAA protocol  MUST  provide  a  means  of
     securing the credentials from end to end of the AAA conversation.  The
     AAA protocol MUST also provide protection against  replay  attacks  in
     this situation.
PRC: Is this the authentication blob, or the username?

     8.4.  Accounting and Auditing Requirements



     8.4.1.  Accounting Protocol Requirements



     8.4.1.1.  Guaranteed Delivery


     The accounting and auditing functions of the AAA protocol are used for
     network planning, resource management,  policy  decisions,  and  other
     functions  that  require  accurate  knowledge of the state of the NAS.
     NAS operators need to be able to engineer their network usage measure-
     ment  systems  to  a predictable level of accuracy.  Therefore, an AAA
     protocol MUST provide a means of  guaranteed  delivery  of  accounting
     information between the NAS and the AAA Server(s).
PRC: My experience is that hop-by-hop delivery acknowledgement is the best way
since it reduces the traffic, and the end-to-end retransmissions known to
exist in RADIUS.

     8.4.1.2.  Real Time Accounting


     NAS  operators  often require a real time view onto the status of ses-
     sions served by a NAS.  Therefore, the AAA protocol MUST support real-
     time  delivery  of  accounting and auditing information.  In this con-
     text, real time is defined as accounting information  delivery  begin-
     ning within one second of the triggering event.


     8.4.1.3.  Batch Accounting


     The AAA protocol SHOULD also support delivery of stored accounting and
     auditing information in batches (non-real time).








     Beadles                 Category: Informational              [Page 11]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     8.4.1.4.  Accounting Time Stamps


     There may be delays associated with the delivery of accounting  infor-
     mation.   The NAS operator will desire to know the time an event actu-
     ally occurred, rather than simply the time when  notification  of  the
     event  was  received.  Therefore, the AAA protocol MUST carry an unam-
     biguous time stamp associated with each accounting event.   This  time
     stamp  MUST  be  unambiguous with regard to time zone.  Note that this
     assumes  that  the  NAS  has  access  to  a   correct   time   source.
PRC: Does it matter if an intermediate node changes the time? Or does the time
have the be protected end-to-end?

     8.4.1.5.  Accounting Events


     At a minimum, the AAA protocol MUST  support  delivery  of  accounting
     information triggered by the following events:
          -Start of a user session

          -End of a user session

          -Expiration  of  a predetermined repeating time interval during a
          user session.  The AAA protocol MUST provide a means for the  AAA
          server  to  request  that a NAS use a certain interval accounting
          time.

          -Dynamic  re-authorization  during  a  user  session  (e.g..,  new
          resources being delivered to the user)

          -Dynamic re-authentication during a user session


     8.4.1.6.  On-demand Accounting


     NAS  operators  need  to  maintain an accurate view onto the status of
     sessions served by a NAS, even  through  failure  of  an  AAA  server.
     Therefore, the AAA protocol MUST support a means of requesting current
     session state and accounting from the NAS on demand.


     8.4.2.  Accounting attribute requirements


     At a minimum, the AAA protocol MUST support delivery of the  following
     types of accounting/auditing data:
          -All parameters used to authenticate a session.

          -Details  of  the  authorization  profile that was applied to the
          session.

          -The duration of the session.

          -The cumulative number of bytes  sent  by  the  user  during  the



     Beadles                 Category: Informational              [Page 12]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


          session.

          -The  cumulative  number of bytes received by the user during the
          session.

          -The cumulative number of packets sent by  the  user  during  the
          session.

          -The cumulative number of packets received by the user during the
          session.

          -Details of the access protocol used  during  the  session  (port
          type, connect speeds, etc.)


     8.4.3.  Accounting Protocol Security Requirements



     8.4.3.1.  Integrity and Confidentiality


     Note  that  accounting  and  auditing data are operationally sensitive
     information.  The AAA protocol MUST provide a means to assure  end-to-
     end  integrity  of this data.  The AAA protocol SHOULD provide a means
     of assuring the end-to-end confidentiality of this data.


     8.4.3.2.  Non-repudiation


     Network operators use accounting data for network  planning,  resource
     management,  and other business-critical functions that require confi-
     dence in the correctness of this data. The AAA protocol SHOULD provide
     a  mechanism to ensure that the source of Accounting data cannot repu-
     diate this data after transmission.
PRC: non-repudiation of both the source and the destination means that sufficient
data must be present in the accounting response. This allows the sender to
prove that the peer did in fact receive, and acknowledge receipt of the
accounting data.

     9.  Device Management Protocols


     This document does not specify any requirements for device  management
     protocols.


     10.  Acknowledgments

     Many  of  the  requirements  in  this document first took form in Glen
     Zorn's "Yet Another  Authentication  Protocol  (YAAP)"  document,  for
     which grateful acknowledgment is made.







     Beadles                 Category: Informational              [Page 13]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     11.  Security considerations


     See above for security requirements for the NAS AAA protocol.

     Where  an  AAA  architecture  spans multiple domains of authority, AAA
     information may need to cross trust boundaries.  In this situation,  a
     NAS  might  operate as a shared device that services multiple adminis-
     trative domains.  Network operators are advised take this into consid-
     eration when deploying NAS's and AAA Servers.


     12.  IANA Considerations


     This document does not directly specify any IANA considerations.  How-
     ever, the following recommendations are made:

     Future development and extension of an AAA protocol will be made  much
     easier  if  new  attributes  and values can be requested or registered
     directly through IANA, rather than  through  an  IETF  Standardization
     process.

     The  AAA  protocol  might  use  enumerated values for some attributes,
     which enumerate already-defined IANA types (such as protocol  number).
     In  these cases, the AAA protocol SHOULD use the IANA assigned numbers
     as the enumerated values.



     13.  References


     [KEYWORDS] S. Bradner.   "Key  words  for  use  in  RFCs  to  Indicate
     Requirement Levels."  RFC 2119, Harvard University, March 1997.

     [NAS-MODEL]  D.  Mitton,  M. Beadles.  "Network Access Server Require-
     ments Next Generation (NASREQNG) NAS Model."  Work in progress.

     [PPPOE] L. Mamakos et al.  "A Method for Transmitting PPP Over  Ether-
     net (PPPoE)."  RFC 2516, UUNET Technologies, Inc., February 1999.

     [L2TP] W. M. Townsley, et al.  "Layer Two Tunneling Protocol  (L2TP)."
     RFC 2661, IBM, Cisco, Ascend, Microsoft, August 1999.

     [PPP] W.  Simpson.  "The Point-to-Point Protocol  (PPP)."   RFC  1661,
     Daydreamer, July 1994.

     [TELNET]  J.  Postel,  J.  Reynolds.  "Telnet Protocol Specification."
     STD 8, RFC 854, ISI, May 1983.

     [ROUTING-REQUIREMENTS] F.  Baker.   "Requirements  for  IP  Version  4
     Routers."  RFC 1812, Cisco Systems, June 1995.




     Beadles                 Category: Informational              [Page 14]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     [IPV6]  S.  Deering,  R. Hinden.  "Internet Protocol, Version 6 (IPv6)
     Specification."  RFC 2460, Cisco, Nokia, December 1998.

     [RFC 2277] H. Alvestrand.  "IETF Policy on  Character  Sets  and  Lan-
     guages."  RFC 2277, UNINETT, January 1998.

     [CHAP]  W.  Simpson.  "PPP Challenge Handshake Authentication Protocol
     (CHAP)."  RFC 1994, Daydreamer, August 1996.

     [EAP] L. Blunk, J. Vollbrecht.  "PPP Extensible Authentication  Proto-
     col (EAP)."  RFC 2284, Merit Network, Inc., March 1998.

     [NAI]  B.   Aboba,  M. Beadles.  "The Network Access Identifier."  RFC
     2486, Microsoft, WorldCom Advanced Networks, January 1999.

     [X.509] ITU-T Recommendation X.509 (1997 E): Information Technology  -
     Open  Systems  Interconnection  - The Directory: Authentication Frame-
     work, June 1997.

     [KERBEROS] J. Kohl, C. Neuman.  "The Kerberos  Network  Authentication
     Service  (V5)."  RFC 1510, Digital Equipment Corporation, ISI, Septem-
     ber 1993.

     [ESP] S.  Kent,  R.  Atkinson.   "IP  Encapsulating  Security  Payload
     (ESP)."  RFC 2406, BBN Corp, @Home Network, November 1998.

     [AH]  S.  Kent,  R.  Atkinson.   "IP Authentication Header (AH)."  RFC
     2402, BBN Corp, @Home Network, November 1998.

     [ROAMING-REQUIREMENTS] B. Aboba, G. Zorn.   "Criteria  for  Evaluating
     Roaming Protocols."  RFC 2477, Microsoft, January 1999.

     [RADIUS]

     [RADIUS-ACCOUNTING]
PRC: Was this intentional? :)


     14.  Author's Address



     Mark Anthony Beadles
     UUNET, an MCI WorldCom Company
     5000 Britton Rd.
     Hilliard, OH 43026

     Phone: 614-723-1941
     EMail: mbeadles@wcom.net









     Beadles                 Category: Informational              [Page 15]





     INTERNET-DRAFT        Criteria for NAS Protocols        6 October 1999


     15.  Full Copyright Statement


     Copyright (C) The Internet Society (1999).  All Rights Reserved.

     This  document  and  translations of it may be copied and furnished to
     others, and derivative works that comment on or otherwise  explain  it
     or  assist in its implmentation may be prepared, copied, published and
     distributed, in whole or in part, without  restriction  of  any  kind,
     provided  that  the  above  copyright  notice  and  this paragraph are
     included on all such copies and derivative works.  However, this docu-
     ment  itself  may  not be modified in any way, such as by removing the
     copyright notice or references to the Internet Society or other Inter-
     net  organizations,  except  as  needed  for the purpose of developing
     Internet standards in which case the procedures for copyrights defined
     in  the Internet Standards process must be followed, or as required to
     translate it into languages other than   English.  The limited permis-
     sions  granted  above  are  perpetual  and  will not be revoked by the
     Internet Society or its successors or assigns.  This document and  the
     information  contained  herein is provided on an "AS IS" basis and THE
     INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL
     WARRANTIES,  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WAR-
     RANTY THAT THE USE OF THE INFORMATION HEREIN  WILL  NOT  INFRINGE  ANY
     RIGHTS  OR  ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A
     PARTICULAR PURPOSE."


     16.  Expiration Date


     This document is  filed  as  <draft-ietf-nasreq-criteria-02.txt>,  and
     expires April 06, 2000.

























     Beadles                 Category: Informational              [Page 16]



--525316457.939256081980.JavaMail.root@westapp--



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct  7 08:12:04 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id IAA12849;
	Thu, 7 Oct 1999 08:12:04 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id HAA04658
	for nasreq-outgoing; Thu, 7 Oct 1999 07:45:12 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from ietf.org (odin.ietf.org [132.151.1.176])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id HAA04654
	for <nasreq@tdmx.rutgers.edu>; Thu, 7 Oct 1999 07:45:10 -0400 (EDT)
Received: from CNRI.Reston.VA.US (localhost [127.0.0.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id GAA10616;
	Thu, 7 Oct 1999 06:53:28 -0400 (EDT)
Message-Id: <199910071053.GAA10616@ietf.org>
Mime-Version: 1.0
Content-Type: Multipart/Mixed; Boundary="NextPart"
To: IETF-Announce: ;
Cc: nasreq@tdmx.rutgers.edu
From: Internet-Drafts@ietf.org
Reply-to: Internet-Drafts@ietf.org
Subject: I-D ACTION:draft-ietf-nasreq-criteria-02.txt
Date: Thu, 07 Oct 1999 06:52:07 -0400
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

--NextPart

A New Internet-Draft is available from the on-line Internet-Drafts directories.
This draft is a work item of the Network Access Server Requirements Working Group of the IETF.

	Title		: Criteria for Evaluating Network Access Server 
                          Protocols
	Author(s)	: M. Beadles
	Filename	: draft-ietf-nasreq-criteria-02.txt
	Pages		: 16
	Date		: 06-Oct-99
	
This document defines  requirements  for  protocols  used  by  Network
Access  Servers  (NAS).   Protocols  used by NAS's may be divided into
four spaces:  Access protocols, Network protocols, AAA protocols,  and
Management  protocols.  Primary attention is given to setting require-
ments for AAA protocols, since that space is currently the least  well
defined.

A URL for this Internet-Draft is:
http://www.ietf.org/internet-drafts/draft-ietf-nasreq-criteria-02.txt

Internet-Drafts are also available by anonymous FTP. Login with the username
"anonymous" and a password of your e-mail address. After logging in,
type "cd internet-drafts" and then
	"get draft-ietf-nasreq-criteria-02.txt".

A list of Internet-Drafts directories can be found in
http://www.ietf.org/shadow.html 
or ftp://ftp.ietf.org/ietf/1shadow-sites.txt


Internet-Drafts can also be obtained by e-mail.

Send a message to:
	mailserv@ietf.org.
In the body type:
	"FILE /internet-drafts/draft-ietf-nasreq-criteria-02.txt".
	
NOTE:	The mail server at ietf.org can return the document in
	MIME-encoded form by using the "mpack" utility.  To use this
	feature, insert the command "ENCODING mime" before the "FILE"
	command.  To decode the response(s), you will need "munpack" or
	a MIME-compliant mail reader.  Different MIME-compliant mail readers
	exhibit different behavior, especially when dealing with
	"multipart" MIME messages (i.e. documents which have been split
	up into multiple messages), so check your local documentation on
	how to manipulate these messages.
		
		
Below is the data which will enable a MIME compliant mail reader
implementation to automatically retrieve the ASCII version of the
Internet-Draft.

--NextPart
Content-Type: Multipart/Alternative; Boundary="OtherAccess"

--OtherAccess
Content-Type: Message/External-body;
	access-type="mail-server";
	server="mailserv@ietf.org"

Content-Type: text/plain
Content-ID:	<19991006104504.I-D@ietf.org>

ENCODING mime
FILE /internet-drafts/draft-ietf-nasreq-criteria-02.txt

--OtherAccess
Content-Type: Message/External-body;
	name="draft-ietf-nasreq-criteria-02.txt";
	site="ftp.ietf.org";
	access-type="anon-ftp";
	directory="internet-drafts"

Content-Type: text/plain
Content-ID:	<19991006104504.I-D@ietf.org>

--OtherAccess--

--NextPart--




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct  7 12:19:05 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA20031;
	Thu, 7 Oct 1999 12:19:05 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id LAA06114
	for nasreq-outgoing; Thu, 7 Oct 1999 11:49:09 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from smtpgw02.wanmail.net (smtpgw02.wanmail.net [209.154.194.61])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id LAA06110
	for <nasreq@tdmx.rutgers.edu>; Thu, 7 Oct 1999 11:49:04 -0400 (EDT)
Received: from wcomexch1.wcom.net (wcomexch1.wcom.net [209.154.236.12])
	by smtpgw02.wanmail.net (8.9.1/8.9.1) with ESMTP id LAA12115
	for <nasreq@tdmx.rutgers.edu>; Thu, 7 Oct 1999 11:48:33 -0400 (EDT)
Received: by wcomexch1.wcom.net with Internet Mail Service (5.5.2448.0)
	id <SAX9F7C3>; Thu, 7 Oct 1999 11:47:35 -0400
Message-ID: <9A74C89D6696D2118E500008C7BA7C560C03B17C@wcomexch1.wcom.net>
From: "Beadles, Mark A." <MBeadles@wcom.net>
To: "Nasreqng List (E-mail)" <nasreq@tdmx.rutgers.edu>
Subject: nasreq FW: A few words on AAA WG relationship with other WGs and standar
	ds bodies
Date: Thu, 7 Oct 1999 11:47:34 -0400 
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

FYI from co-chair of AAA WG.

+ Mark Anthony Beadles + Co-chair, IETF NASREQ WG   + 
+  mbeadles@wcom.net   + http://nasreq.rutgers.edu/ +
+  Voice 614.723.1941  +       Fax 614.723.8407     +

-----Original Message-----
From: Bernard Aboba [mailto:aboba@internaut.com]
Sent: Wednesday, 06 October 1999 3:55 PM
To: aaa-wg@merit.edu
Subject: A few words on AAA WG relationship with other WGs and standards
bodies


I thought I'd say a few words outlining the
documents we will be incorporating from other
WGs and standards bodies, and describing the
process for incorporating that comments on
those submissions.  

The idea behind incorporating the work of other
groups is to leverage their expertise and energy
as well as to ensure that the requirements 
represent the consensus of those groups. This
ensures that we will design a base protocol
that those groups can then use in their own
work. It also provides a way of ensuring
proper coordination between AAA WG and other
groups. That is, we see AAA WG as a supplier to
those other groups, not an outside force 
re-designing their architecture. Defining a 
clear supplier-customer relationship lessens
our work load,  prevents duplication of
effort and reduces the likelihood of potential
conflicts with other groups. 

The external documents to be incorporated
include:

From NASREQ:
draft-ietf-nasreq-criteria-02.txt

From ROAMOPS:
RFC 2477

From Mobile IP (joint discussion with AAA):
draft-ietf-aaa-mobile-ip-req-00.txt

From TIA 45.6: 
Work in progress, not submitted yet.
contact: Tom Hiller, tom.hiller@lucent.com. 

In order to allow discussion of the
above inputs, it is highly desirable that
they get in some time before the Oct. 22
deadline. Getting them in by Oct. 15 would
both permit some time for AAA WG discussion
as well as possibly allow time for composition
of a summary document. 

Since the authors of the above documents 
may not necessarily be closely following 
the AAA WG mailing list, it makes sense 
to post comments to the relevant WG 
mailing list as well as possibly
to cc: the authors.  

Please note that if you are developing 
requirements that relate to the above
areas, and intend this to be taken into
account by the AAA WG, then you MUST get
your input incorporated by the authors of
the above documents. 

Note that the decision on whether to accept
your input is up to the WG and authors in
question. If your input is rejected by the
WG in question, you should not expect to
get that decision overturned in an appeal
to the AAA WG. Since we will focus on
incorporating documents from other WGs,
if you wish to present alternatives or
contest decisions of other WGs, then
the place to focus those energies is
within the other bodies, not AAA WG. 




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct  7 14:54:46 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA23818;
	Thu, 7 Oct 1999 14:54:46 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id OAA07221
	for nasreq-outgoing; Thu, 7 Oct 1999 14:14:30 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from drawbridge.ascend.com (drawbridge.ascend.com [198.4.92.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id OAA07217
	for <nasreq@tdmx.rutgers.edu>; Thu, 7 Oct 1999 14:14:29 -0400 (EDT)
Received: from fw-ext.ascend.com (fw-ext [198.4.92.5])
	by drawbridge.ascend.com (8.9.1a/8.9.1) with SMTP id LAA08434;
	Thu, 7 Oct 1999 11:09:14 -0700 (PDT)
Received: from russet.ascend.com by fw-ext.ascend.com
          via smtpd (for drawbridge.ascend.com [198.4.92.1]) with SMTP; 7 Oct 1999 18:14:22 UT
Received: from porky (porky.ascend.com [192.207.23.83])
	by russet.ascend.com (8.9.1a/8.9.1) with ESMTP id LAA13968;
	Thu, 7 Oct 1999 11:14:21 -0700 (PDT)
Received: from ascend.com by ascend.com
Message-Id: <4.2.0.58.19991007084106.00983220@porky>
X-Sender: mhold@porky
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 
Date: Thu, 07 Oct 1999 08:46:14 -0700
To: Pat.Calhoun@eng.sun.com
From: Matt Holdrege <matt@ascend.com>
Subject: Re: Comments on draft-ietf-nasreq-criteria-02.txt
Cc: nasreq@tdmx.rutgers.edu
In-Reply-To: <525314688.939256082195.JavaMail.pcalhoun@hsmpka.Eng.Sun.CO
 M>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

At 05:28 PM 10/6/99 -0700, Pat.Calhoun@eng.sun.com wrote:

>Mark,
>
>Here are my comments on the draft. Sorry they are so extensive that it
>looks like most of the I-D is included in this e-mail. I did it this way
>because in most cases some context is needed to understand the comment.
>
>Good work, btw.

I don't think I had any serious problems with Pat's concerns. But a high 
level question for me would be Radius backwards compatibility of this whole 
exercise.

I think that it is mandatory that our criteria and the solutions that 
result show backwards compatibility with RADIUS or it won't be of much 
value to the industry. OTOH, it is good that we describe the limitations of 
RADIUS in a way that points us to a better solution. Phrasing that can be 
difficult, but some of us have had a lot of practice lately. :)



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct  7 14:59:32 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA23873;
	Thu, 7 Oct 1999 14:59:32 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id OAA07406
	for nasreq-outgoing; Thu, 7 Oct 1999 14:27:12 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from dfssl.exchange.microsoft.com (dfssl.exchange.microsoft.com [131.107.88.59])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id OAA07402
	for <nasreq@tdmx.rutgers.edu>; Thu, 7 Oct 1999 14:27:10 -0400 (EDT)
Received: by dfssl with Internet Mail Service (5.5.2650.21)
	id <426VJTWJ>; Thu, 7 Oct 1999 11:26:24 -0700
Message-ID: <078292D50C98D2118D090008C7E9C6A60225812B@STAY.platinum.corp.microsoft.com>
From: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
To: "'Matt Holdrege'" <matt@ascend.com>, Pat.Calhoun@eng.sun.com
Cc: nasreq@tdmx.rutgers.edu
Subject: RE: Comments on draft-ietf-nasreq-criteria-02.txt
Date: Thu, 7 Oct 1999 11:26:21 -0700 
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

In terms of RADIUS backward compatibility, it's worthwhile
to describe the level of backward compatibility required,
and, as importantly, why it is needed. 

Some possibilities (in order of increasing stringency):

1. New protocol and RADIUS can listen on the same port. 
This style of compatibility was utilized in L2TP for
backwards compatibility with L2F. Note that in this
case the new and old protocols were quite different. 

2. It is possible to gateway between the new protocol and
legacy RADIUS devices. This enables use of a proxy as a
transition mechanism. This is the level of compatibility
required in RFC 2477. Note that this does not imply that
the new protocol be very close to RADIUS in design, only
that it be possible to build a gateway between them. An
analogy would be gateways between SMTP and proprietary
mail systems, which were popular in the 1980s. 

3. Legacy devices can speak RADIUS to the new
protocol server, and receive replies that they
will understand, while at the same time new
devices can speak the new protocol to the same
server. 

This is the level of backwards compatibility that
was achieved in DHCP, which superceeded BOOTP. 


-----Original Message-----
From: Matt Holdrege [mailto:matt@ascend.com]
Sent: Thursday, October 07, 1999 8:46 AM
To: Pat.Calhoun@eng.sun.com
Cc: nasreq@tdmx.rutgers.edu
Subject: Re: Comments on draft-ietf-nasreq-criteria-02.txt


At 05:28 PM 10/6/99 -0700, Pat.Calhoun@eng.sun.com wrote:

>Mark,
>
>Here are my comments on the draft. Sorry they are so extensive that it
>looks like most of the I-D is included in this e-mail. I did it this way
>because in most cases some context is needed to understand the comment.
>
>Good work, btw.

I don't think I had any serious problems with Pat's concerns. But a high 
level question for me would be Radius backwards compatibility of this whole 
exercise.

I think that it is mandatory that our criteria and the solutions that 
result show backwards compatibility with RADIUS or it won't be of much 
value to the industry. OTOH, it is good that we describe the limitations of 
RADIUS in a way that points us to a better solution. Phrasing that can be 
difficult, but some of us have had a lot of practice lately. :)


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct  7 17:31:58 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA25815;
	Thu, 7 Oct 1999 17:31:58 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id QAA08134
	for nasreq-outgoing; Thu, 7 Oct 1999 16:52:54 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id QAA08130
	for <nasreq@tdmx.rutgers.edu>; Thu, 7 Oct 1999 16:52:52 -0400 (EDT)
Received: from engmail2.Eng.Sun.COM ([129.146.1.25])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id NAA14618;
	Thu, 7 Oct 1999 13:52:50 -0700 (PDT)
Received: from hsmpka.eng.sun.com (phys-hsmpka.Eng.Sun.COM [129.146.95.37])
	by engmail2.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id NAA13695;
	Thu, 7 Oct 1999 13:52:49 -0700 (PDT)
Received: from hsmpka.Eng.Sun.COM by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id NAA07345; Thu, 7 Oct 1999 13:52:25 -0700
From: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Message-Id: <199910072052.NAA07345@hsmpka.eng.sun.com>
Date: Thu, 7 Oct 1999 13:52:26 -0700
To: "Matt Holdrege" <matt@ascend.com>, <nasreq@tdmx.rutgers.edu>
Cc: <pcalhoun@eng.sun.com>
Reply-To: <Pat.Calhoun@eng.sun.com>
Subject: nasreq Re: bounce
X-Mailer: Sun NetMail 2.3
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

I've CC'ed myself on this e-mail because pat.calhoun@eng seemed to stop working
this morning. Although I've fixed it, it will take 24 hours to take effect,
therefore I would appreciate if you reply to keep pcalhoun@eng.sun.com in the
recipient list.

I agree with you that describing the RADIUS limitations would be an interesting
exercise, and as it turns out I just finished doing so. The new DIAMETER 
framework document was recently re-written from scratch and has a whole
section on the know RADIUS problems. The document then continues to describe
how the DIAMETER protocol fixes these problems.

I agree that backward RADIUS compatibility would be nice, but complete
compatibility with the existing protocol is nearly impossible. What needs to
be done is to develop a protocol that can:
1. allow servers to perform protocol translation with minimal overhead
2. minimize configuration by re-using the existing RADIUS dictionary, and
3. allow RADIUS user profiles to be used unchanged.

More than this we are talking about no changes to the protocol, and that would
unfortunately not solve the known RADIUS problems.

PatC
BTW, The new framework (DIAMETER) document was re-written from scratch (finally)
and has a whole section on RADIUS problems, followed by a section on how 
DIAMETER solves them.

As far as backward compatibility
>>
>>>Mark,
>>>
>>>Here are my comments on the draft. Sorry they are so extensive that it
>>>looks like most of the I-D is included in this e-mail. I did it this way
>>>because in most cases some context is needed to understand the comment.
>>>
>>>Good work, btw.
>>
>>I don't think I had any serious problems with Pat's concerns. But a high 
>>level question for me would be Radius backwards compatibility of this 
>>whole exercise.
>>
>>I think that it is mandatory that our criteria and the solutions that 
>>result show backwards compatibility with RADIUS or it won't be of much 
>>value to the industry. OTOH, it is good that we describe the limitations 
>>of RADIUS in a way that points us to a better solution. Phrasing that can 
>>be difficult, but some of us have had a lot of practice lately. :)
>>
>




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Mon Oct 11 07:19:57 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA26854;
	Mon, 11 Oct 1999 07:19:56 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id GAA01238
	for nasreq-outgoing; Mon, 11 Oct 1999 06:59:19 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from ietf.org (odin.ietf.org [132.151.1.176])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id GAA01234
	for <nasreq@tdmx.rutgers.edu>; Mon, 11 Oct 1999 06:59:16 -0400 (EDT)
Received: from CNRI.Reston.VA.US (localhost [127.0.0.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id GAA26365;
	Mon, 11 Oct 1999 06:58:45 -0400 (EDT)
Message-Id: <199910111058.GAA26365@ietf.org>
Mime-Version: 1.0
Content-Type: Multipart/Mixed; Boundary="NextPart"
To: IETF-Announce: ;
Cc: nasreq@tdmx.rutgers.edu
From: Internet-Drafts@ietf.org
Reply-to: Internet-Drafts@ietf.org
Subject: I-D ACTION:draft-ietf-nasreq-nasmodel-01.txt
Date: Mon, 11 Oct 1999 06:58:45 -0400
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

--NextPart

A New Internet-Draft is available from the on-line Internet-Drafts directories.
This draft is a work item of the Network Access Server Requirements Working Group of the IETF.

	Title		: Network Access Server Requirements Next Generation  
                          (NASREQNG)NAS Model
	Author(s)	: D. Mitton, M. Beadles
	Filename	: draft-ietf-nasreq-nasmodel-01.txt
	Pages		: 19
	Date		: 08-Oct-99
	
This document describes the terminology and gives a model of typical
Network Access Server (NAS).  The purpose of this effort is to set the
reference space for describing and evaluating NAS service protocols,
such as RADIUS (RFC 2138, 2139)[1],[2] and follow-on efforts like AAA
Working Group, and the Diameter protocol [3].  These are protocols for
carrying user service information for authentication, authorization,
accounting, and auditing, between a Network Access Server which desires
to authenticate its incoming calls and a shared authentication server.

A URL for this Internet-Draft is:
http://www.ietf.org/internet-drafts/draft-ietf-nasreq-nasmodel-01.txt

Internet-Drafts are also available by anonymous FTP. Login with the username
"anonymous" and a password of your e-mail address. After logging in,
type "cd internet-drafts" and then
	"get draft-ietf-nasreq-nasmodel-01.txt".

A list of Internet-Drafts directories can be found in
http://www.ietf.org/shadow.html 
or ftp://ftp.ietf.org/ietf/1shadow-sites.txt


Internet-Drafts can also be obtained by e-mail.

Send a message to:
	mailserv@ietf.org.
In the body type:
	"FILE /internet-drafts/draft-ietf-nasreq-nasmodel-01.txt".
	
NOTE:	The mail server at ietf.org can return the document in
	MIME-encoded form by using the "mpack" utility.  To use this
	feature, insert the command "ENCODING mime" before the "FILE"
	command.  To decode the response(s), you will need "munpack" or
	a MIME-compliant mail reader.  Different MIME-compliant mail readers
	exhibit different behavior, especially when dealing with
	"multipart" MIME messages (i.e. documents which have been split
	up into multiple messages), so check your local documentation on
	how to manipulate these messages.
		
		
Below is the data which will enable a MIME compliant mail reader
implementation to automatically retrieve the ASCII version of the
Internet-Draft.

--NextPart
Content-Type: Multipart/Alternative; Boundary="OtherAccess"

--OtherAccess
Content-Type: Message/External-body;
	access-type="mail-server";
	server="mailserv@ietf.org"

Content-Type: text/plain
Content-ID:	<19991008135936.I-D@ietf.org>

ENCODING mime
FILE /internet-drafts/draft-ietf-nasreq-nasmodel-01.txt

--OtherAccess
Content-Type: Message/External-body;
	name="draft-ietf-nasreq-nasmodel-01.txt";
	site="ftp.ietf.org";
	access-type="anon-ftp";
	directory="internet-drafts"

Content-Type: text/plain
Content-ID:	<19991008135936.I-D@ietf.org>

--OtherAccess--

--NextPart--




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Mon Oct 11 17:26:40 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA16011;
	Mon, 11 Oct 1999 17:26:39 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id RAA04192
	for nasreq-outgoing; Mon, 11 Oct 1999 17:11:41 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id RAA04188
	for <nasreq@tdmx.rutgers.edu>; Mon, 11 Oct 1999 17:11:39 -0400 (EDT)
Received: from engmail3.Eng.Sun.COM ([129.144.170.5])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id OAA00674
	for <nasreq@tdmx.rutgers.edu>; Mon, 11 Oct 1999 14:11:38 -0700 (PDT)
Received: from hsmpka.eng.sun.com (hsmpka.Eng.Sun.COM [129.146.93.47])
	by engmail3.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id OAA26026
	for <nasreq@tdmx.rutgers.edu>; Mon, 11 Oct 1999 14:11:02 -0700 (PDT)
Received: from hsmpka.Eng.Sun.COM by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id OAA00394; Mon, 11 Oct 1999 14:10:45 -0700
From: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Message-Id: <199910112110.OAA00394@hsmpka.eng.sun.com>
Date: Mon, 11 Oct 1999 14:10:15 -0700
To: <nasreq@tdmx.rutgers.edu>
Reply-To: <Pat.Calhoun@eng.sun.com>
Subject: Question on draft-ietf-nasreq-criteria-02.txt
X-Mailer: Sun NetMail 2.3
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

I have a question about a potential requirement that does not show up in
the current -02 spec. The requirement would be that this next generation
protocol SHOULD allow for combination of authentication and authorization
in a single request. This is especially important since it is the way that
RADIUS currently works, but some people within the AAA WG may not be aware
of this. 

It may also be a requirement that authentication and authorization be
sent separately. This may be more important once a session has already
been established, where a user has to be re-authentication and/or
re-authorized.

PatC



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Mon Oct 11 21:13:00 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA18899;
	Mon, 11 Oct 1999 21:12:59 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id UAA05117
	for nasreq-outgoing; Mon, 11 Oct 1999 20:53:42 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mailman.cisco.com (mailman.cisco.com [171.68.225.9])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id UAA05113
	for <nasreq@tdmx.rutgers.edu>; Mon, 11 Oct 1999 20:53:41 -0400 (EDT)
Received: from chsharp-tecra (chsharp-isdn.cisco.com [171.68.116.221]) by mailman.cisco.com (8.8.8+Sun/CISCO.SERVER.1.2) with ESMTP id RAA25717; Mon, 11 Oct 1999 17:53:08 -0700 (PDT)
Message-Id: <4.2.0.58.19991011205203.00b72c80@dogwood.cisco.com>
X-Sender: chsharp@dogwood.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 
Date: Mon, 11 Oct 1999 20:52:24 -0400
To: <Pat.Calhoun@eng.sun.com>, <nasreq@tdmx.rutgers.edu>
From: Chip Sharp <chsharp@cisco.com>
Subject: Re: Question on draft-ietf-nasreq-criteria-02.txt
In-Reply-To: <199910112110.OAA00394@hsmpka.eng.sun.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

At 02:10 PM 10/11/99 -0700, Patrice Calhoun wrote:
>I have a question about a potential requirement that does not show up in
>the current -02 spec. The requirement would be that this next generation
>protocol SHOULD allow for combination of authentication and authorization
>in a single request. This is especially important since it is the way that
>RADIUS currently works, but some people within the AAA WG may not be aware
>of this.
>
>It may also be a requirement that authentication and authorization be
>sent separately. This may be more important once a session has already
>been established, where a user has to be re-authentication and/or
>re-authorized.

Or authorized for a newly requested service.


>PatC

Support NetAid!  http://www.netaid.org
--------------------------------------------------
Chip Sharp                 Consulting Engineering
Cisco Systems              Telco Bio-region
Reality - Love it or Leave it.			
--------------------------------------------------


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Mon Oct 11 21:58:19 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA20325;
	Mon, 11 Oct 1999 21:58:18 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id VAA05281
	for nasreq-outgoing; Mon, 11 Oct 1999 21:39:09 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from dfssl.exchange.microsoft.com (dfssl.exchange.microsoft.com [131.107.88.59])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id VAA05277
	for <nasreq@tdmx.rutgers.edu>; Mon, 11 Oct 1999 21:39:08 -0400 (EDT)
Received: by dfssl with Internet Mail Service (5.5.2650.21)
	id <4RCW445C>; Mon, 11 Oct 1999 18:38:28 -0700
Message-ID: <078292D50C98D2118D090008C7E9C6A602258147@STAY.platinum.corp.microsoft.com>
From: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
To: "'Chip Sharp'" <chsharp@cisco.com>, Pat.Calhoun@eng.sun.com,
        nasreq@tdmx.rutgers.edu
Subject: RE: Question on draft-ietf-nasreq-criteria-02.txt
Date: Mon, 11 Oct 1999 18:38:37 -0700
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

>>It may also be a requirement that authentication and authorization be
>>sent separately. This may be more important once a session has already
>>been established, where a user has to be re-authentication and/or
>>re-authorized.

>Or authorized for a newly requested service.

I've always wondered whether this was really a AAA 
requirement, or more of a requirement of the underlying 
connection protocol. For example, for PPP one
could create an EAP-Kerberos method to achieve this goal without any
changes to existing AAA protocols, so that meeting the
requirement is trivial from the AAA point of view. However,
the issues come with respect to the access protocol. 

The basic issue here is that in
a network access situation, the client does not have full Internet
connectivity and so cannot reach the KDC directly. This may
require the device to obtain tickets on the user's behalf, so
that delegation support is required. Also, if the access protocol 
involved has limited round trips it can be tricky
to wedge all the packets in. 




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Mon Oct 11 23:51:04 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id XAA22702;
	Mon, 11 Oct 1999 23:51:04 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id XAA05728
	for nasreq-outgoing; Mon, 11 Oct 1999 23:30:05 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from smtpgw02.wanmail.net (smtpgw02.wanmail.net [209.154.194.61])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id XAA05724
	for <nasreq@tdmx.rutgers.edu>; Mon, 11 Oct 1999 23:30:03 -0400 (EDT)
Received: from wcomexch1.wcom.net (wcomexch1.wcom.net [209.154.236.12])
	by smtpgw02.wanmail.net (8.9.1/8.9.1) with ESMTP id XAA10580;
	Mon, 11 Oct 1999 23:29:00 -0400 (EDT)
Received: by wcomexch1.wcom.net with Internet Mail Service (5.5.2448.0)
	id <SAX9H6V6>; Mon, 11 Oct 1999 23:27:57 -0400
Message-ID: <9A74C89D6696D2118E500008C7BA7C560C03B195@wcomexch1.wcom.net>
From: "Beadles, Mark A." <MBeadles@wcom.net>
To: "'Bernard Aboba (Exchange)'" <bernarda@Exchange.Microsoft.com>,
        "'Matt Holdrege'" <matt@ascend.com>, Pat.Calhoun@eng.sun.com
Cc: nasreq@tdmx.rutgers.edu
Subject: RE: Comments on draft-ietf-nasreq-criteria-02.txt
Date: Mon, 11 Oct 1999 23:27:52 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="windows-1252"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

I agree that we need to operationally support backwards compatibility.  I
think that (3) below is too stringent.  For example, the new AAA protocol
may have a radically different Accounting transport, making (3) difficult.

I will add some kind of text to Criteria talking about this.

+ Mark Anthony Beadles +  UUNET Product Development  + 
+  mbeadles@wcom.net   + http://networking.us.uu.net +
+  Voice 614.723.1941  +       Fax 614.723.8407      +



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Mon Oct 11 23:53:19 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id XAA22714;
	Mon, 11 Oct 1999 23:53:19 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id XAA05751
	for nasreq-outgoing; Mon, 11 Oct 1999 23:33:56 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from smtpgw02.wanmail.net (smtpgw02.wanmail.net [209.154.194.61])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id XAA05747
	for <nasreq@tdmx.rutgers.edu>; Mon, 11 Oct 1999 23:33:55 -0400 (EDT)
Received: from wcomexch1.wcom.net (wcomexch1.wcom.net [209.154.236.12])
	by smtpgw02.wanmail.net (8.9.1/8.9.1) with ESMTP id XAA12900;
	Mon, 11 Oct 1999 23:32:52 -0400 (EDT)
Received: by wcomexch1.wcom.net with Internet Mail Service (5.5.2448.0)
	id <SAX9H6XG>; Mon, 11 Oct 1999 23:31:49 -0400
Message-ID: <9A74C89D6696D2118E500008C7BA7C560C03B196@wcomexch1.wcom.net>
From: "Beadles, Mark A." <MBeadles@wcom.net>
To: "'Bernard Aboba (Exchange)'" <bernarda@Exchange.Microsoft.com>,
        "'Matt Holdrege'" <matt@ascend.com>, Pat.Calhoun@eng.sun.com
Cc: nasreq@tdmx.rutgers.edu
Subject: RE: Comments on draft-ietf-nasreq-criteria-02.txt
Date: Mon, 11 Oct 1999 23:31:40 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="windows-1252"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

It is operationally very important to support the many thousands of devices
making up the Internet today, which speak
RADIUS.	Therefore, devices which implement the AAA protocol SHOULD support
some means of compatibility with devices
that implement RADIUS.  This requirement MAY be met by simply listening on
the same port as RADIUS;  note that in this
case the AAA and RADIUS protocols might be quite different.  Alternatively,
the requirement MAY be met by creating a
gateway between RADIUS and the AAA protocol.  Other alternatives are
possible; the intent is to ensure the continued
operation of the currently-deployed RADIUS infrastructure during deployment
of the AAA protocol.

+ Mark Anthony Beadles +  UUNET Product Development  + 
+  mbeadles@wcom.net   + http://networking.us.uu.net +
+  Voice 614.723.1941  +       Fax 614.723.8407      +


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Tue Oct 12 00:07:40 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id AAA22810;
	Tue, 12 Oct 1999 00:07:39 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id XAA05874
	for nasreq-outgoing; Mon, 11 Oct 1999 23:47:42 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from smtpgw02.wanmail.net (smtpgw02.wanmail.net [209.154.194.61])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id XAA05870
	for <nasreq@tdmx.rutgers.edu>; Mon, 11 Oct 1999 23:47:40 -0400 (EDT)
Received: from wcomexch1.wcom.net (wcomexch1.wcom.net [209.154.236.12])
	by smtpgw02.wanmail.net (8.9.1/8.9.1) with ESMTP id XAA21121
	for <nasreq@tdmx.rutgers.edu>; Mon, 11 Oct 1999 23:47:10 -0400 (EDT)
Received: by wcomexch1.wcom.net with Internet Mail Service (5.5.2448.0)
	id <SAX9H66A>; Mon, 11 Oct 1999 23:46:07 -0400
Message-ID: <9A74C89D6696D2118E500008C7BA7C560C03B198@wcomexch1.wcom.net>
From: "Beadles, Mark A." <MBeadles@wcom.net>
To: "Nasreqng List (E-mail)" <nasreq@tdmx.rutgers.edu>
Subject: nasreq Criteria-03
Date: Mon, 11 Oct 1999 23:46:05 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

I have just posted criteria-03.  There are quite a few minor changes (mostly
clarifications), but a couple major changes:

1 - I have added an explicit requirement for RADIUS compatibility.

2 - I have removed the requirement to explicitly support TCP AND UDP.
Instead, I have focused on transport characteristics, and separated
accounting transport from auth/auth transport.



+ Mark Anthony Beadles +  UUNET Product Development  + 
+  mbeadles@wcom.net   + http://networking.us.uu.net +
+  Voice 614.723.1941  +       Fax 614.723.8407      +




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Tue Oct 12 07:09:37 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA08784;
	Tue, 12 Oct 1999 07:09:36 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id GAA07527
	for nasreq-outgoing; Tue, 12 Oct 1999 06:57:05 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from ietf.org (odin.ietf.org [132.151.1.176])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id GAA07523
	for <nasreq@tdmx.rutgers.edu>; Tue, 12 Oct 1999 06:57:01 -0400 (EDT)
Received: from CNRI.Reston.VA.US (localhost [127.0.0.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id GAA08146;
	Tue, 12 Oct 1999 06:56:25 -0400 (EDT)
Message-Id: <199910121056.GAA08146@ietf.org>
Mime-Version: 1.0
Content-Type: Multipart/Mixed; Boundary="NextPart"
To: IETF-Announce: ;
Cc: nasreq@tdmx.rutgers.edu
From: Internet-Drafts@ietf.org
Reply-to: Internet-Drafts@ietf.org
Subject: I-D ACTION:draft-ietf-nasreq-ext-radiuspract-01.txt
Date: Tue, 12 Oct 1999 06:56:25 -0400
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

--NextPart

A New Internet-Draft is available from the on-line Internet-Drafts directories.
This draft is a work item of the Network Access Server Requirements Working Group of the IETF.

	Title		: Network Access Servers Requirements: Extended RADIUS 
                          Practices
	Author(s)	: D. Mitton
	Filename	: draft-ietf-nasreq-ext-radiuspract-01.txt
	Pages		: 16
	Date		: 11-Oct-99
	
This document describes current practices implemented in NAS products
that go beyond the scope of the RADIUS RFCs (2138, 2139). The purpose of
this effort is to give examples that show the need for addressing and
standardizing these types of ad-hoc functions.  Since many of these
features require a matching server support component, the ability to
deploy and manage interoperable NAS and AAA server products is severely
hindered.
These practices are documented here to show functions that are obviously
desired in developing future AAA protocols for NAS deployment.
This document is a draft submission of the Network Access Server
Requirements Next Generation (NASREQNG) Working Group of the Internet
Engineering Task Force (IETF).  Comments should be submitted to the
mailing list nasreq@tdmx.rutgers.edu.

A URL for this Internet-Draft is:
http://www.ietf.org/internet-drafts/draft-ietf-nasreq-ext-radiuspract-01.txt

Internet-Drafts are also available by anonymous FTP. Login with the username
"anonymous" and a password of your e-mail address. After logging in,
type "cd internet-drafts" and then
	"get draft-ietf-nasreq-ext-radiuspract-01.txt".

A list of Internet-Drafts directories can be found in
http://www.ietf.org/shadow.html 
or ftp://ftp.ietf.org/ietf/1shadow-sites.txt


Internet-Drafts can also be obtained by e-mail.

Send a message to:
	mailserv@ietf.org.
In the body type:
	"FILE /internet-drafts/draft-ietf-nasreq-ext-radiuspract-01.txt".
	
NOTE:	The mail server at ietf.org can return the document in
	MIME-encoded form by using the "mpack" utility.  To use this
	feature, insert the command "ENCODING mime" before the "FILE"
	command.  To decode the response(s), you will need "munpack" or
	a MIME-compliant mail reader.  Different MIME-compliant mail readers
	exhibit different behavior, especially when dealing with
	"multipart" MIME messages (i.e. documents which have been split
	up into multiple messages), so check your local documentation on
	how to manipulate these messages.
		
		
Below is the data which will enable a MIME compliant mail reader
implementation to automatically retrieve the ASCII version of the
Internet-Draft.

--NextPart
Content-Type: Multipart/Alternative; Boundary="OtherAccess"

--OtherAccess
Content-Type: Message/External-body;
	access-type="mail-server";
	server="mailserv@ietf.org"

Content-Type: text/plain
Content-ID:	<19991011110744.I-D@ietf.org>

ENCODING mime
FILE /internet-drafts/draft-ietf-nasreq-ext-radiuspract-01.txt

--OtherAccess
Content-Type: Message/External-body;
	name="draft-ietf-nasreq-ext-radiuspract-01.txt";
	site="ftp.ietf.org";
	access-type="anon-ftp";
	directory="internet-drafts"

Content-Type: text/plain
Content-ID:	<19991011110744.I-D@ietf.org>

--OtherAccess--

--NextPart--




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Tue Oct 12 10:04:04 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA15554;
	Tue, 12 Oct 1999 10:04:03 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id JAA08222
	for nasreq-outgoing; Tue, 12 Oct 1999 09:48:29 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from segue.merit.edu (segue.merit.edu [198.108.1.41])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id JAA08206
	for <nasreq@tdmx.rutgers.edu>; Tue, 12 Oct 1999 09:48:27 -0400 (EDT)
Received: from merit.edu (stevie.merit.edu [198.108.62.185])
	by segue.merit.edu (Postfix) with ESMTP
	id 21BC85DDD3; Tue, 12 Oct 1999 09:48:23 -0400 (EDT)
Message-ID: <38033B99.B024A02@merit.edu>
Date: Tue, 12 Oct 1999 09:46:01 -0400
From: John Vollbrecht <jrv@merit.edu>
Reply-To: jrv@merit.edu
Organization: Merit Network Inc
X-Mailer: Mozilla 4.6 [en] (WinNT; U)
X-Accept-Language: en
MIME-Version: 1.0
To: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
Cc: "'Chip Sharp'" <chsharp@cisco.com>, Pat.Calhoun@eng.sun.com,
        nasreq@tdmx.rutgers.edu
Subject: Re: Question on draft-ietf-nasreq-criteria-02.txt
References: <078292D50C98D2118D090008C7E9C6A602258147@STAY.platinum.corp.microsoft.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit



"Bernard Aboba (Exchange)" wrote:

> >>It may also be a requirement that authentication and authorization be
> >>sent separately. This may be more important once a session has already
> >>been established, where a user has to be re-authentication and/or
> >>re-authorized.
>
> >Or authorized for a newly requested service.
>
> I've always wondered whether this was really a AAA
> requirement, or more of a requirement of the underlying
> connection protocol. For example, for PPP one
> could create an EAP-Kerberos method to achieve this goal without any
> changes to existing AAA protocols, so that meeting the
> requirement is trivial from the AAA point of view. However,
> the issues come with respect to the access protocol.
>
> The basic issue here is that in
> a network access situation, the client does not have full Internet
> connectivity and so cannot reach the KDC directly. This may
> require the device to obtain tickets on the user's behalf, so
> that delegation support is required. Also, if the access protocol
> involved has limited round trips it can be tricky
> to wedge all the packets in.

One issue with EAP Kerberos is that it is not inter-domain.  I am not sure
how the EAP over RADIUS would work in an interdomain situation, but it is
certainly a requirement that if it is part of the AAA protocol that it be
able to work across domains.

A second point is that reauthentication and reauthorization are done for
different reasons.  Reauthentication is done to verify that this is still
the same user.  Reauthorization is done either because new resources are to
needed to support the existing service (e.g. adding another ISDN channel to
a connection) or to verify that the current authorization is still valid.

A final comment - I think the concept of delegation is very interesting
here.  The idea of delegating the authentication to the NAS by the user.  In
other contexts, an authorization server may delegate some or all
authorization to other authorization servers.  I am not sure how to
formalize this, but it might be worth a try.




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Tue Oct 12 13:20:58 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA22131;
	Tue, 12 Oct 1999 13:20:58 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id MAA09028
	for nasreq-outgoing; Tue, 12 Oct 1999 12:56:22 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from dfssl.exchange.microsoft.com (dfssl.exchange.microsoft.com [131.107.88.59])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id MAA09024
	for <nasreq@tdmx.rutgers.edu>; Tue, 12 Oct 1999 12:56:17 -0400 (EDT)
Received: by dfssl with Internet Mail Service (5.5.2650.21)
	id <4RCWVA07>; Tue, 12 Oct 1999 09:55:37 -0700
Message-ID: <078292D50C98D2118D090008C7E9C6A60598437C@STAY.platinum.corp.microsoft.com>
From: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
To: jrv@merit.edu
Cc: "'Chip Sharp'" <chsharp@cisco.com>, Pat.Calhoun@eng.sun.com,
        nasreq@tdmx.rutgers.edu
Subject: RE: Question on draft-ietf-nasreq-criteria-02.txt
Date: Tue, 12 Oct 1999 09:55:45 -0700
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

>One issue with EAP Kerberos is that it is not inter-domain.  I am not sure
>how the EAP over RADIUS would work in an interdomain situation, but it is
>certainly a requirement that if it is part of the AAA protocol that it be
>able to work across domains.

I'm curious as to why you think that EAP-KRB can't work inter-realm. From
what I can see, the main obstacles relate to topology and the limitations of
PPP authentication (and Kerberos). At first glance, EAP over RADIUS doesn't
seem like an obstacle. 

In dialup we need to assume that the client isn't connected to the Internet
until after IPCP completes. As a result, it isn't possible for the client to
have general Internet connectivity during the authentication. This means
that the client can talk to the AAA server via EAP but doesn't have the
general ability to reach any node on the Internet. 

This means that the client can't get to its home KDC directly in order to
obtain a TGT, and then ask for a ticket to the local realm's KDC. The way
around this is for the AAA server to forward packets to home KDCs on the
user's behalf. 

For example, the client can send the AAA server an AS_REQ and the AAA server
can forward this to the home KDC on the client's behalf and get the AS_REP
back and send this to the client. As long as the client is corresponding
with KDCs that the AAA server is willing to forward to then it would seem
like this could work. Note that certificate-based EAP protocols like EAP-TLS
don't encounter this issue since the local AAA server can validate the cert.


I would note that the above issues only occur during the first
authentication. Subsequent to that the client can use the PPP service ticket
that it obtained in the first authentication, and re-authenticate in an
expedited manner. 

So from what I can see there isn't much that a AAA protocol can do to make
this easier. 

>A second point is that reauthentication and reauthorization are done for
>different reasons.  Reauthentication is done to verify that this is still
>the same user.  Reauthorization is done either because new resources 
>are to needed to support the existing service (e.g. adding another ISDN 
>channel to a connection) or to verify that the current authorization is >
>still valid.

Sure. While the authentication might be expedited due to the client's
posession of a reusable ticket, this doesn't affect the authorization
process. 

>A final comment - I think the concept of delegation is very interesting
>here.  The idea of delegating the authentication to the NAS by the user.  

Actually, the forwarding described above would be from the user to the AAA
server (KRB Forwardable flag set). Since few NASes understand Kerberos (or
want to) I don't think that the NAS will generally have the smarts to handle
this. This creates a bit of an issue since the session key will be held by
the AAA server and the client, not by the client and the NAS. So if you want
to do something with the session key (like encrypt the PPP session) you've
got an issue. 

>In other contexts, an authorization server may delegate some or all
>authorization to other authorization servers.  I am not sure how to
>formalize this, but it might be worth a try.

Well, in Kerberos it's possible for the KDC to send back authorization in
the PACS field. I'd have to think through the sequence of events but
potentially the local KDC could take the PACS field into account in issuing
its own tickets to the user. 


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 13 07:14:34 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA17853;
	Wed, 13 Oct 1999 07:14:34 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id GAA13835
	for nasreq-outgoing; Wed, 13 Oct 1999 06:54:09 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from ietf.org (odin.ietf.org [132.151.1.176])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id GAA13831
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 06:54:08 -0400 (EDT)
Received: from CNRI.Reston.VA.US (localhost [127.0.0.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id GAA17189;
	Wed, 13 Oct 1999 06:53:32 -0400 (EDT)
Message-Id: <199910131053.GAA17189@ietf.org>
Mime-Version: 1.0
Content-Type: Multipart/Mixed; Boundary="NextPart"
To: IETF-Announce: ;
Cc: nasreq@tdmx.rutgers.edu
From: Internet-Drafts@ietf.org
Reply-to: Internet-Drafts@ietf.org
Subject: I-D ACTION:draft-ietf-nasreq-criteria-03.txt
Date: Wed, 13 Oct 1999 06:53:32 -0400
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

--NextPart

A New Internet-Draft is available from the on-line Internet-Drafts directories.
This draft is a work item of the Network Access Server Requirements Working Group of the IETF.

	Title		: Criteria for Evaluating Network Access Server  
                          Protocols
	Author(s)	: M. Beadles
	Filename	: draft-ietf-nasreq-criteria-03.txt
	Pages		: 17
	Date		: 12-Oct-99
	
This document defines  requirements  for  protocols  used  by  Network
Access  Servers  (NAS).   Protocols  used by NAS's may be divided into
four spaces:  Access protocols, Network protocols, AAA protocols,  and
Management  protocols.  Primary attention is given to setting require-
ments for AAA protocols, since that space is currently the least  well
defined.

A URL for this Internet-Draft is:
http://www.ietf.org/internet-drafts/draft-ietf-nasreq-criteria-03.txt

Internet-Drafts are also available by anonymous FTP. Login with the username
"anonymous" and a password of your e-mail address. After logging in,
type "cd internet-drafts" and then
	"get draft-ietf-nasreq-criteria-03.txt".

A list of Internet-Drafts directories can be found in
http://www.ietf.org/shadow.html 
or ftp://ftp.ietf.org/ietf/1shadow-sites.txt


Internet-Drafts can also be obtained by e-mail.

Send a message to:
	mailserv@ietf.org.
In the body type:
	"FILE /internet-drafts/draft-ietf-nasreq-criteria-03.txt".
	
NOTE:	The mail server at ietf.org can return the document in
	MIME-encoded form by using the "mpack" utility.  To use this
	feature, insert the command "ENCODING mime" before the "FILE"
	command.  To decode the response(s), you will need "munpack" or
	a MIME-compliant mail reader.  Different MIME-compliant mail readers
	exhibit different behavior, especially when dealing with
	"multipart" MIME messages (i.e. documents which have been split
	up into multiple messages), so check your local documentation on
	how to manipulate these messages.
		
		
Below is the data which will enable a MIME compliant mail reader
implementation to automatically retrieve the ASCII version of the
Internet-Draft.

--NextPart
Content-Type: Multipart/Alternative; Boundary="OtherAccess"

--OtherAccess
Content-Type: Message/External-body;
	access-type="mail-server";
	server="mailserv@ietf.org"

Content-Type: text/plain
Content-ID:	<19991012102311.I-D@ietf.org>

ENCODING mime
FILE /internet-drafts/draft-ietf-nasreq-criteria-03.txt

--OtherAccess
Content-Type: Message/External-body;
	name="draft-ietf-nasreq-criteria-03.txt";
	site="ftp.ietf.org";
	access-type="anon-ftp";
	directory="internet-drafts"

Content-Type: text/plain
Content-ID:	<19991012102311.I-D@ietf.org>

--OtherAccess--

--NextPart--




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 13 10:50:47 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA26298;
	Wed, 13 Oct 1999 10:50:47 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id KAA14710
	for nasreq-outgoing; Wed, 13 Oct 1999 10:29:57 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from smtpgw02.wanmail.net (smtpgw02.wanmail.net [209.154.194.61])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id KAA14706
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 10:29:55 -0400 (EDT)
Received: from wcomexch1.wcom.net (wcomexch1.wcom.net [209.154.236.12])
	by smtpgw02.wanmail.net (8.9.1/8.9.1) with ESMTP id KAA15495
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 10:29:25 -0400 (EDT)
Received: by wcomexch1.wcom.net with Internet Mail Service (5.5.2448.0)
	id <SAX925Z6>; Wed, 13 Oct 1999 10:28:21 -0400
Message-ID: <9A74C89D6696D2118E500008C7BA7C560C03B1B4@wcomexch1.wcom.net>
From: "Beadles, Mark A." <MBeadles@wcom.net>
To: "Nasreqng List (E-mail)" <nasreq@tdmx.rutgers.edu>
Subject: nasreq Call for Agenda
Date: Wed, 13 Oct 1999 10:28:20 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

Please submit agenda items for the November IETF.  Agenda cutoff is 3
November, so please submit agenda items no later than 1 November.

+ Mark Anthony Beadles + Co-chair, IETF NASREQ WG   + 
+  mbeadles@wcom.net   + http://nasreq.rutgers.edu/ +
+  Voice 614.723.1941  +       Fax 614.723.8407     +






From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 13 11:48:49 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA27782;
	Wed, 13 Oct 1999 11:48:47 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id LAA15067
	for nasreq-outgoing; Wed, 13 Oct 1999 11:22:53 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id LAA15063
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 11:22:52 -0400 (EDT)
Received: from engmail4.Eng.Sun.COM ([129.144.134.6])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id IAA17901
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 08:22:50 -0700 (PDT)
Received: from hsmpka.eng.sun.com (phys-hsmpka.Eng.Sun.COM [129.146.63.37])
	by engmail4.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id IAA15756
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 08:22:14 -0700 (PDT)
Received: from hsmpka.Eng.Sun.COM by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id IAA01717; Wed, 13 Oct 1999 08:22:06 -0700
From: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Message-Id: <199910131522.IAA01717@hsmpka.eng.sun.com>
Date: Wed, 13 Oct 1999 08:21:09 -0700
To: <nasreq@tdmx.rutgers.edu>
Reply-To: <Pat.Calhoun@eng.sun.com>
Subject: comments on draft-ietf-nasreq-criteria-03.txt
X-Mailer: Sun NetMail 2.3
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

I just took a look at the latest document, and I like it. Thanks to Mark for
the enormous effort it must have taken to get the document in shape for
the I-D deadline.

I do, however, have come up two more requirements. I am not sure that this
requirement really needs to be incorporated into a new revision before the
deadline, perhaps it can wait. In RFC 2477 [ROAMOPS], the following is stated:

    -  capabilities adjustment.  Each manufacturer's equipment has
       different capabilities.  Thus, the AAAL may have to modify the
       authorization information from the AAAH.

This is a very important requirement for two reasons:

1. In the RADIUS world, there is a proliferation of vendor specific attributes.
When a client talks to a server (or a server to another server), it has NO
idea what the manufacturer of the equipment is, and MAY attempt to send vendor
specific attributes. If both peers could know a priori what each other supports,
it would make operations that much simpler.

2. In the NASes that I am aware of, when one configures the RADIUS servers, 
one must specificy which ones is the accounting server, and which one isn't.
As new services become available, the NAS must then have additional 
configuration so it knows which server to ask for specific services. If the
NAS could get an idea of which servers should be used for what purpose, this
would reduce the overall configuration of the NAS, and again make the operations
simpler.

It would be nice if NASes could use a service discovery scheme to locate
AAA servers. Let's use SLP as an example. When a NAS comes up, it could use
SLP to locate the servers in the network. Imagine a network where one simply
installs a new server, and NO configuration is necessary on the NAS. It simply
locates the new server, and makes use of it. 

Comments?

PatC



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 13 15:42:36 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA02879;
	Wed, 13 Oct 1999 15:42:34 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id PAA16571
	for nasreq-outgoing; Wed, 13 Oct 1999 15:25:31 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from dfssl.exchange.microsoft.com (dfssl.exchange.microsoft.com [131.107.88.59])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id PAA16567
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 15:25:29 -0400 (EDT)
Received: by dfssl with Internet Mail Service (5.5.2650.21)
	id <4RCWWK5M>; Wed, 13 Oct 1999 12:23:59 -0700
Message-ID: <078292D50C98D2118D090008C7E9C6A602258153@STAY.platinum.corp.microsoft.com>
From: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
To: "'Pat.Calhoun@eng.sun.com'" <Pat.Calhoun@eng.sun.com>,
        nasreq@tdmx.rutgers.edu
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
Date: Wed, 13 Oct 1999 12:23:57 -0700
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="windows-1252"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

I agree with Pat.

One of the major reasons for insertion of RADIUS proxies
is capabilities adjustment. Given the different dialects
that are spoken nowadays, getting the editing right is
already quite tedious and might be likely to become
unbearable in the future if we don't do something about
it. 

There are several possible solutions:

a. Capabilities negotiation. This is the approach taken
in BGP. This approach seems to work well when there are
lots of extensions sets, thus making it likely that each side
might not be able to communicate well unless they reach
an agreement beforehand. 

b. Use a Mandatory bit as in L2TP. In this approach, typically 
only standardized attributes should be marked Mandatory, so as to
avoid potential "bullying." However, if there are sets of
standardized attributes that might not be implemented by a
client or server (as might be the case when there is a base
protocol and extension sets), then this might not work so well.  


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 13 15:57:06 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA03083;
	Wed, 13 Oct 1999 15:57:06 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id PAA16676
	for nasreq-outgoing; Wed, 13 Oct 1999 15:36:17 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id PAA16668
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 15:36:15 -0400 (EDT)
Received: from engmail4.Eng.Sun.COM ([129.144.134.6])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id MAA26731;
	Wed, 13 Oct 1999 12:36:14 -0700 (PDT)
Received: from hsmpka.eng.sun.com (hsmpka.Eng.Sun.COM [129.146.101.47])
	by engmail4.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id MAA06107;
	Wed, 13 Oct 1999 12:35:01 -0700 (PDT)
Received: from hsmpka.Eng.Sun.COM by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id MAA27148; Wed, 13 Oct 1999 12:34:09 -0700
From: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Message-Id: <199910131934.MAA27148@hsmpka.eng.sun.com>
Date: Wed, 13 Oct 1999 12:33:45 -0700
To: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>,
        <nasreq@tdmx.rutgers.edu>,
        "'Pat.Calhoun@eng.sun.com'" <Pat.Calhoun@eng.sun.com>
Reply-To: <Pat.Calhoun@eng.sun.com>
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
X-Mailer: Sun NetMail 2.3
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Actually, I do not see the two solutions provided as being mutually
exclusive. I think that both can be used simultaneously. The former
allows for one to identify standards-track extensions, and the later
to identify vendor-specific attributes.

PatC
>I agree with Pat.
>
>One of the major reasons for insertion of RADIUS proxies
>is capabilities adjustment. Given the different dialects
>that are spoken nowadays, getting the editing right is
>already quite tedious and might be likely to become
>unbearable in the future if we don't do something about
>it. 
>
>There are several possible solutions:
>
>a. Capabilities negotiation. This is the approach taken
>in BGP. This approach seems to work well when there are
>lots of extensions sets, thus making it likely that each side
>might not be able to communicate well unless they reach
>an agreement beforehand. 
>
>b. Use a Mandatory bit as in L2TP. In this approach, typically 
>only standardized attributes should be marked Mandatory, so as to
>avoid potential "bullying." However, if there are sets of
>standardized attributes that might not be implemented by a
>client or server (as might be the case when there is a base
>protocol and extension sets), then this might not work so well.  




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 13 16:29:28 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA03603;
	Wed, 13 Oct 1999 16:29:28 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id QAA16981
	for nasreq-outgoing; Wed, 13 Oct 1999 16:10:54 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from dfssl.exchange.microsoft.com (dfssl.exchange.microsoft.com [131.107.88.59])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id QAA16977
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 16:10:49 -0400 (EDT)
Received: by dfssl with Internet Mail Service (5.5.2650.21)
	id <4RCWWM4W>; Wed, 13 Oct 1999 13:09:41 -0700
Message-ID: <078292D50C98D2118D090008C7E9C6A602258155@STAY.platinum.corp.microsoft.com>
From: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
To: "'Pat.Calhoun@eng.sun.com'" <Pat.Calhoun@eng.sun.com>,
        "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>,
        nasreq@tdmx.rutgers.edu
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
Date: Wed, 13 Oct 1999 13:09:39 -0700
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

If you knew that the other side could handle a capability (via
capabilities negotiation) why would you *also* need to mark
an attribute as Mandatory? Does this have something to do
with proxy editing?

-----Original Message-----
From: Pat.Calhoun@eng.sun.com [mailto:Pat.Calhoun@eng.sun.com]
Sent: Wednesday, October 13, 1999 12:34 PM
To: Bernard Aboba (Exchange); nasreq@tdmx.rutgers.edu;
'Pat.Calhoun@eng.sun.com'
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt


Actually, I do not see the two solutions provided as being mutually
exclusive. I think that both can be used simultaneously. The former
allows for one to identify standards-track extensions, and the later
to identify vendor-specific attributes.

PatC
>I agree with Pat.
>
>One of the major reasons for insertion of RADIUS proxies
>is capabilities adjustment. Given the different dialects
>that are spoken nowadays, getting the editing right is
>already quite tedious and might be likely to become
>unbearable in the future if we don't do something about
>it. 
>
>There are several possible solutions:
>
>a. Capabilities negotiation. This is the approach taken
>in BGP. This approach seems to work well when there are
>lots of extensions sets, thus making it likely that each side
>might not be able to communicate well unless they reach
>an agreement beforehand. 
>
>b. Use a Mandatory bit as in L2TP. In this approach, typically 
>only standardized attributes should be marked Mandatory, so as to
>avoid potential "bullying." However, if there are sets of
>standardized attributes that might not be implemented by a
>client or server (as might be the case when there is a base
>protocol and extension sets), then this might not work so well.  



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 13 16:30:44 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA03627;
	Wed, 13 Oct 1999 16:30:43 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id QAA17012
	for nasreq-outgoing; Wed, 13 Oct 1999 16:15:12 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from rip.psg.com (rip.psg.com [147.28.0.39])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id QAA17008
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 16:15:10 -0400 (EDT)
Received: from randy by rip.psg.com with local (Exim 3.03 #1)
	id 11bUnh-000KMO-00; Wed, 13 Oct 1999 13:15:05 -0700
From: Randy Bush <randy@psg.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Cc: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>,
        <nasreq@tdmx.rutgers.edu>
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
References: <199910131934.MAA27148@hsmpka.eng.sun.com>
Message-Id: <E11bUnh-000KMO-00@rip.psg.com>
Date: Wed, 13 Oct 1999 13:15:05 -0700
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

> Actually, I do not see the two solutions provided as being mutually
> exclusive. I think that both can be used simultaneously. The former
> allows for one to identify standards-track extensions, and the later
> to identify vendor-specific attributes.
> 
> PatC
>> I agree with Pat.
>> 
>> One of the major reasons for insertion of RADIUS proxies
>> is capabilities adjustment. Given the different dialects
>> that are spoken nowadays, getting the editing right is
>> already quite tedious and might be likely to become
>> unbearable in the future if we don't do something about
>> it. 
>> 
>> There are several possible solutions:
>> 
>> a. Capabilities negotiation. This is the approach taken
>> in BGP. This approach seems to work well when there are
>> lots of extensions sets, thus making it likely that each side
>> might not be able to communicate well unless they reach
>> an agreement beforehand. 
>> 
>> b. Use a Mandatory bit as in L2TP. In this approach, typically 
>> only standardized attributes should be marked Mandatory, so as to
>> avoid potential "bullying." However, if there are sets of
>> standardized attributes that might not be implemented by a
>> client or server (as might be the case when there is a base
>> protocol and extension sets), then this might not work so well.  

as we have learned that we can not make vendor-specific non-standard
extensions truely mandatory, it devolves into a form of negotiation.

randy


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 13 16:43:21 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA03913;
	Wed, 13 Oct 1999 16:43:20 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id QAA17084
	for nasreq-outgoing; Wed, 13 Oct 1999 16:23:45 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id QAA17080
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 16:23:43 -0400 (EDT)
Received: from engmail4.Eng.Sun.COM ([129.144.134.6])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id NAA17627;
	Wed, 13 Oct 1999 13:23:42 -0700 (PDT)
Received: from hsmpka.eng.sun.com (hsmpka.Eng.Sun.COM [129.146.65.47])
	by engmail4.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id NAA15726;
	Wed, 13 Oct 1999 13:22:52 -0700 (PDT)
Received: from hsmpka.Eng.Sun.COM by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id NAA12024; Wed, 13 Oct 1999 13:22:41 -0700
From: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Message-Id: <199910132022.NAA12024@hsmpka.eng.sun.com>
Date: Wed, 13 Oct 1999 13:21:45 -0700
To: "Randy Bush" <randy@psg.com>
Cc: <nasreq@tdmx.rutgers.edu>
Reply-To: <Pat.Calhoun@eng.sun.com>
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
X-Mailer: Sun NetMail 2.3
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit


>> Actually, I do not see the two solutions provided as being mutually
>> exclusive. I think that both can be used simultaneously. The former
>> allows for one to identify standards-track extensions, and the later
>> to identify vendor-specific attributes.
>> 
>> PatC
>>> I agree with Pat.
>>> 
>>> One of the major reasons for insertion of RADIUS proxies
>>> is capabilities adjustment. Given the different dialects
>>> that are spoken nowadays, getting the editing right is
>>> already quite tedious and might be likely to become
>>> unbearable in the future if we don't do something about
>>> it. 
>>> 
>>> There are several possible solutions:
>>> 
>>> a. Capabilities negotiation. This is the approach taken
>>> in BGP. This approach seems to work well when there are
>>> lots of extensions sets, thus making it likely that each side
>>> might not be able to communicate well unless they reach
>>> an agreement beforehand. 
>>> 
>>> b. Use a Mandatory bit as in L2TP. In this approach, typically 
>>> only standardized attributes should be marked Mandatory, so as to
>>> avoid potential "bullying." However, if there are sets of
>>> standardized attributes that might not be implemented by a
>>> client or server (as might be the case when there is a base
>>> protocol and extension sets), then this might not work so well.  
>
>as we have learned that we can not make vendor-specific non-standard
>extensions truely mandatory, it devolves into a form of negotiation.
>
Correct. The question is whether this negotiation occurs prior to
any other AAA communication (read: bootstrap message), or on the fly 
(read: Sorry, I don't like attribute xyz). Proposal A (above) is the
former, while B is the latter.

PatC



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 13 16:50:42 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA04100;
	Wed, 13 Oct 1999 16:50:42 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id QAA17046
	for nasreq-outgoing; Wed, 13 Oct 1999 16:18:13 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id QAA17041
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 16:18:11 -0400 (EDT)
Received: from engmail4.Eng.Sun.COM ([129.144.134.6])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id NAA14899;
	Wed, 13 Oct 1999 13:18:07 -0700 (PDT)
Received: from hsmpka.eng.sun.com (phys-hsmpka.Eng.Sun.COM [129.146.65.37])
	by engmail4.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id NAA14540;
	Wed, 13 Oct 1999 13:17:15 -0700 (PDT)
Received: from hsmpka.Eng.Sun.COM by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id NAA09951; Wed, 13 Oct 1999 13:16:55 -0700
From: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Message-Id: <199910132016.NAA09951@hsmpka.eng.sun.com>
Date: Wed, 13 Oct 1999 13:16:05 -0700
To: <nasreq@tdmx.rutgers.edu>,
        "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
Reply-To: <Pat.Calhoun@eng.sun.com>
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
X-Mailer: Sun NetMail 2.3
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Not as I see it.

Let's take the hypothetical case where a base protocol exists with several
"extensions". Peers will only attempt to communicate together if they
know that the peers support an extension in question. 

However, vendor A may wish to send additional information in vendor specific
AVPs. Of course, he *should* only do this if the peer is of the same
manufacturer (one would think), but the RADIUS protocol has proved this to
be untrue. Server vendors are in the business of trying to figure out who
has done what with the protocol and implement it. Therefore, simply sending
a "vendor string" is not sufficient.

One other approach would be to send the vendor IDs that an implementation
supports, in addition to the extensions. But there may be hundreds of vendor
specific attributes for the vendor, and not all may be supported. Conversely,
one could send the whole dictionary as part of the capabilities exchange....
let's not even go there :)

This, I believe, is a tough question to answer (now that I've typed it in).
I strongly believe in capabilities negotiation. I also believe that a back-up
plan to support mandatory bits is also another way to ensure extensibility.
One alternative, of course, is to state that if you receive *any* attribute
that is unknown, the request must be rejected. The problem here is that the
attribute in question may contain purely informational data.

PatC
>If you knew that the other side could handle a capability (via
>capabilities negotiation) why would you *also* need to mark
>an attribute as Mandatory? Does this have something to do
>with proxy editing?
>
>-----Original Message-----
>From: Pat.Calhoun@eng.sun.com [mailto:Pat.Calhoun@eng.sun.com]
>Sent: Wednesday, October 13, 1999 12:34 PM
>To: Bernard Aboba (Exchange); nasreq@tdmx.rutgers.edu;
>'Pat.Calhoun@eng.sun.com'
>Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
>
>
>Actually, I do not see the two solutions provided as being mutually
>exclusive. I think that both can be used simultaneously. The former
>allows for one to identify standards-track extensions, and the later
>to identify vendor-specific attributes.
>
>PatC
>>I agree with Pat.
>>
>>One of the major reasons for insertion of RADIUS proxies
>>is capabilities adjustment. Given the different dialects
>>that are spoken nowadays, getting the editing right is
>>already quite tedious and might be likely to become
>>unbearable in the future if we don't do something about
>>it. 
>>
>>There are several possible solutions:
>>
>>a. Capabilities negotiation. This is the approach taken
>>in BGP. This approach seems to work well when there are
>>lots of extensions sets, thus making it likely that each side
>>might not be able to communicate well unless they reach
>>an agreement beforehand. 
>>
>>b. Use a Mandatory bit as in L2TP. In this approach, typically 
>>only standardized attributes should be marked Mandatory, so as to
>>avoid potential "bullying." However, if there are sets of
>>standardized attributes that might not be implemented by a
>>client or server (as might be the case when there is a base
>>protocol and extension sets), then this might not work so well.  
>




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 13 17:44:35 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA04955;
	Wed, 13 Oct 1999 17:44:35 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id RAA17543
	for nasreq-outgoing; Wed, 13 Oct 1999 17:23:17 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from rip.psg.com (rip.psg.com [147.28.0.39])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id RAA17539
	for <nasreq@tdmx.rutgers.edu>; Wed, 13 Oct 1999 17:23:15 -0400 (EDT)
Received: from randy by rip.psg.com with local (Exim 3.03 #1)
	id 11bVre-000KsI-00; Wed, 13 Oct 1999 14:23:14 -0700
From: Randy Bush <randy@psg.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Cc: <nasreq@tdmx.rutgers.edu>,
        "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
References: <199910132016.NAA09951@hsmpka.eng.sun.com>
Message-Id: <E11bVre-000KsI-00@rip.psg.com>
Date: Wed, 13 Oct 1999 14:23:14 -0700
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

> Let's take the hypothetical case where a base protocol exists with several
> "extensions". Peers will only attempt to communicate together if they
> know that the peers support an extension in question. 

s/communicate together/use the extension/



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 08:44:32 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id IAA27977;
	Thu, 14 Oct 1999 08:44:32 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id IAA21526
	for nasreq-outgoing; Thu, 14 Oct 1999 08:30:11 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from btm4r4.alcatel.be (btm4r4.alcatel.be [195.207.101.110])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id IAA21522
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 08:30:09 -0400 (EDT)
Received: from btmq9s.rc.bel.alcatel.be (btmq9s.rc.bel.alcatel.be [138.203.65.182])
	by btm4r4.alcatel.be (8.9.1a/8.9.1) with ESMTP id OAA10608;
	Thu, 14 Oct 1999 14:29:12 +0200 (MET DST)
Received: from alcatel.be (btm0l8 [138.203.65.218])
	by btmq9s.rc.bel.alcatel.be (8.8.8+Sun/8.8.8) with ESMTP id OAA11962;
	Thu, 14 Oct 1999 14:30:02 +0200 (MET DST)
Message-ID: <3805CC87.5A9753AD@alcatel.be>
Date: Thu, 14 Oct 1999 14:28:55 +0200
From: "Yves T'Joens" <yves.tjoens@alcatel.be>
Organization: Alcatel
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
X-Accept-Language: en
MIME-Version: 1.0
To: Pat.Calhoun@eng.sun.com
CC: nasreq@tdmx.rutgers.edu,
        "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
References: <199910132016.NAA09951@hsmpka.eng.sun.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Patrice Calhoun wrote:
> 
Hi Pat et al.,

> Not as I see it.
> 
> Let's take the hypothetical case where a base protocol exists with several
> "extensions". Peers will only attempt to communicate together if they
> know that the peers support an extension in question.
> 
> However, vendor A may wish to send additional information in vendor specific
> AVPs. Of course, he *should* only do this if the peer is of the same
> manufacturer (one would think), but the RADIUS protocol has proved this to
> be untrue. Server vendors are in the business of trying to figure out who
> has done what with the protocol and implement it. Therefore, simply sending
> a "vendor string" is not sufficient.
> 
> One other approach would be to send the vendor IDs that an implementation
> supports, in addition to the extensions. But there may be hundreds of vendor
> specific attributes for the vendor, and not all may be supported. Conversely,
> one could send the whole dictionary as part of the capabilities exchange....
> let's not even go there :)
> 
> This, I believe, is a tough question to answer (now that I've typed it in).
> I strongly believe in capabilities negotiation. I also believe that a back-up
> plan to support mandatory bits is also another way to ensure extensibility.
> One alternative, of course, is to state that if you receive *any* attribute
> that is unknown, the request must be rejected. The problem here is that the
> attribute in question may contain purely informational data.
> 

just to add a bit to the confusion.

Something that exists and is in use in e.g., ATM signalling protocols,
is the notion of pass along.
What happens if one has a proxy situation (e.g. over two aaa proxies,
the middle RADIUS server belonging to a broker), where some specific
attribute needs to be send end to end, but the next hop doesn't know it.

would it be a valid requirement to the protocol to be able to indicate
in the Attribute the behaviour expected from the 'proxy in the middle'.
Such behaviour could be e.g., if you do not understand it, please
however pass it trough to the other side.

this is an extension to the mandatory field, however specific to the
proxying behaviour and the distinction between end to end and hop by
hop.
if assumed a valid requirement, i can make up some text for it, just
shout.

cheers,
Yves



> PatC
> >If you knew that the other side could handle a capability (via
> >capabilities negotiation) why would you *also* need to mark
> >an attribute as Mandatory? Does this have something to do
> >with proxy editing?
> >
> >-----Original Message-----
> >From: Pat.Calhoun@eng.sun.com [mailto:Pat.Calhoun@eng.sun.com]
> >Sent: Wednesday, October 13, 1999 12:34 PM
> >To: Bernard Aboba (Exchange); nasreq@tdmx.rutgers.edu;
> >'Pat.Calhoun@eng.sun.com'
> >Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
> >
> >
> >Actually, I do not see the two solutions provided as being mutually
> >exclusive. I think that both can be used simultaneously. The former
> >allows for one to identify standards-track extensions, and the later
> >to identify vendor-specific attributes.
> >
> >PatC
> >>I agree with Pat.
> >>
> >>One of the major reasons for insertion of RADIUS proxies
> >>is capabilities adjustment. Given the different dialects
> >>that are spoken nowadays, getting the editing right is
> >>already quite tedious and might be likely to become
> >>unbearable in the future if we don't do something about
> >>it.
> >>
> >>There are several possible solutions:
> >>
> >>a. Capabilities negotiation. This is the approach taken
> >>in BGP. This approach seems to work well when there are
> >>lots of extensions sets, thus making it likely that each side
> >>might not be able to communicate well unless they reach
> >>an agreement beforehand.
> >>
> >>b. Use a Mandatory bit as in L2TP. In this approach, typically
> >>only standardized attributes should be marked Mandatory, so as to
> >>avoid potential "bullying." However, if there are sets of
> >>standardized attributes that might not be implemented by a
> >>client or server (as might be the case when there is a base
> >>protocol and extension sets), then this might not work so well.
> >


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 09:46:44 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA00563;
	Thu, 14 Oct 1999 09:46:44 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id JAA21791
	for nasreq-outgoing; Thu, 14 Oct 1999 09:32:39 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from btm4r4.alcatel.be (btm4r4.alcatel.be [195.207.101.110])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id JAA21787
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 09:32:34 -0400 (EDT)
Received: from btmq9s.rc.bel.alcatel.be (btmq9s.rc.bel.alcatel.be [138.203.65.182])
	by btm4r4.alcatel.be (8.9.1a/8.9.1) with ESMTP id PAA25587;
	Thu, 14 Oct 1999 15:30:53 +0200 (MET DST)
Received: from alcatel.be (btm0l8 [138.203.65.218])
	by btmq9s.rc.bel.alcatel.be (8.8.8+Sun/8.8.8) with ESMTP id PAA01108;
	Thu, 14 Oct 1999 15:31:43 +0200 (MET DST)
Message-ID: <3805DAFD.B160F2E9@alcatel.be>
Date: Thu, 14 Oct 1999 15:30:37 +0200
From: "Yves T'Joens" <yves.tjoens@alcatel.be>
Organization: Alcatel
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
X-Accept-Language: en
MIME-Version: 1.0
To: nasreq <nasreq@tdmx.rutgers.edu>, mbeadles@wcom.net
Subject: nasreq comments on criteria-03
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Hi Mark,

Very nice doc, 
some very minor comments/questions on the document.

section 8.1.2.1 fast fail over.

do you mean here the detection that fast fail over is necessary or the
fail over mechanism itself ?
To be more precise, if it would be the detection, would that mean that a
keep alive mechanism must work on such a (pretty fast ;-) ) time scheme,
or do you refer more exactly that no whole database synchronization
should happen between client and server on fail over...

section 8.1.3 Attribute value protocol

this section only deals with attribute extensibility, should message
extensibility also be mentioned explicitly ??

section 8.1.3.1 Attribute data types

this section words the requirement as "... simple attribute data types,
including integer, enumeration, text string...." Does the 'including'
refer to the fact that many more simple data types may be defined, if
so, then perhaps something like 'not limited to' might be added to the
sentence. I was thinking on opaque values to be included in the AVP...

section  8.3.2.2 Authorizatoin attribute requirements

-Quality of Service : this requirement seems to be so wide that it is
hardly applicable, should we somehow more specify it, e.g., should
support authorization to intserv based resource requests (rsvp) or
diffserv forwarding behaviour... ?



cheers,
Yves


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 09:46:49 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA00580;
	Thu, 14 Oct 1999 09:46:48 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id JAA21738
	for nasreq-outgoing; Thu, 14 Oct 1999 09:19:57 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from dfssl.exchange.microsoft.com (dfssl.exchange.microsoft.com [131.107.88.59])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id JAA21734
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 09:19:55 -0400 (EDT)
Received: by dfssl with Internet Mail Service (5.5.2650.21)
	id <4RCWW051>; Thu, 14 Oct 1999 06:19:00 -0700
Message-ID: <078292D50C98D2118D090008C7E9C6A60225816B@STAY.platinum.corp.microsoft.com>
From: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
To: "'Yves T'Joens'" <yves.tjoens@alcatel.be>, Pat.Calhoun@eng.sun.com
Cc: nasreq@tdmx.rutgers.edu,
        "Bernard Aboba (Exchange)"
	 <bernarda@Exchange.Microsoft.com>
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
Date: Thu, 14 Oct 1999 06:18:59 -0700
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

>What happens if one has a proxy situation (e.g. over two aaa proxies,
>the middle RADIUS server belonging to a broker), where some specific
>attribute needs to be send end to end, but the next hop doesn't know it.

Interesting case. There are a couple of mechanisms being talked about
in the requirements. One is end-to-end security. This allows you
to indicate that the attribute in question cannot be edited and must be
passed
along. However the question is how the endpoint can know that this
attribute can be understood by its end destination. Capabilities
negotiation won't give you this because it's only a hop-by-hop
mechanism. Since you need this attribute to be passed along, if
it is a standard one then I'd presume it would be marked
Mandatory. However, that doesn't influence proxy behavior; it 
just tells the end destination how to behave in case it doesn't
understand the attribute. So it isn't entirely clear to me that
this case is covered by what we've been talking about.


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 10:00:57 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA01103;
	Thu, 14 Oct 1999 10:00:56 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id JAA21810
	for nasreq-outgoing; Thu, 14 Oct 1999 09:36:06 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from btm4r4.alcatel.be (btm4r4.alcatel.be [195.207.101.110])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id JAA21806
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 09:36:04 -0400 (EDT)
Received: from btmq9s.rc.bel.alcatel.be (btmq9s.rc.bel.alcatel.be [138.203.65.182])
	by btm4r4.alcatel.be (8.9.1a/8.9.1) with ESMTP id PAA26764;
	Thu, 14 Oct 1999 15:35:33 +0200 (MET DST)
Received: from alcatel.be (btm0l8 [138.203.65.218])
	by btmq9s.rc.bel.alcatel.be (8.8.8+Sun/8.8.8) with ESMTP id PAA07009;
	Thu, 14 Oct 1999 15:36:27 +0200 (MET DST)
Message-ID: <3805DC19.2D13D44F@alcatel.be>
Date: Thu, 14 Oct 1999 15:35:21 +0200
From: "Yves T'Joens" <yves.tjoens@alcatel.be>
Organization: Alcatel
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
X-Accept-Language: en
MIME-Version: 1.0
To: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
CC: Pat.Calhoun@eng.sun.com, nasreq@tdmx.rutgers.edu
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
References: <078292D50C98D2118D090008C7E9C6A60225816B@STAY.platinum.corp.microsoft.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

"Bernard Aboba (Exchange)" wrote:
> 
> >What happens if one has a proxy situation (e.g. over two aaa proxies,
> >the middle RADIUS server belonging to a broker), where some specific
> >attribute needs to be send end to end, but the next hop doesn't know it.
> 
> Interesting case. There are a couple of mechanisms being talked about
> in the requirements. One is end-to-end security. This allows you
> to indicate that the attribute in question cannot be edited and must be
> passed
> along. However the question is how the endpoint can know that this
> attribute can be understood by its end destination. Capabilities
> negotiation won't give you this because it's only a hop-by-hop
> mechanism. Since you need this attribute to be passed along, if
> it is a standard one then I'd presume it would be marked
> Mandatory. However, that doesn't influence proxy behavior; it
> just tells the end destination how to behave in case it doesn't
> understand the attribute. So it isn't entirely clear to me that
> this case is covered by what we've been talking about.

use of the mandatory bit indeed doesn't help here.
Note that if a 'generic' AAA protocol is in use, it would occur a lot I
presume that extensions are written to it, that may be understood by
both end points, while the 'proxy in the middle' still runs an older
version... so somehow, I think we should certainly have a look at this
requirement. As I said before, one can learn something from the way this
has been handled in ATM signalling (correct in the ITU, messed up in the
ATM forum...)


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 10:37:41 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA02790;
	Thu, 14 Oct 1999 10:37:40 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id KAA21979
	for nasreq-outgoing; Thu, 14 Oct 1999 10:11:04 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id KAA21975
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 10:11:02 -0400 (EDT)
Received: from engmail4.Eng.Sun.COM ([129.144.134.6])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id HAA18721;
	Thu, 14 Oct 1999 07:10:54 -0700 (PDT)
Received: from hsmpka.eng.sun.com (hsmpka.Eng.Sun.COM [129.146.105.47])
	by engmail4.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id HAA23556;
	Thu, 14 Oct 1999 07:10:49 -0700 (PDT)
Received: from mordor by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id HAA00062; Thu, 14 Oct 1999 07:10:48 -0700
Date: Thu, 14 Oct 1999 07:10:48 -0700 (PDT)
From: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>
Reply-To: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
To: "Yves T'Joens" <yves.tjoens@alcatel.be>
Cc: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>,
        Pat.Calhoun@eng.sun.com, nasreq@tdmx.rutgers.edu
In-Reply-To: "Your message with ID" <3805DC19.2D13D44F@alcatel.be>
Message-ID: <Roam.SIMC.2.0.6.939910248.17295.pcalhoun@hsmpka>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk


> use of the mandatory bit indeed doesn't help here.
> Note that if a 'generic' AAA protocol is in use, it would occur a lot I
> presume that extensions are written to it, that may be understood by
> both end points, while the 'proxy in the middle' still runs an older
> version... so somehow, I think we should certainly have a look at this
> requirement. As I said before, one can learn something from the way this
> has been handled in ATM signalling (correct in the ITU, messed up in the
> ATM forum...)

One of the things that we seem to be forgetting is the fact that AAA
communication can occur over several intermediate hosts, or proxies. The way
that I view capabilities exchange is not end to end. It is VERY difficult to
do end to end capabilities exchange, especially since it would need to be done
prior to the actual auth request.

I view capabilities exchange as a hop-by-hop thing that allows two peers to
know what each other support. Let's take an example of a NAS installed in a
network with 4 AAA servers. Using capabilities exchange it could find out that
server B is to be used for Mobile IP Auth, while server A does PPP and server
C + D support accounting. Today this would have to be manually configured on a
NAS, which is administratively tedious.

Moving away from the NAS, when the AAA server talks to a broker, there may be
different servers within the broker network for each service. Similarly, when
the broker's server talks to the home network's AAA server specific servers
may perform specific functions. This is what I view capabilities exchange for,
reduction of manual configuration.

When the user dials into a NAS, the request reaches the local Server, then the
broker and finally the home server. Having the NAS perform some capabilities
exchange with the home network, in which each intermediate node needs to adjust
the capabilities to reflect their own supported extensions, makes this a VERY
difficult problem and dramatically increases the latency involved in the call
setup. This would not fly for the wireless space.

For this reason, I would prefer to have both capabilities exchange, and the
mandatory bit. The AAA protocol MUST allow a server to return an error if it
receives an unknown attribute, and it SHOULD even specify which attribute
caused the error(s). This would allow a clever implementation to try again
without using the offending attribute, if it wants to.

PatC



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 10:43:25 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA03024;
	Thu, 14 Oct 1999 10:43:24 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id KAA22041
	for nasreq-outgoing; Thu, 14 Oct 1999 10:21:56 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from rip.psg.com (rip.psg.com [147.28.0.39])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id KAA22037
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 10:21:55 -0400 (EDT)
Received: from randy by rip.psg.com with local (Exim 3.03 #1)
	id 11bllP-0003D7-00; Thu, 14 Oct 1999 07:21:51 -0700
From: Randy Bush <randy@psg.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: Bernard Aboba <bernarda@Exchange.Microsoft.com>
Cc: nasreq@tdmx.rutgers.edu
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
References: <078292D50C98D2118D090008C7E9C6A60225816B@STAY.platinum.corp.microsoft.com>
Message-Id: <E11bllP-0003D7-00@rip.psg.com>
Date: Thu, 14 Oct 1999 07:21:51 -0700
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

<ad hat off>

i am slow.

given that security and trust are usually end-to-end due to a long history
of middle weaknesses, why are smart (i.e. able to look at content) relay nas
devices needed?  it seems to me that they will be an ever-expanding source
of complications.

as the answer is probably obvious to all here except for me, private
education may be more appropriate than major distraction on the list.

randy


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 10:51:33 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA03226;
	Thu, 14 Oct 1999 10:51:31 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id KAA22104
	for nasreq-outgoing; Thu, 14 Oct 1999 10:35:30 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from btm4r4.alcatel.be (btm4r4.alcatel.be [195.207.101.110])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id KAA22100
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 10:35:23 -0400 (EDT)
Received: from btmq9s.rc.bel.alcatel.be (btmq9s.rc.bel.alcatel.be [138.203.65.182])
	by btm4r4.alcatel.be (8.9.1a/8.9.1) with ESMTP id QAA10177;
	Thu, 14 Oct 1999 16:34:52 +0200 (MET DST)
Received: from alcatel.be (btm0l8 [138.203.65.218])
	by btmq9s.rc.bel.alcatel.be (8.8.8+Sun/8.8.8) with ESMTP id QAA14234;
	Thu, 14 Oct 1999 16:35:47 +0200 (MET DST)
Message-ID: <3805EA01.EAC6F261@alcatel.be>
Date: Thu, 14 Oct 1999 16:34:41 +0200
From: "Yves T'Joens" <yves.tjoens@alcatel.be>
Organization: Alcatel
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
X-Accept-Language: en
MIME-Version: 1.0
To: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>,
        nasreq <nasreq@tdmx.rutgers.edu>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
References: <Roam.SIMC.2.0.6.939910248.17295.pcalhoun@hsmpka>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

"pcalhoun@eng.sun.com" wrote:
> 
> > use of the mandatory bit indeed doesn't help here.
> > Note that if a 'generic' AAA protocol is in use, it would occur a lot I
> > presume that extensions are written to it, that may be understood by
> > both end points, while the 'proxy in the middle' still runs an older
> > version... so somehow, I think we should certainly have a look at this
> > requirement. As I said before, one can learn something from the way this
> > has been handled in ATM signalling (correct in the ITU, messed up in the
> > ATM forum...)
> 
> One of the things that we seem to be forgetting is the fact that AAA
> communication can occur over several intermediate hosts, or proxies. The way
> that I view capabilities exchange is not end to end. It is VERY difficult to
> do end to end capabilities exchange, especially since it would need to be done
> prior to the actual auth request.

I did not even mention the difference between end to end and hop by hop
capabilities exchange. I know it is VERY difficult, the question is, is
it necessary.

> 
> I view capabilities exchange as a hop-by-hop thing that allows two peers to
> know what each other support. Let's take an example of a NAS installed in a
> network with 4 AAA servers. Using capabilities exchange it could find out that
> server B is to be used for Mobile IP Auth, while server A does PPP and server
> C + D support accounting. Today this would have to be manually configured on a
> NAS, which is administratively tedious.

The NAS still needs to be configured on how to reach the other server
(either manually or by other means), so why not add an indication of the
server capabilities at the same time.

> 
> Moving away from the NAS, when the AAA server talks to a broker, there may be
> different servers within the broker network for each service. Similarly, when
> the broker's server talks to the home network's AAA server specific servers
> may perform specific functions. This is what I view capabilities exchange for,
> reduction of manual configuration.
> 
> When the user dials into a NAS, the request reaches the local Server, then the
> broker and finally the home server. Having the NAS perform some capabilities
> exchange with the home network, in which each intermediate node needs to adjust
> the capabilities to reflect their own supported extensions, makes this a VERY
> difficult problem and dramatically increases the latency involved in the call
> setup. This would not fly for the wireless space.

Indeed in such a configuration, a sort of trial and error scheme would
be better than reaching agreement prior to attribute exchange, given
that 99% of the trials will work, while the other 1% fail.

> 
> For this reason, I would prefer to have both capabilities exchange, and the
> mandatory bit. The AAA protocol MUST allow a server to return an error if it
> receives an unknown attribute, and it SHOULD even specify which attribute
> caused the error(s). This would allow a clever implementation to try again
> without using the offending attribute, if it wants to.
> 

This imposes a bit of an configuration burden. If one imagines a
consortium of ISPs that allow roaming of any kind, then one is surely
limited by the capabilities of the broker proxy.
Let me try to give an example.
Assume within the present roaming consortium, some local ISP starts
offering QoS like services to its customers. Roaming users that have
also these capabilities at home, may request them while roaming. A
supporting local ISP may thus add to the authorization request some new
AVP (totally unknown AVP to the broker), just to indicate this (assume
these capabilities to be already 'standard' but not implemented by all
ISPs and not by the broker). 
My question is then, will they be able to pass the broker proxy ? 
The local ISP may mark the attribute as mandatory, since it is
'standard', but of no concern to the broker. It would be nice if the
local ISP could indicate in the AVP that the information has only end to
end value, and that the broker should pass it unharmed, even if not
understood.

I don't see this requirement in the present nasreq document.

Note that this is orthogonal to the issue of capability negotiation
between the two endpoints (although related).


> PatC

Yves


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 11:03:22 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA03529;
	Thu, 14 Oct 1999 11:03:21 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id KAA22163
	for nasreq-outgoing; Thu, 14 Oct 1999 10:44:53 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id KAA22159
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 10:44:51 -0400 (EDT)
Received: from engmail2.Eng.Sun.COM ([129.146.1.25])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id HAA07743;
	Thu, 14 Oct 1999 07:44:47 -0700 (PDT)
Received: from hsmpka.eng.sun.com (phys-hsmpka.Eng.Sun.COM [129.146.63.37])
	by engmail2.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id HAA05429;
	Thu, 14 Oct 1999 07:44:47 -0700 (PDT)
Received: from mordor by hsmpka.eng.sun.com (SMI-8.6/SMI-SVR4)
	id HAA06337; Thu, 14 Oct 1999 07:44:45 -0700
Date: Thu, 14 Oct 1999 07:44:45 -0700 (PDT)
From: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>
Reply-To: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
To: "Yves T'Joens" <yves.tjoens@alcatel.be>
Cc: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>,
        nasreq <nasreq@tdmx.rutgers.edu>
In-Reply-To: "Your message with ID" <3805EA01.EAC6F261@alcatel.be>
Message-ID: <Roam.SIMC.2.0.6.939912285.19073.pcalhoun@hsmpka>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

> "pcalhoun@eng.sun.com" wrote:
> > 
> > > use of the mandatory bit indeed doesn't help here.
> > > Note that if a 'generic' AAA protocol is in use, it would occur a lot I
> > > presume that extensions are written to it, that may be understood by
> > > both end points, while the 'proxy in the middle' still runs an older
> > > version... so somehow, I think we should certainly have a look at this
> > > requirement. As I said before, one can learn something from the way this
> > > has been handled in ATM signalling (correct in the ITU, messed up in the
> > > ATM forum...)
> > 
> > One of the things that we seem to be forgetting is the fact that AAA
> > communication can occur over several intermediate hosts, or proxies. The way
> > that I view capabilities exchange is not end to end. It is VERY difficult to
> > do end to end capabilities exchange, especially since it would need to be done
> > prior to the actual auth request.
> 
> I did not even mention the difference between end to end and hop by hop
> capabilities exchange. I know it is VERY difficult, the question is, is
> it necessary.
> 
> > 
> > I view capabilities exchange as a hop-by-hop thing that allows two peers to
> > know what each other support. Let's take an example of a NAS installed in a
> > network with 4 AAA servers. Using capabilities exchange it could find out that
> > server B is to be used for Mobile IP Auth, while server A does PPP and server
> > C + D support accounting. Today this would have to be manually configured on a
> > NAS, which is administratively tedious.
> 
> The NAS still needs to be configured on how to reach the other server
> (either manually or by other means), so why not add an indication of the
> server capabilities at the same time.

I thought this was my point....

> 
> > 
> > Moving away from the NAS, when the AAA server talks to a broker, there may be
> > different servers within the broker network for each service. Similarly, when
> > the broker's server talks to the home network's AAA server specific servers
> > may perform specific functions. This is what I view capabilities exchange for,
> > reduction of manual configuration.
> > 
> > When the user dials into a NAS, the request reaches the local Server, then the
> > broker and finally the home server. Having the NAS perform some capabilities
> > exchange with the home network, in which each intermediate node needs to adjust
> > the capabilities to reflect their own supported extensions, makes this a VERY
> > difficult problem and dramatically increases the latency involved in the call
> > setup. This would not fly for the wireless space.
> 
> Indeed in such a configuration, a sort of trial and error scheme would
> be better than reaching agreement prior to attribute exchange, given
> that 99% of the trials will work, while the other 1% fail.

I think that the reality of it all is that protocol extensibility via vendor
specific attributes works well intra-domain. For inter-domain cases, any
vendor specific attributes should NOT be sent. Perhaps this is the rule that
needs to be applied, and no protocol requirements would be required.

> 
> > 
> > For this reason, I would prefer to have both capabilities exchange, and the
> > mandatory bit. The AAA protocol MUST allow a server to return an error if it
> > receives an unknown attribute, and it SHOULD even specify which attribute
> > caused the error(s). This would allow a clever implementation to try again
> > without using the offending attribute, if it wants to.
> > 
> 
> This imposes a bit of an configuration burden. If one imagines a
> consortium of ISPs that allow roaming of any kind, then one is surely
> limited by the capabilities of the broker proxy.
> Let me try to give an example.
> Assume within the present roaming consortium, some local ISP starts
> offering QoS like services to its customers. Roaming users that have
> also these capabilities at home, may request them while roaming. A
> supporting local ISP may thus add to the authorization request some new
> AVP (totally unknown AVP to the broker), just to indicate this (assume
> these capabilities to be already 'standard' but not implemented by all
> ISPs and not by the broker). 
> My question is then, will they be able to pass the broker proxy ? 
> The local ISP may mark the attribute as mandatory, since it is
> 'standard', but of no concern to the broker. It would be nice if the
> local ISP could indicate in the AVP that the information has only end to
> end value, and that the broker should pass it unharmed, even if not
> understood.
> 
hmmm... well if the broker is in it for the long run, and plans to keep its
customers, it very well better start deploying new protocol extensions.
Otherwise ISPs will be forced to look for better service from another broker
(and even today we do not have a shortage of brokers).

Even having had this discussion, I am still convinced that capabilities
exchange (hop by hop) provides enough functionality to make most of the
network work well. I think that my statement above that vendor specific stuff
should be stripped when requests are sent inter-domain would ensure that we do
not run into the end-to-end problem.

PatC



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 11:29:09 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA04309;
	Thu, 14 Oct 1999 11:29:08 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id LAA22299
	for nasreq-outgoing; Thu, 14 Oct 1999 11:07:53 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from btm4r4.alcatel.be (btm4r4.alcatel.be [195.207.101.110])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id LAA22287
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 11:07:46 -0400 (EDT)
Received: from btmq9s.rc.bel.alcatel.be (btmq9s.rc.bel.alcatel.be [138.203.65.182])
	by btm4r4.alcatel.be (8.9.1a/8.9.1) with ESMTP id RAA16038;
	Thu, 14 Oct 1999 17:07:14 +0200 (MET DST)
Received: from alcatel.be (btm0l8 [138.203.65.218])
	by btmq9s.rc.bel.alcatel.be (8.8.8+Sun/8.8.8) with ESMTP id RAA16542;
	Thu, 14 Oct 1999 17:08:09 +0200 (MET DST)
Message-ID: <3805F197.A1603B63@alcatel.be>
Date: Thu, 14 Oct 1999 17:07:03 +0200
From: "Yves T'Joens" <yves.tjoens@alcatel.be>
Organization: Alcatel
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
X-Accept-Language: en
MIME-Version: 1.0
To: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>
CC: nasreq <nasreq@tdmx.rutgers.edu>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
References: <Roam.SIMC.2.0.6.939912285.19073.pcalhoun@hsmpka>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

"pcalhoun@eng.sun.com" wrote:
> 
> > "pcalhoun@eng.sun.com" wrote:
> > >
[...]
> > >
> > > I view capabilities exchange as a hop-by-hop thing that allows two peers to
> > > know what each other support. Let's take an example of a NAS installed in a
> > > network with 4 AAA servers. Using capabilities exchange it could find out that
> > > server B is to be used for Mobile IP Auth, while server A does PPP and server
> > > C + D support accounting. Today this would have to be manually configured on a
> > > NAS, which is administratively tedious.
> >
> > The NAS still needs to be configured on how to reach the other server
> > (either manually or by other means), so why not add an indication of the
> > server capabilities at the same time.
> 
> I thought this was my point....

By that means, SNMP would be used to configure a NAS with the list of
all its peer AAA servers, while the capability exchange would be used to
see what the nature of the servers is, and the range of attributes they
support. My intention was to say, that the nature of the server could be
configured alongside its IP address. For the support of attributes, I
agree...

> > > When the user dials into a NAS, the request reaches the local Server, then the
> > > broker and finally the home server. Having the NAS perform some capabilities
> > > exchange with the home network, in which each intermediate node needs to adjust
> > > the capabilities to reflect their own supported extensions, makes this a VERY
> > > difficult problem and dramatically increases the latency involved in the call
> > > setup. This would not fly for the wireless space.
> >
> > Indeed in such a configuration, a sort of trial and error scheme would
> > be better than reaching agreement prior to attribute exchange, given
> > that 99% of the trials will work, while the other 1% fail.
> 
> I think that the reality of it all is that protocol extensibility via vendor
> specific attributes works well intra-domain. For inter-domain cases, any
> vendor specific attributes should NOT be sent. Perhaps this is the rule that
> needs to be applied, and no protocol requirements would be required.
> 

I was not exactly refering to vendor specific attributes, but to all
attributes, in a world, where capabilities change and the main concern
is to have differentiation in services, and where one wants to avoid the
need to upgrade the whole network just to get some information end to
end...

> >
> > >
> > > For this reason, I would prefer to have both capabilities exchange, and the
> > > mandatory bit. The AAA protocol MUST allow a server to return an error if it
> > > receives an unknown attribute, and it SHOULD even specify which attribute
> > > caused the error(s). This would allow a clever implementation to try again
> > > without using the offending attribute, if it wants to.
> > >
> >
> > This imposes a bit of an configuration burden. If one imagines a
> > consortium of ISPs that allow roaming of any kind, then one is surely
> > limited by the capabilities of the broker proxy.
> > Let me try to give an example.
> > Assume within the present roaming consortium, some local ISP starts
> > offering QoS like services to its customers. Roaming users that have
> > also these capabilities at home, may request them while roaming. A
> > supporting local ISP may thus add to the authorization request some new
> > AVP (totally unknown AVP to the broker), just to indicate this (assume
> > these capabilities to be already 'standard' but not implemented by all
> > ISPs and not by the broker).
> > My question is then, will they be able to pass the broker proxy ?
> > The local ISP may mark the attribute as mandatory, since it is
> > 'standard', but of no concern to the broker. It would be nice if the
> > local ISP could indicate in the AVP that the information has only end to
> > end value, and that the broker should pass it unharmed, even if not
> > understood.
> >
> hmmm... well if the broker is in it for the long run, and plans to keep its
> customers, it very well better start deploying new protocol extensions.
> Otherwise ISPs will be forced to look for better service from another broker
> (and even today we do not have a shortage of brokers).

well....

> 
> Even having had this discussion, I am still convinced that capabilities
> exchange (hop by hop) provides enough functionality to make most of the
> network work well. I think that my statement above that vendor specific stuff
> should be stripped when requests are sent inter-domain would ensure that we do
> not run into the end-to-end problem.
> 
> PatC


Yves


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 11:29:10 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA04313;
	Thu, 14 Oct 1999 11:29:09 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id LAA22292
	for nasreq-outgoing; Thu, 14 Oct 1999 11:07:48 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mailman.cisco.com (mailman.cisco.com [171.68.225.9])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id LAA22288
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 11:07:46 -0400 (EDT)
Received: from chsharp-tecra (chsharp-isdn.cisco.com [171.68.116.221]) by mailman.cisco.com (8.8.8+Sun/CISCO.SERVER.1.2) with ESMTP id IAA10709; Thu, 14 Oct 1999 08:06:35 -0700 (PDT)
Message-Id: <4.2.0.58.19991014104557.00a70160@dogwood.cisco.com>
X-Sender: chsharp@dogwood.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 
Date: Thu, 14 Oct 1999 10:57:27 -0400
To: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>,
        "Yves T'Joens" <yves.tjoens@alcatel.be>
From: Chip Sharp <chsharp@cisco.com>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
Cc: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>,
        nasreq@tdmx.rutgers.edu
In-Reply-To: <Roam.SIMC.2.0.6.939910248.17295.pcalhoun@hsmpka>
References: <"Your message with ID" <3805DC19.2D13D44F@alcatel.be>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

At 07:10 AM 10/14/99 -0700, pcalhoun@eng.sun.com wrote:

> > use of the mandatory bit indeed doesn't help here.
> > Note that if a 'generic' AAA protocol is in use, it would occur a lot I
> > presume that extensions are written to it, that may be understood by
> > both end points, while the 'proxy in the middle' still runs an older
> > version... so somehow, I think we should certainly have a look at this
> > requirement. As I said before, one can learn something from the way this
> > has been handled in ATM signalling (correct in the ITU, messed up in the
> > ATM forum...)
...snip...
>For this reason, I would prefer to have both capabilities exchange, and the
>mandatory bit. The AAA protocol MUST allow a server to return an error if it
>receives an unknown attribute, and it SHOULD even specify which attribute
>caused the error(s). This would allow a clever implementation to try again
>without using the offending attribute, if it wants to.

If the AAA protocol allows a server (or client) to return an error if 
receives an unknown attribute, why is a Mandatory bit needed?  If the 
originator of a request receives an error response for an attribute that it 
thinks is mandatory, it should be able to take appropriate action (e.g., 
clear the call, cancel the request, send an SNMP Trap, etc.).  If the two 
endpoints are so out of synch that one is using mandatory attributes the 
other doesn't recognize, it is more than just a transitory failure.

As discovered in the L2TP saga, the message from IESG seems to be "Just Say 
No" to Mandatory bits.

Chip

>PatC

Support NetAid!  http://www.netaid.org
--------------------------------------------------
Chip Sharp                 Consulting Engineering
Cisco Systems              Telco Bio-region
Reality - Love it or Leave it.			
--------------------------------------------------


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 11:41:57 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA04773;
	Thu, 14 Oct 1999 11:41:56 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id LAA22362
	for nasreq-outgoing; Thu, 14 Oct 1999 11:15:54 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from btm4r4.alcatel.be (btm4r4.alcatel.be [195.207.101.110])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id LAA22358
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 11:15:49 -0400 (EDT)
Received: from btmq9s.rc.bel.alcatel.be (btmq9s.rc.bel.alcatel.be [138.203.65.182])
	by btm4r4.alcatel.be (8.9.1a/8.9.1) with ESMTP id RAA17512;
	Thu, 14 Oct 1999 17:15:18 +0200 (MET DST)
Received: from alcatel.be (btm0l8 [138.203.65.218])
	by btmq9s.rc.bel.alcatel.be (8.8.8+Sun/8.8.8) with ESMTP id RAA17009;
	Thu, 14 Oct 1999 17:16:18 +0200 (MET DST)
Message-ID: <3805F380.5BA8B257@alcatel.be>
Date: Thu, 14 Oct 1999 17:15:12 +0200
From: "Yves T'Joens" <yves.tjoens@alcatel.be>
Organization: Alcatel
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
X-Accept-Language: en
MIME-Version: 1.0
To: Chip Sharp <chsharp@cisco.com>
CC: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>,
        "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>,
        nasreq@tdmx.rutgers.edu
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
References: <"Your message with ID" <3805DC19.2D13D44F@alcatel.be> <4.2.0.58.19991014104557.00a70160@dogwood.cisco.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Chip Sharp wrote:
> 
> At 07:10 AM 10/14/99 -0700, pcalhoun@eng.sun.com wrote:
> 
> > > use of the mandatory bit indeed doesn't help here.
> > > Note that if a 'generic' AAA protocol is in use, it would occur a lot I
> > > presume that extensions are written to it, that may be understood by
> > > both end points, while the 'proxy in the middle' still runs an older
> > > version... so somehow, I think we should certainly have a look at this
> > > requirement. As I said before, one can learn something from the way this
> > > has been handled in ATM signalling (correct in the ITU, messed up in the
> > > ATM forum...)
> ...snip...
> >For this reason, I would prefer to have both capabilities exchange, and the
> >mandatory bit. The AAA protocol MUST allow a server to return an error if it
> >receives an unknown attribute, and it SHOULD even specify which attribute
> >caused the error(s). This would allow a clever implementation to try again
> >without using the offending attribute, if it wants to.
> 
> If the AAA protocol allows a server (or client) to return an error if
> receives an unknown attribute, why is a Mandatory bit needed?  If the
> originator of a request receives an error response for an attribute that it
> thinks is mandatory, it should be able to take appropriate action (e.g.,
> clear the call, cancel the request, send an SNMP Trap, etc.).  If the two
> endpoints are so out of synch that one is using mandatory attributes the
> other doesn't recognize, it is more than just a transitory failure.

there is good reason for the mandatory bit. It can be used to indicate :
if you don't understand this, too bad, just throw it away and proceed.
no error messages.

> 
> As discovered in the L2TP saga, the message from IESG seems to be "Just Say
> No" to Mandatory bits.

... in vendor specific attributes...
> 
> Chip

yves
> 
> >PatC
> 
> Support NetAid!  http://www.netaid.org
> --------------------------------------------------
> Chip Sharp                 Consulting Engineering
> Cisco Systems              Telco Bio-region
> Reality - Love it or Leave it.
> --------------------------------------------------


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 11:46:23 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA04938;
	Thu, 14 Oct 1999 11:46:23 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id LAA22383
	for nasreq-outgoing; Thu, 14 Oct 1999 11:16:28 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from rip.psg.com (rip.psg.com [147.28.0.39])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id LAA22379
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 11:16:27 -0400 (EDT)
Received: from randy by rip.psg.com with local (Exim 3.03 #1)
	id 11bmc5-0003iy-00; Thu, 14 Oct 1999 08:16:17 -0700
From: Randy Bush <randy@psg.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>
Cc: "Yves T'Joens" <yves.tjoens@alcatel.be>, nasreq <nasreq@tdmx.rutgers.edu>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
References: <3805EA01.EAC6F261@alcatel.be>
	<Roam.SIMC.2.0.6.939912285.19073.pcalhoun@hsmpka>
Message-Id: <E11bmc5-0003iy-00@rip.psg.com>
Date: Thu, 14 Oct 1999 08:16:17 -0700
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

> I think that the reality of it all is that protocol extensibility via
> vendor specific attributes works well intra-domain.

does this imply an assumption of intra-domain vendor homogeniety?  if so,
and if by 'domain' your mean e.g. one isp.  this is certainly not the case
in many very significant domains of which i am aware.

randy


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 12:57:14 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA07212;
	Thu, 14 Oct 1999 12:57:14 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id MAA22717
	for nasreq-outgoing; Thu, 14 Oct 1999 12:29:36 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from mailman.cisco.com (mailman.cisco.com [171.68.225.9])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id MAA22713
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 12:29:34 -0400 (EDT)
Received: from chsharp-tecra (chsharp-isdn.cisco.com [171.68.116.221]) by mailman.cisco.com (8.8.8+Sun/CISCO.SERVER.1.2) with ESMTP id JAA05289; Thu, 14 Oct 1999 09:25:32 -0700 (PDT)
Message-Id: <4.2.0.58.19991014113419.00b1b100@dogwood.cisco.com>
X-Sender: chsharp@dogwood.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 
Date: Thu, 14 Oct 1999 11:47:07 -0400
To: "Yves T'Joens" <yves.tjoens@alcatel.be>
From: Chip Sharp <chsharp@cisco.com>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
Cc: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>,
        "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>,
        nasreq@tdmx.rutgers.edu
In-Reply-To: <3805F380.5BA8B257@alcatel.be>
References: <"Your message with ID" <3805DC19.2D13D44F@alcatel.be>
 <4.2.0.58.19991014104557.00a70160@dogwood.cisco.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

At 05:15 PM 10/14/99 +0200, Yves T'Joens wrote:
>Chip Sharp wrote:
> >
> > At 07:10 AM 10/14/99 -0700, pcalhoun@eng.sun.com wrote:
> >
> > > > use of the mandatory bit indeed doesn't help here.
> > > > Note that if a 'generic' AAA protocol is in use, it would occur a lot I
> > > > presume that extensions are written to it, that may be understood by
> > > > both end points, while the 'proxy in the middle' still runs an older
> > > > version... so somehow, I think we should certainly have a look at this
> > > > requirement. As I said before, one can learn something from the way 
> this
> > > > has been handled in ATM signalling (correct in the ITU, messed up 
> in the
> > > > ATM forum...)
> > ...snip...
> > >For this reason, I would prefer to have both capabilities exchange, 
> and the
> > >mandatory bit. The AAA protocol MUST allow a server to return an error 
> if it
> > >receives an unknown attribute, and it SHOULD even specify which attribute
> > >caused the error(s). This would allow a clever implementation to try again
> > >without using the offending attribute, if it wants to.
> >
> > If the AAA protocol allows a server (or client) to return an error if
> > receives an unknown attribute, why is a Mandatory bit needed?  If the
> > originator of a request receives an error response for an attribute that it
> > thinks is mandatory, it should be able to take appropriate action (e.g.,
> > clear the call, cancel the request, send an SNMP Trap, etc.).  If the two
> > endpoints are so out of synch that one is using mandatory attributes the
> > other doesn't recognize, it is more than just a transitory failure.
>
>there is good reason for the mandatory bit. It can be used to indicate :
>if you don't understand this, too bad, just throw it away and proceed.
>no error messages.

Hmm.  Perhaps someone should define "mandatory" in this discussion since I 
had a different definition of the word.  My undestanding of a mandatory 
attribute one that is required to be supported or the exchange (or call) fails.


> >
> > As discovered in the L2TP saga, the message from IESG seems to be "Just Say
> > No" to Mandatory bits.
>
>... in vendor specific attributes...

In the case of standard attributes, there will be attributes that MUST be 
supported (mandatory) and attributes that MAY be supported (optional). In 
the former case, a mandatory bit isn't required.  In the latter case, what 
does a mandatory bit mean?

> >
> > Chip
>
>yves
> >
> > >PatC
> >
> > Support NetAid!  http://www.netaid.org
> > --------------------------------------------------
> > Chip Sharp                 Consulting Engineering
> > Cisco Systems              Telco Bio-region
> > Reality - Love it or Leave it.
> > --------------------------------------------------

Support NetAid!  http://www.netaid.org
--------------------------------------------------
Chip Sharp                 Consulting Engineering
Cisco Systems              Telco Bio-region
Reality - Love it or Leave it.			
--------------------------------------------------


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 14 16:00:07 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA12179;
	Thu, 14 Oct 1999 16:00:06 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) id PAA24864
	for nasreq-outgoing; Thu, 14 Oct 1999 15:23:31 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from baynet.baynetworks.com (ns1.BayNetworks.COM [134.177.3.20])
	by the-dread-pirate-roberts.rutgers.edu (8.8.8/8.8.8) with ESMTP id PAA24837
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 15:23:27 -0400 (EDT)
Received: from mailhost.BayNetworks.COM (h8754.s84f5.BayNetworks.COM [132.245.135.84])
	by baynet.baynetworks.com (8.9.1/8.9.1) with ESMTP id MAA23870
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 12:21:40 -0700 (PDT)
Received: from pobox.engeast.BayNetworks.COM (pobox.engeast.baynetworks.com [192.32.61.6])
	by mailhost.BayNetworks.COM (8.9.1/8.8.8) with ESMTP id PAA07364
	for <nasreq@tdmx.rutgers.edu>; Thu, 14 Oct 1999 15:18:54 -0400 (EDT)
Received: from ltrpluim.corpemea.baynetworks.com ([141.251.166.11])
	by pobox.engeast.BayNetworks.COM (SMI-8.6/BNET-97/04/24-S) with ESMTP
	id PAA26675; Thu, 14 Oct 1999 15:22:52 -0400
	for <nasreq@tdmx.rutgers.edu>
Received: (from rpluim@localhost)
	by ltrpluim.corpemea.baynetworks.com (8.8.7/8.8.7) id VAA31946;
	Thu, 14 Oct 1999 21:18:31 +0200
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <14342.11399.350059.853786@ltrpluim.corpemea.baynetworks.com>
Date: Thu, 14 Oct 1999 21:18:31 +0200 (CEST)
To: nasreq@tdmx.rutgers.edu
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
In-Reply-To: <078292D50C98D2118D090008C7E9C6A60225816B@STAY.platinum.corp.microsoft.com>
References: <078292D50C98D2118D090008C7E9C6A60225816B@STAY.platinum.corp.microsoft.com>
X-Mailer: VM 6.75 under 21.2  (beta19) "Shinjuku" XEmacs Lucid
X-Attribution: Robert
From: Robert Pluim <rpluim@bigfoot.com>
Reply-To: Robert Pluim <rpluim@bigfoot.com>
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Bernard Aboba writes:
 > >What happens if one has a proxy situation (e.g. over two aaa proxies,
 > >the middle RADIUS server belonging to a broker), where some specific
 > >attribute needs to be send end to end, but the next hop doesn't know it.
 > 
 > Interesting case. There are a couple of mechanisms being talked about
 > in the requirements. One is end-to-end security. This allows you
 > to indicate that the attribute in question cannot be edited and must be
 > passed along.

What if one of the two servers disagrees about whether a non-editable
attribute should be editable or not? Do you drop the packet? Or do you
remove the offending attribute? (I can think of at least one real-life
example here, involving multilink).

Robert
-- 


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Fri Oct 15 23:14:54 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id XAA26363;
	Fri, 15 Oct 1999 23:14:54 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id WAA03523 for nasreq-outgoing; Fri, 15 Oct 1999 22:58:11 -0400 (EDT)
Received: from pinky.rte.microsoft.com ([131.107.152.16]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id WAA03519 for <nasreq@tdmx.rutgers.edu>; Fri, 15 Oct 1999 22:58:09 -0400 (EDT)
Received: from localhost (gwz@localhost) by pinky.rte.microsoft.com (8.7.6/8.7.3) with SMTP id TAA21727; Fri, 15 Oct 1999 19:45:22 -0700 (PDT)
X-Authentication-Warning: pinky.rte.microsoft.com: gwz owned process doing -bs
Date: Fri, 15 Oct 1999 19:45:22 -0700 (PDT)
From: Glen Zorn <gwz@acm.org>
To: Randy Bush <randy@psg.com>
cc: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>,
        "Yves T'Joens" <yves.tjoens@alcatel.be>,
        nasreq <nasreq@tdmx.rutgers.edu>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
In-Reply-To: <E11bmc5-0003iy-00@rip.psg.com>
Message-ID: <Pine.BSF.3.96.991015193855.5367F-100000@pinky.rte.microsoft.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

On Thu, 14 Oct 1999, Randy Bush wrote:

> > I think that the reality of it all is that protocol extensibility via
> > vendor specific attributes works well intra-domain.
> 
> does this imply an assumption of intra-domain vendor homogeniety?  

Not necessarily; however, it does, I think, imply intra-domain knowledge
of vendors (which may be essentially the same thing).

>if so,
> and if by 'domain' your mean e.g. one isp.  this is certainly not the case
> in many very significant domains of which i am aware.
> 
> randy
> 



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Fri Oct 15 23:55:24 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id XAA26970;
	Fri, 15 Oct 1999 23:55:24 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id XAA03700 for nasreq-outgoing; Fri, 15 Oct 1999 23:39:43 -0400 (EDT)
Received: from NOD.RESTON.MCI.NET (nod.Reston.mci.net [166.45.6.38]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id XAA03696 for <nasreq@tdmx.rutgers.edu>; Fri, 15 Oct 1999 23:39:41 -0400 (EDT)
Received: from adsl-63-193-107-75.dsl.snfc21.pacbell.net ([166.45.3.11])
 by shoe.reston.mci.net (PMDF V5.2-32 #33823)
 with ESMTP id <01JH6JQ7ZXGO95MRZY@shoe.reston.mci.net> for
 nasreq@tdmx.rutgers.edu; Fri, 15 Oct 1999 23:39:40 EDT
Date: Fri, 15 Oct 1999 20:39:37 -0700
From: Paul Krumviede <paul.krumviede@wcom.com>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
In-reply-to: <Pine.BSF.3.96.991015193855.5367F-100000@pinky.rte.microsoft.com>
Originator-info: login-id=krumviede; server=localhost
To: Glen Zorn <gwz@acm.org>, Randy Bush <randy@psg.com>
Cc: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>,
        "Yves T'Joens" <yves.tjoens@alcatel.be>,
        nasreq <nasreq@tdmx.rutgers.edu>
Message-id: <3717107102.940019977@adsl-63-193-107-75.dsl.snfc21.pacbell.net>
MIME-version: 1.0
X-Mailer: Mulberry (Win32) [1.4.4, s/n P005-300888-001]
Content-type: text/plain; charset=us-ascii
Content-disposition: inline
Content-transfer-encoding: 7BIT
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7BIT

--On Friday, October 15, 1999, 7:45 PM -0700 Glen Zorn <gwz@acm.org> wrote:

> On Thu, 14 Oct 1999, Randy Bush wrote:
> 
>> > I think that the reality of it all is that protocol extensibility via
>> > vendor specific attributes works well intra-domain.
>> 
>> does this imply an assumption of intra-domain vendor homogeniety?  

it better not. acquisitions, particularly international ones, break any such
assumption.

> Not necessarily; however, it does, I think, imply intra-domain knowledge
> of vendors (which may be essentially the same thing).

even if i have such knowledge now, it can be overtaken by events.
of course, suitable choice of the scope of a domain can make this
true by definition, but that seems to be an unproductive exercise.

-paul



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Sat Oct 16 00:06:41 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id AAA27054;
	Sat, 16 Oct 1999 00:06:36 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id XAA03804 for nasreq-outgoing; Fri, 15 Oct 1999 23:54:11 -0400 (EDT)
Received: from pinky.rte.microsoft.com ([131.107.152.16]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id XAA03800 for <nasreq@tdmx.rutgers.edu>; Fri, 15 Oct 1999 23:54:10 -0400 (EDT)
Received: from localhost (gwz@localhost) by pinky.rte.microsoft.com (8.7.6/8.7.3) with SMTP id UAA21789; Fri, 15 Oct 1999 20:41:21 -0700 (PDT)
X-Authentication-Warning: pinky.rte.microsoft.com: gwz owned process doing -bs
Date: Fri, 15 Oct 1999 20:41:21 -0700 (PDT)
From: Glen Zorn <gwz@acm.org>
To: Paul Krumviede <paul.krumviede@wcom.com>
cc: Randy Bush <randy@psg.com>,
        "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>,
        "Yves T'Joens" <yves.tjoens@alcatel.be>,
        nasreq <nasreq@tdmx.rutgers.edu>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
In-Reply-To: <3717107102.940019977@adsl-63-193-107-75.dsl.snfc21.pacbell.net>
Message-ID: <Pine.BSF.3.96.991015203201.5367K-100000@pinky.rte.microsoft.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

On Fri, 15 Oct 1999, Paul Krumviede wrote:

> --On Friday, October 15, 1999, 7:45 PM -0700 Glen Zorn <gwz@acm.org> wrote:
> 
> > On Thu, 14 Oct 1999, Randy Bush wrote:
> > 
> >> > I think that the reality of it all is that protocol extensibility via
> >> > vendor specific attributes works well intra-domain.
> >> 
> >> does this imply an assumption of intra-domain vendor homogeniety?  
> 
> it better not. acquisitions, particularly international ones, break any such
> assumption.
> 
> > Not necessarily; however, it does, I think, imply intra-domain knowledge
> > of vendors (which may be essentially the same thing).
> 
> even if i have such knowledge now, it can be overtaken by events.

True, but (in theory, at least) the effects of such events are suspected,
(if not known) in advance. 

> of course, suitable choice of the scope of a domain can make this
> true by definition, but that seems to be an unproductive exercise.
> 

Precisely why we have standards, right?  

> -paul
> 
> 



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Sat Oct 16 00:59:15 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id AAA27807;
	Sat, 16 Oct 1999 00:59:14 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id AAA04208 for nasreq-outgoing; Sat, 16 Oct 1999 00:46:36 -0400 (EDT)
Received: from rip.psg.com (rip.psg.com [147.28.0.39]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id AAA04204 for <nasreq@tdmx.rutgers.edu>; Sat, 16 Oct 1999 00:46:34 -0400 (EDT)
Received: from randy by rip.psg.com with local (Exim 3.03 #1)
	id 11cLjl-0000Cs-00; Fri, 15 Oct 1999 21:46:33 -0700
From: Randy Bush <randy@psg.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: Paul Krumviede <paul.krumviede@wcom.com>
Cc: nasreq <nasreq@tdmx.rutgers.edu>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
References: <Pine.BSF.3.96.991015193855.5367F-100000@pinky.rte.microsoft.com>
	<3717107102.940019977@adsl-63-193-107-75.dsl.snfc21.pacbell.net>
Message-Id: <E11cLjl-0000Cs-00@rip.psg.com>
Date: Fri, 15 Oct 1999 21:46:33 -0700
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

>>>> I think that the reality of it all is that protocol extensibility via
>>>> vendor specific attributes works well intra-domain.
>>> does this imply an assumption of intra-domain vendor homogeniety?  
> it better not. acquisitions, particularly international ones, break any such
> assumption.
>> Not necessarily; however, it does, I think, imply intra-domain knowledge
>> of vendors (which may be essentially the same thing).
> even if i have such knowledge now, it can be overtaken by events.
> of course, suitable choice of the scope of a domain can make this
> true by definition, but that seems to be an unproductive exercise.

it may have been bernard who explained to me that intradomain proxy use is
really needed for facades to provide homogenity, i.e. hide a bunch of
incompatible stuff so our front-end systems don't have to know the kinks
of the 42 different vendors equipment we've acquired.  this makes sense.

but, in this case, we can have the differences under administrative
configuration control.  we can just say no to configuring an intermediate
proxy that will obscure what needs to be made visible.  i.e. this may not
be a problem that must be solved by protocol.

whatcha think?

randy


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Sat Oct 16 01:40:43 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id BAA01441;
	Sat, 16 Oct 1999 01:40:43 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id BAA04368 for nasreq-outgoing; Sat, 16 Oct 1999 01:27:08 -0400 (EDT)
Received: from NOD.RESTON.MCI.NET (nod.Reston.mci.net [166.45.6.38]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id BAA04364 for <nasreq@tdmx.rutgers.edu>; Sat, 16 Oct 1999 01:27:06 -0400 (EDT)
Received: from adsl-63-193-107-75.dsl.snfc21.pacbell.net ([166.45.3.11])
 by shoe.reston.mci.net (PMDF V5.2-32 #33823)
 with ESMTP id <01JH6NHFB3XO94DPGF@shoe.reston.mci.net> for
 nasreq@tdmx.rutgers.edu; Sat, 16 Oct 1999 01:27:06 EDT
Date: Fri, 15 Oct 1999 22:27:01 -0700
From: Paul Krumviede <paul.krumviede@wcom.com>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
In-reply-to: <E11cLjl-0000Cs-00@rip.psg.com>
Originator-info: login-id=krumviede; server=localhost
To: Randy Bush <randy@psg.com>
Cc: nasreq <nasreq@tdmx.rutgers.edu>
Message-id: <3723551452.940026421@adsl-63-193-107-75.dsl.snfc21.pacbell.net>
MIME-version: 1.0
X-Mailer: Mulberry (Win32) [1.4.4, s/n P005-300888-001]
Content-type: text/plain; charset=us-ascii
Content-disposition: inline
Content-transfer-encoding: 7BIT
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7BIT

--On Friday, October 15, 1999, 9:46 PM -0700 Randy Bush <randy@psg.com> wrote:

>>>>> I think that the reality of it all is that protocol extensibility via
>>>>> vendor specific attributes works well intra-domain.
>>>> does this imply an assumption of intra-domain vendor homogeniety?  
>> it better not. acquisitions, particularly international ones, break any such
>> assumption.
>>> Not necessarily; however, it does, I think, imply intra-domain knowledge
>>> of vendors (which may be essentially the same thing).
>> even if i have such knowledge now, it can be overtaken by events.
>> of course, suitable choice of the scope of a domain can make this
>> true by definition, but that seems to be an unproductive exercise.
> 
> it may have been bernard who explained to me that intradomain proxy use is
> really needed for facades to provide homogenity, i.e. hide a bunch of
> incompatible stuff so our front-end systems don't have to know the kinks
> of the 42 different vendors equipment we've acquired.  this makes sense.

oh, i agree that some mechanism, such as intradomain proxies, is
needed here. but it isn't always easy to hide equipment differences
for more complex services that require vendor-specific attributes be
delivered to the end box(es), such as a NAS.  Now where this vendor
specific stuff gets inserted into the response(s) going back to the
requestor is the interesting, and hard, part.  It is currently hard to
say "this caller needs a secure tunnel to gateway X" in a vendor
neutral fashion (as least as far as I know it is; I wouldn't mind being
proven wrong).

> but, in this case, we can have the differences under administrative
> configuration control.  we can just say no to configuring an intermediate
> proxy that will obscure what needs to be made visible.  i.e. this may not
> be a problem that must be solved by protocol.
> 
> whatcha think?

This is what I recommend where possible. As long as the
protocol(s) allow me to transport the right level of abstraction
to specify the services I want to offer (now and in the future)
I can probably make this work and push the gory detail
into the places that need to know it.

One reason for commenting on this thread is that I'm not so
sure that vendor specific attributes work all that well in more
complex intradomain situations. Administrative scoping
within a domain, with something that can do the right
intrepretation at boundaries, seems like a very reasonable
approach, but I need a reasonable way to communicate
the necessary service configuration info, and this seems to
impose requirements on the data represenation.

One reason this whole discussion can be messy is
that the scope of a domain could be considered to
align with a business (e.g. MCI WorldCom) or with
some finer granularity (e.g. UUnet and vBNS) within
the business.

-paul



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Sat Oct 16 03:09:27 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA09253;
	Sat, 16 Oct 1999 03:09:27 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id CAA05048 for nasreq-outgoing; Sat, 16 Oct 1999 02:53:18 -0400 (EDT)
Received: from dfssl.exchange.microsoft.com (dfssl.exchange.microsoft.com [131.107.88.59]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id CAA05044 for <nasreq@tdmx.rutgers.edu>; Sat, 16 Oct 1999 02:53:16 -0400 (EDT)
Received: by dfssl with Internet Mail Service (5.5.2650.21)
	id <48A7PSMF>; Fri, 15 Oct 1999 23:53:12 -0700
Message-ID: <078292D50C98D2118D090008C7E9C6A602258183@STAY.platinum.corp.microsoft.com>
From: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
To: "'Randy Bush'" <randy@psg.com>, Paul Krumviede <paul.krumviede@wcom.com>
Cc: nasreq <nasreq@tdmx.rutgers.edu>
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
Date: Fri, 15 Oct 1999 23:53:11 -0700
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

>hide a bunch of incompatible stuff so our front-end systems don't 
>have to know the kinks of the 42 different vendors equipment 
>we've acquired.  this makes sense.

Well, that is one reason proxies have been used, yes. But once
again, it is really a response to a protocol inadequacy. The
objective of capabilities negotiation should be to remove the
need to deploy a proxy for this purpose, making it possible 
for servers to talk to NASes (or local proxies) without the
need for intermediaries. 


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Sat Oct 16 13:07:20 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA13003;
	Sat, 16 Oct 1999 13:07:19 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id MAA07380 for nasreq-outgoing; Sat, 16 Oct 1999 12:51:13 -0400 (EDT)
Received: from lobster.baynetworks.com (ns3.BayNetworks.COM [192.32.253.3]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id MAA07375 for <nasreq@tdmx.rutgers.edu>; Sat, 16 Oct 1999 12:51:12 -0400 (EDT)
Received: from mailhost.BayNetworks.COM (h8754.s84f5.BayNetworks.COM [132.245.135.84])
	by lobster.baynetworks.com (8.9.1/8.9.1) with ESMTP id MAA21198;
	Sat, 16 Oct 1999 12:54:01 -0400 (EDT)
Received: from pobox.engeast.BayNetworks.COM (pobox.engeast.baynetworks.com [192.32.61.6])
	by mailhost.BayNetworks.COM (8.9.1/8.8.8) with ESMTP id MAA08072;
	Sat, 16 Oct 1999 12:46:38 -0400 (EDT)
Received: from ltrpluim.corpemea.baynetworks.com ([141.251.166.11])
	by pobox.engeast.BayNetworks.COM (SMI-8.6/BNET-97/04/24-S) with ESMTP
	id MAA28183; Sat, 16 Oct 1999 12:50:31 -0400
	for 
Received: (from rpluim@localhost)
	by ltrpluim.corpemea.baynetworks.com (8.8.7/8.8.7) id SAA00339;
	Sat, 16 Oct 1999 18:45:54 +0200
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <14344.43965.133608.179173@ltrpluim.corpemea.baynetworks.com>
Date: Sat, 16 Oct 1999 18:45:49 +0200 (CEST)
To: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
Cc: nasreq@tdmx.rutgers.edu
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
In-Reply-To: <078292D50C98D2118D090008C7E9C6A602258177@STAY.platinum.corp.microsoft.com>
References: <078292D50C98D2118D090008C7E9C6A602258177@STAY.platinum.corp.microsoft.com>
X-Mailer: VM 6.75 under 21.2  (beta19) "Shinjuku" XEmacs Lucid
X-Attribution: Robert
From: Robert Pluim <rpluim@bigfoot.com>
Reply-To: Robert Pluim <rpluim@bigfoot.com>
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Bernard Aboba writes:
 > >What if one of the two servers disagrees about whether a non-editable
 > >attribute should be editable or not? Do you drop the packet? Or do you
 > >remove the offending attribute? (I can think of at least one real-life
 > >example here, involving multilink).
 > 
 > Can you describe the real-life example?

(I already discussed this with Randy Bush, and we decided it was more
applicable to aaa than nasreq, but here goes anyway).

Suppose someone is dialing in to an ISP that is not his home one. The
AAA request for him goes to his home ISP via a broker. There is a
contract in place between the two ISPs, but it does not cover
multink. The home server returns an accept with a Port-Limit of 2,
say, and marks it as non-editable (otherwise someone intermediary could
increase it). The intermediate broker wants to change this Port-Limit
to 1, since it has no contract covering multink in place, but it
can't, since the attribute is non-editable.


-- 
Robert Pluim                               Voice:  +33 4 92 96 17 43
Systems Development Engineer                 Fax:  +33 4 92 96 15 32
EMEA Data R&D
Nortel Networks               <URL:mailto:rpluim@nortelnetworks.com>


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Sun Oct 17 03:26:09 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA02821;
	Sun, 17 Oct 1999 03:26:09 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id DAA10460 for nasreq-outgoing; Sun, 17 Oct 1999 03:13:07 -0400 (EDT)
Received: from rip.psg.com (rip.psg.com [147.28.0.39]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id DAA10456 for <nasreq@tdmx.rutgers.edu>; Sun, 17 Oct 1999 03:13:06 -0400 (EDT)
Received: from randy by rip.psg.com with local (Exim 3.03 #1)
	id 11ckV1-0000nk-00; Sun, 17 Oct 1999 00:12:59 -0700
From: Randy Bush <randy@psg.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: Robert Pluim <rpluim@bigfoot.com>
Cc: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>,
        nasreq@tdmx.rutgers.edu
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
References: <078292D50C98D2118D090008C7E9C6A602258177@STAY.platinum.corp.microsoft.com>
	<14344.43965.133608.179173@ltrpluim.corpemea.baynetworks.com>
Message-Id: <E11ckV1-0000nk-00@rip.psg.com>
Date: Sun, 17 Oct 1999 00:12:59 -0700
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

> Suppose someone is dialing in to an ISP that is not his home one. The
> AAA request for him goes to his home ISP via a broker. There is a
> contract in place between the two ISPs, but it does not cover
> multink. The home server returns an accept with a Port-Limit of 2,
> say, and marks it as non-editable (otherwise someone intermediary could
> increase it). The intermediate broker wants to change this Port-Limit
> to 1, since it has no contract covering multink in place, but it
> can't, since the attribute is non-editable.

why does the broker care?  the two end systems make the decisions and
the broker merely has to help the client find the authentication and
authorization server.

randy


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Sun Oct 17 10:06:25 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA08353;
	Sun, 17 Oct 1999 10:06:24 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id JAA11868 for nasreq-outgoing; Sun, 17 Oct 1999 09:52:52 -0400 (EDT)
Received: from dfssl.exchange.microsoft.com (dfssl.exchange.microsoft.com [131.107.88.59]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id JAA11864 for <nasreq@tdmx.rutgers.edu>; Sun, 17 Oct 1999 09:52:50 -0400 (EDT)
Received: by dfssl with Internet Mail Service (5.5.2650.21)
	id <48A7PZ0K>; Sun, 17 Oct 1999 06:52:39 -0700
Message-ID: <078292D50C98D2118D090008C7E9C6A602258187@STAY.platinum.corp.microsoft.com>
From: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
To: "'Randy Bush'" <randy@psg.com>, Robert Pluim <rpluim@bigfoot.com>
Cc: nasreq@tdmx.rutgers.edu
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
Date: Sun, 17 Oct 1999 06:52:38 -0700
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="windows-1252"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

>> Suppose someone is dialing in to an ISP that is not his home one. The
>> AAA request for him goes to his home ISP via a broker. There is a
>> contract in place between the two ISPs, but it does not cover
>> multink. The home server returns an accept with a Port-Limit of 2,
>> say, and marks it as non-editable (otherwise someone intermediary could
>> increase it). The intermediate broker wants to change this Port-Limit
>> to 1, since it has no contract covering multink in place, but it
>> can't, since the attribute is non-editable.

>why does the broker care?  the two end systems make the decisions and
>the broker merely has to help the client find the authentication and
>authorization server.

Yes, I think that the issue is not really the broker here, it is that
the server and client have a way of coming to agreement on the values
of parameters. Today in RADIUS there are no error messages, and thus
no way for the NAS to say "sorry, can't do Port-Limit=2." Note that
this is a distinct problem from the ability to understand the
attribute, since it is the *value* that is unacceptable, not the
attribute itself.


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Sun Oct 17 11:18:32 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA08775;
	Sun, 17 Oct 1999 11:18:32 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id LAA12106 for nasreq-outgoing; Sun, 17 Oct 1999 11:02:09 -0400 (EDT)
Received: from rip.psg.com (rip.psg.com [147.28.0.39]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id LAA12102 for <nasreq@tdmx.rutgers.edu>; Sun, 17 Oct 1999 11:02:08 -0400 (EDT)
Received: from randy by rip.psg.com with local (Exim 3.03 #1)
	id 11crox-0004vO-00; Sun, 17 Oct 1999 08:02:03 -0700
From: Randy Bush <randy@psg.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: Bernard Aboba <bernarda@exchange.microsoft.com>
Cc: nasreq@tdmx.rutgers.edu
Subject: RE: comments on draft-ietf-nasreq-criteria-03.txt
References: <078292D50C98D2118D090008C7E9C6A602258187@STAY.platinum.corp.microsoft.com>
Message-Id: <E11crox-0004vO-00@rip.psg.com>
Date: Sun, 17 Oct 1999 08:02:03 -0700
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

>>> Suppose someone is dialing in to an ISP that is not his home one. The
>>> AAA request for him goes to his home ISP via a broker. There is a
>>> contract in place between the two ISPs, but it does not cover
>>> multink. The home server returns an accept with a Port-Limit of 2,
>>> say, and marks it as non-editable (otherwise someone intermediary could
>>> increase it). The intermediate broker wants to change this Port-Limit
>>> to 1, since it has no contract covering multink in place, but it
>>> can't, since the attribute is non-editable.
>> why does the broker care?  the two end systems make the decisions and
>> the broker merely has to help the client find the authentication and
>> authorization server.
> Yes, I think that the issue is not really the broker here, it is that
> the server and client have a way of coming to agreement on the values
> of parameters. Today in RADIUS there are no error messages, and thus
> no way for the NAS to say "sorry, can't do Port-Limit=2." Note that
> this is a distinct problem from the ability to understand the
> attribute, since it is the *value* that is unacceptable, not the
> attribute itself.

if broker != proxy, this kind of stuff becomes irrelevant, yes?

randy


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Sun Oct 17 22:21:08 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id WAA12610;
	Sun, 17 Oct 1999 22:21:06 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id WAA14293 for nasreq-outgoing; Sun, 17 Oct 1999 22:01:33 -0400 (EDT)
Received: from home.merit.edu (home.merit.edu [198.108.60.42]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id WAA14289 for <nasreq@tdmx.rutgers.edu>; Sun, 17 Oct 1999 22:01:31 -0400 (EDT)
Received: from merit.edu (pm457-05.dialip.mich.net [198.110.21.111]) by home.merit.edu (8.8.8/merit-2.0) with ESMTP id WAA25071; Sun, 17 Oct 1999 22:01:26 -0400 (EDT)
Message-ID: <380A93A4.6F614137@merit.edu>
Date: Sun, 17 Oct 1999 23:27:33 -0400
From: John Vollbrecht <jrv@merit.edu>
X-Mailer: Mozilla 4.6 [en] (Win95; U)
X-Accept-Language: en
MIME-Version: 1.0
To: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
CC: "'Randy Bush'" <randy@psg.com>, Robert Pluim <rpluim@bigfoot.com>,
        nasreq@tdmx.rutgers.edu
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
References: <078292D50C98D2118D090008C7E9C6A602258187@STAY.platinum.corp.microsoft.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit



"Bernard Aboba (Exchange)" wrote:

> >> Suppose someone is dialing in to an ISP that is not his home one. The
> >> AAA request for him goes to his home ISP via a broker. There is a
> >> contract in place between the two ISPs, but it does not cover
> >> multink. The home server returns an accept with a Port-Limit of 2,
> >> say, and marks it as non-editable (otherwise someone intermediary could
> >> increase it). The intermediate broker wants to change this Port-Limit
> >> to 1, since it has no contract covering multink in place, but it
> >> can't, since the attribute is non-editable.
>
> >why does the broker care?  the two end systems make the decisions and
> >the broker merely has to help the client find the authentication and
> >authorization server.
>
> Yes, I think that the issue is not really the broker here, it is that
> the server and client have a way of coming to agreement on the values
> of parameters. Today in RADIUS there are no error messages, and thus
> no way for the NAS to say "sorry, can't do Port-Limit=2." Note that
> this is a distinct problem from the ability to understand the
> attribute, since it is the *value* that is unacceptable, not the
> attribute itself.

Seems to me that there are several issues here.

There is the negotiation issue -- does the protocol need to be able to allow
the nas or an intermediate broker say - "you can have one port, not two - will
you take a connection with one port?"

Another issue is the semantics question - what if a NAS vendor comes up with a
new feature and wants to let the user specify how it is to be configured,  how
does the vendor let the user do that without an inordinate amount of
beaurocratic hassle?  The Vendor specific attributes are a way to do that.
Other ways have been suggested.

Then there is the issue that the broker might want to offer a generic service
with attributes different than what is provided by the NAS.  For example it
might want to include a "huntgroup" attribute or a "charge class" attribute.
Such a broker might also want to have a single set of attributes which it
translates into appropriate NAS specific attributes.  The common example for
this is the filter setting - which is implemented differently by NAS Vendors.
A broker might have its own way of allowing users to specify a filter and then
translate that into what is needed by each NAS that it supports.

Then there is the issue that has been discussed a lot on this list already --
how do you handle the situation where one server does not understand an
attribute sent by another.  This could be handled in a number of ways. 1) one
way is to outlaw this by requiring all servers to negotiate with each other
about which attributes it understands, and sending anything it doesn't
understand is cause for dropping its association with the other server.  2)
another approach is to have a "mandatory" bit which is set by a sender if it
expects the receiver to understand the attribute.  The request is refused if
the receiver does not understand.  3) Another approach is to have the receiver
return a "conditional" acceptance with the caveat that it didn't understand
what the attribute meant, so if it was important the sender should not allow
the acceptance to go through.   This is a complicated situation, made even
more complicated by having the possibility of intermediate brokers between NAS
and user server.  I think this is a very important issue for the ng aaa
protocol, and any proposed solution needs to be well thought out and have well
documented the way it will work for a varitey of inter-organization AAA
environments.

John.








From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Mon Oct 18 17:47:57 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA19963;
	Mon, 18 Oct 1999 17:47:52 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id RAA20644 for nasreq-outgoing; Mon, 18 Oct 1999 17:18:54 -0400 (EDT)
Received: from pinky.rte.microsoft.com ([131.107.152.16]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id RAA20640 for <nasreq@tdmx.rutgers.edu>; Mon, 18 Oct 1999 17:18:52 -0400 (EDT)
Received: from localhost (gwz@localhost) by pinky.rte.microsoft.com (8.7.6/8.7.3) with SMTP id OAA26924; Mon, 18 Oct 1999 14:06:24 -0700 (PDT)
X-Authentication-Warning: pinky.rte.microsoft.com: gwz owned process doing -bs
Date: Mon, 18 Oct 1999 14:06:24 -0700 (PDT)
From: Glen Zorn <gwz@acm.org>
To: Paul Krumviede <paul.krumviede@wcom.com>
cc: Randy Bush <randy@psg.com>, nasreq <nasreq@tdmx.rutgers.edu>
Subject: Re: comments on draft-ietf-nasreq-criteria-03.txt
In-Reply-To: <3723551452.940026421@adsl-63-193-107-75.dsl.snfc21.pacbell.net>
Message-ID: <Pine.BSF.3.96.991018140218.5367L-100000@pinky.rte.microsoft.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

On Fri, 15 Oct 1999, Paul Krumviede wrote:

> --On Friday, October 15, 1999, 9:46 PM -0700 Randy Bush <randy@psg.com> wrote:
> 
> >>>>> I think that the reality of it all is that protocol extensibility via
> >>>>> vendor specific attributes works well intra-domain.
> >>>> does this imply an assumption of intra-domain vendor homogeniety?  
> >> it better not. acquisitions, particularly international ones, break any such
> >> assumption.
> >>> Not necessarily; however, it does, I think, imply intra-domain knowledge
> >>> of vendors (which may be essentially the same thing).
> >> even if i have such knowledge now, it can be overtaken by events.
> >> of course, suitable choice of the scope of a domain can make this
> >> true by definition, but that seems to be an unproductive exercise.
> > 
> > it may have been bernard who explained to me that intradomain proxy use is
> > really needed for facades to provide homogenity, i.e. hide a bunch of
> > incompatible stuff so our front-end systems don't have to know the kinks
> > of the 42 different vendors equipment we've acquired.  this makes sense.
> 
> oh, i agree that some mechanism, such as intradomain proxies, is
> needed here. but it isn't always easy to hide equipment differences
> for more complex services that require vendor-specific attributes be
> delivered to the end box(es), such as a NAS.  Now where this vendor
> specific stuff gets inserted into the response(s) going back to the
> requestor is the interesting, and hard, part.  

I have long been of the opinion that this stuff should be inserted at the
proxy closest to the NAS.

> It is currently hard to
> say "this caller needs a secure tunnel to gateway X" in a vendor
> neutral fashion (as least as far as I know it is; I wouldn't mind being
> proven wrong).
> 

I don't think it's _that_ hard; see draft-ietf-radius-tunnel-auth-09.txt
(delayed for many months, apparently for political reasons).

> > but, in this case, we can have the differences under administrative
> > configuration control.  we can just say no to configuring an intermediate
> > proxy that will obscure what needs to be made visible.  i.e. this may not
> > be a problem that must be solved by protocol.
> > 
> > whatcha think?
> 
> This is what I recommend where possible. As long as the
> protocol(s) allow me to transport the right level of abstraction
> to specify the services I want to offer (now and in the future)
> I can probably make this work and push the gory detail
> into the places that need to know it.
> 
> One reason for commenting on this thread is that I'm not so
> sure that vendor specific attributes work all that well in more
> complex intradomain situations. Administrative scoping
> within a domain, with something that can do the right
> intrepretation at boundaries, seems like a very reasonable
> approach, but I need a reasonable way to communicate
> the necessary service configuration info, and this seems to
> impose requirements on the data represenation.
> 
> One reason this whole discussion can be messy is
> that the scope of a domain could be considered to
> align with a business (e.g. MCI WorldCom) or with
> some finer granularity (e.g. UUnet and vBNS) within
> the business.
> 
> -paul
> 
> 



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Tue Oct 19 08:40:32 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id IAA12804;
	Tue, 19 Oct 1999 08:40:31 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id IAA24413 for nasreq-outgoing; Tue, 19 Oct 1999 08:20:08 -0400 (EDT)
Received: from btm4r4.alcatel.be (btm4r4.alcatel.be [195.207.101.110]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id IAA24409 for <nasreq@tdmx.rutgers.edu>; Tue, 19 Oct 1999 08:19:58 -0400 (EDT)
Received: from btmq9s.rc.bel.alcatel.be (btmq9s.rc.bel.alcatel.be [138.203.65.182])
	by btm4r4.alcatel.be (8.9.1a/8.9.1) with ESMTP id OAA23563
	for <nasreq@tdmx.rutgers.edu>; Tue, 19 Oct 1999 14:19:10 +0200 (MET DST)
Received: from alcatel.be (btm0l8 [138.203.65.218])
	by btmq9s.rc.bel.alcatel.be (8.8.8+Sun/8.8.8) with ESMTP id OAA00232
	for <nasreq@tdmx.rutgers.edu>; Tue, 19 Oct 1999 14:19:59 +0200 (MET DST)
Message-ID: <380C61A7.D868ED24@alcatel.be>
Date: Tue, 19 Oct 1999 14:18:47 +0200
From: "Yves T'Joens" <yves.tjoens@alcatel.be>
Organization: Alcatel
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
X-Accept-Language: en
MIME-Version: 1.0
To: nasreq <nasreq@tdmx.rutgers.edu>
Subject: nasreq attribute handling
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Hi Nasreq,

I started off some mails back with a discussion on handling of
attributes/messages in a proxy situation, however during the discussion,
the topic diverged to handling of vendor specific attributes.

Here is a set of requirements that could be included in the nasreq
document, specific to proxy handling in an inter-domain scenario. Note
that this is irrespective of the vendor specific or standardized nature
of the attribute/message.

Handling of attributes/ messages
--------------------------------

The protocol MUST allow the sending implementation to indicate to the
next hop in the proxy chain to behave as follows on the receipt of an
unknown attribute/message 
(a) silently discard and proceed
(b) interrupt the session associated with the sending of this
attribute/message
(c) send back an error message and proceed unmodified
(d) forward the attribute/message unmodified to the next hop (no
semantic interpretation required)

The protocol MUST allow the sending implementation to indicate to the
end system in the chain to behave as follows on the receipt of an
unknown attribute/message
(a) silently discard and proceed
(b) interrup the session associated with the sending of this
attribute/message
(c) send back an error message and proceed unmodified

Yves


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Tue Oct 19 11:23:14 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA21653;
	Tue, 19 Oct 1999 11:23:13 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id KAA25322 for nasreq-outgoing; Tue, 19 Oct 1999 10:57:48 -0400 (EDT)
Received: from gateway.metrored.com.ar (gateway.metrored.com.ar [200.47.31.243] (may be forged)) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id KAA25318 for <nasreq@tdmx.rutgers.edu>; Tue, 19 Oct 1999 10:57:41 -0400 (EDT)
Received: by GATEWAY with Internet Mail Service (5.5.2448.0)
	id <4951GZFC>; Tue, 19 Oct 1999 12:01:34 -0300
Message-ID: <FA1A6476E9A7D211BAAB002035E7BE0B09A4C2@EXCHANGE>
From: =?iso-8859-1?Q?=22Bugallo=2C_Hern=E1n=22?=
	 <bugalloh@metrored.com.ar>
To: "'Yves T'Joens'" <yves.tjoens@alcatel.be>,
        "'nasreq'"
	 <nasreq@tdmx.rutgers.edu>
Subject: nasreq PROXY AAA SERVER Request information
Date: Tue, 19 Oct 1999 11:57:35 -0300
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by ietf.org id LAA21653

Hello:
They could you say to me, how an AAA server like RADIUS works, in a proxy
situation, or in another case, could you tellme where I can find this
information.
Thanks you.
                
                        _\\|//_
                        ( O-O )
 ~~~~~~~~~~~~~~~~~~~~o00~~(_)~~00o~~~~~~~~~~~~~~~~~~~~
	Hernán Bugallo.
	Tecnologia & Networking.
	MetroRED Telecomunicaciones
	Paseo Colon 505-5º Piso (Buenos Aires)-ARGENTINA
	C.P.:(1063). Tel: +(5411) +4344-7700 +(7829)
                     oooO
 ~~~~~~~~~~~~~~~~~~~~(   )~~~Oooo~~~~~~~~~~~~~~~~~~~~~~
                      \ (    (   )
                       \_)    ) /
                             (_/




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Tue Oct 19 11:54:54 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA23519;
	Tue, 19 Oct 1999 11:54:53 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id LAA25455 for nasreq-outgoing; Tue, 19 Oct 1999 11:25:31 -0400 (EDT)
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id LAA25451 for <nasreq@tdmx.rutgers.edu>; Tue, 19 Oct 1999 11:25:29 -0400 (EDT)
Received: from engmail2.Eng.Sun.COM ([129.146.1.25])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id IAA29869;
	Tue, 19 Oct 1999 08:24:51 -0700 (PDT)
Received: from ha1mpk-mail.eng.sun.com (phys-ha1mpkb.Eng.Sun.COM [129.146.93.35])
	by engmail2.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id IAA02876;
	Tue, 19 Oct 1999 08:24:50 -0700 (PDT)
Received: from mordor by ha1mpk-mail.eng.sun.com (SMI-8.6/SMI-SVR4)
	id IAA01383; Tue, 19 Oct 1999 08:24:49 -0700
Date: Tue, 19 Oct 1999 08:24:49 -0700 (PDT)
From: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>
Reply-To: "pcalhoun@eng.sun.com" <Pat.Calhoun@eng.sun.com>
Subject: Re: nasreq PROXY AAA SERVER Request information
To: bugalloh@metrored.com.ar
Cc: "'Yves T'Joens'" <yves.tjoens@alcatel.be>,
        "'nasreq'" <nasreq@tdmx.rutgers.edu>
In-Reply-To: "Your message with ID" <FA1A6476E9A7D211BAAB002035E7BE0B09A4C2@EXCHANGE>
Message-ID: <Roam.SIMC.2.0.6.940346689.24085.pcalhoun@ha1mpk-mail>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=iso-8859-1
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from QUOTED-PRINTABLE to 8bit by ietf.org id LAA23519

ftp://ftp.isi.edu/in-notes/rfc2607.txt

PatC
> Hello:
> They could you say to me, how an AAA server like RADIUS works, in a proxy
> situation, or in another case, could you tellme where I can find this
> information.
> Thanks you.
>                 
>                         _\\|//_
>                         ( O-O )
>  ~~~~~~~~~~~~~~~~~~~~o00~~(_)~~00o~~~~~~~~~~~~~~~~~~~~
> 	Hernán Bugallo.
> 	Tecnologia & Networking.
> 	MetroRED Telecomunicaciones
> 	Paseo Colon 505-5º Piso (Buenos Aires)-ARGENTINA
> 	C.P.:(1063). Tel: +(5411) +4344-7700 +(7829)
>                      oooO
>  ~~~~~~~~~~~~~~~~~~~~(   )~~~Oooo~~~~~~~~~~~~~~~~~~~~~~
>                       \ (    (   )
>                        \_)    ) /
>                              (_/
> 
> 




From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 20 03:56:41 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA21246;
	Wed, 20 Oct 1999 03:56:40 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id DAA29944 for nasreq-outgoing; Wed, 20 Oct 1999 03:39:21 -0400 (EDT)
Received: from scallop.baynetworks.com (ns5.baynetworks.com [194.133.90.101]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id DAA29940 for <nasreq@tdmx.rutgers.edu>; Wed, 20 Oct 1999 03:39:19 -0400 (EDT)
Received: from mailhost.BayNetworks.COM (h8754.s84f5.BayNetworks.COM [132.245.135.84])
	by scallop.baynetworks.com (8.9.1/8.9.1) with ESMTP id JAA10732
	for <nasreq@tdmx.rutgers.edu>; Wed, 20 Oct 1999 09:41:49 +0200 (MET DST)
Received: from pobox.engeast.BayNetworks.COM (pobox.engeast.baynetworks.com [192.32.61.6])
	by mailhost.BayNetworks.COM (8.9.1/8.8.8) with ESMTP id DAA20198
	for <nasreq@tdmx.rutgers.edu>; Wed, 20 Oct 1999 03:35:07 -0400 (EDT)
Received: from ltrpluim.corpemea.baynetworks.com ([141.251.166.11])
	by pobox.engeast.BayNetworks.COM (SMI-8.6/BNET-97/04/24-S) with ESMTP
	id DAA16069; Wed, 20 Oct 1999 03:39:05 -0400
	for <nasreq@tdmx.rutgers.edu>
Received: (from rpluim@localhost)
	by ltrpluim.corpemea.baynetworks.com (8.8.7/8.8.7) id JAA02586;
	Wed, 20 Oct 1999 09:34:16 +0200
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <14349.28788.240535.140448@ltrpluim.corpemea.baynetworks.com>
Date: Wed, 20 Oct 1999 09:34:12 +0200 (CEST)
To: nasreq <nasreq@tdmx.rutgers.edu>
Subject: Re: nasreq attribute handling
In-Reply-To: <380C61A7.D868ED24@alcatel.be>
References: <380C61A7.D868ED24@alcatel.be>
X-Mailer: VM 6.75 under 21.2  (beta19) "Shinjuku" XEmacs Lucid
X-Attribution: Robert
From: Robert Pluim <rpluim@bigfoot.com>
Reply-To: Robert Pluim <rpluim@bigfoot.com>
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Yves T'Joens writes:
 > Hi Nasreq,
 > 
 > I started off some mails back with a discussion on handling of
 > attributes/messages in a proxy situation, however during the discussion,
 > the topic diverged to handling of vendor specific attributes.
 > 
 > Here is a set of requirements that could be included in the nasreq
 > document, specific to proxy handling in an inter-domain scenario. Note
 > that this is irrespective of the vendor specific or standardized nature
 > of the attribute/message.
 > 
 > Handling of attributes/ messages
 > --------------------------------
 > 
 > The protocol MUST allow the sending implementation to indicate to the
 > next hop in the proxy chain to behave as follows on the receipt of an
 > unknown attribute/message 
 > (a) silently discard and proceed
 > (b) interrupt the session associated with the sending of this
 > attribute/message

?? What's the difference between (a) and (b). Discarding implies
interruption.

 > (c) send back an error message and proceed unmodified
 > (d) forward the attribute/message unmodified to the next hop (no
 > semantic interpretation required)
 >

If I were even a mildly paranoid ISP, I would never allow (c) and
scream at anyone that implemented (d). Who knows what that
attribute/message could be controlling?

Robert

-- 


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Wed Oct 20 11:54:10 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA09013;
	Wed, 20 Oct 1999 11:54:10 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id LAA01728 for nasreq-outgoing; Wed, 20 Oct 1999 11:25:35 -0400 (EDT)
Received: from smtpgw01.wanmail.net (smtpgw01.wanmail.net [209.154.194.62]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id LAA01724 for <nasreq@tdmx.rutgers.edu>; Wed, 20 Oct 1999 11:25:29 -0400 (EDT)
Received: from wcomexch1.wcom.net (wcomexch1.wcom.net [209.154.236.12])
	by smtpgw01.wanmail.net (8.9.1/8.9.1) with ESMTP id LAA00165
	for <nasreq@tdmx.rutgers.edu>; Wed, 20 Oct 1999 11:25:29 -0400 (EDT)
Received: by wcomexch1.wcom.net with Internet Mail Service (5.5.2448.0)
	id <49W508P9>; Wed, 20 Oct 1999 11:24:18 -0400
Message-ID: <9A74C89D6696D2118E500008C7BA7C560C03B21D@wcomexch1.wcom.net>
From: "Beadles, Mark A." <MBeadles@wcom.net>
To: "Nasreqng List (E-mail)" <nasreq@tdmx.rutgers.edu>
Subject: nasreq New address information
Date: Wed, 20 Oct 1999 11:24:16 -0400
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)
Content-Type: text/plain;
	charset="iso-8859-1"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

NASREQ WG members:

Effective immediately, please use the e-mail address
mailto:mark.beadles@columbus.rr.com for all correspondence related to the
WG.  Thanks!

+     Mark Anthony Beadles     + Co-chair, IETF NASREQ WG   + 
+ mark.beadles@columbus.rr.com + http://nasreq.rutgers.edu/ +
+      Voice 614.723.1941      +      Fax 614.723.8407      +





From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 21 03:39:34 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA12232;
	Thu, 21 Oct 1999 03:39:33 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id DAA18793 for nasreq-outgoing; Thu, 21 Oct 1999 03:15:25 -0400 (EDT)
Received: from btm4r4.alcatel.be (btm4r4.alcatel.be [195.207.101.110]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id DAA18789 for <nasreq@tdmx.rutgers.edu>; Thu, 21 Oct 1999 03:15:23 -0400 (EDT)
Received: from btmq9s.rc.bel.alcatel.be (btmq9s.rc.bel.alcatel.be [138.203.65.182])
	by btm4r4.alcatel.be (8.9.1a/8.9.1) with ESMTP id JAA27062;
	Thu, 21 Oct 1999 09:15:04 +0200 (MET DST)
Received: from alcatel.be (btm0l8 [138.203.65.218])
	by btmq9s.rc.bel.alcatel.be (8.8.8+Sun/8.8.8) with ESMTP id JAA01637;
	Thu, 21 Oct 1999 09:16:04 +0200 (MET DST)
Message-ID: <380EBD68.509EAEE3@alcatel.be>
Date: Thu, 21 Oct 1999 09:14:48 +0200
From: "Yves T'Joens" <yves.tjoens@alcatel.be>
Organization: Alcatel
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
X-Accept-Language: en
MIME-Version: 1.0
To: Robert Pluim <rpluim@bigfoot.com>
CC: nasreq <nasreq@tdmx.rutgers.edu>
Subject: Re: nasreq attribute handling
References: <380C61A7.D868ED24@alcatel.be> <14349.28788.240535.140448@ltrpluim.corpemea.baynetworks.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Robert Pluim wrote:
> 
> Yves T'Joens writes:
>  > Hi Nasreq,
>  >
>  > I started off some mails back with a discussion on handling of
>  > attributes/messages in a proxy situation, however during the discussion,
>  > the topic diverged to handling of vendor specific attributes.
>  >
>  > Here is a set of requirements that could be included in the nasreq
>  > document, specific to proxy handling in an inter-domain scenario. Note
>  > that this is irrespective of the vendor specific or standardized nature
>  > of the attribute/message.
>  >
>  > Handling of attributes/ messages
>  > --------------------------------
>  >
>  > The protocol MUST allow the sending implementation to indicate to the
>  > next hop in the proxy chain to behave as follows on the receipt of an
>  > unknown attribute/message
>  > (a) silently discard and proceed
>  > (b) interrupt the session associated with the sending of this
>  > attribute/message
> 
> ?? What's the difference between (a) and (b). Discarding implies
> interruption.

I would not say so. Some AVPs may just be discarded, while the session
will survive, only with deteriorated service handling. This is in my
opinion how most of the vendor specific additions to any base protocol
should be handled.

> 
>  > (c) send back an error message and proceed unmodified
>  > (d) forward the attribute/message unmodified to the next hop (no
>  > semantic interpretation required)
>  >
> 
> If I were even a mildly paranoid ISP, I would never allow (c) and

well, (c) would only allow some limited debugging, however this req. may
be dropped, I don't care too much...

> scream at anyone that implemented (d). Who knows what that
> attribute/message could be controlling?
> 

Given a single administrative domain, and a proxy situation. Trust of
operation is not necessarily an issue.

Assume the NAS to be of vendor X, the proxy of vendor Y and the end AAA
server of vendor Z. 
Assume a piece of information V that only needs to be send from the
NAS_X to the end server_Z, without any semantic interpretation of the
proxy. (assume e.g., the proxy is just there for scaleability of the
configuration, and possible attribute modification)
Assume the proxy has not yet the software upgrade to syntactically and
semantically interprete V. 
In such a case, my opinion is that requirement (d) might be valuable.

note that specific to _interdomain_ proxy operation, the pass-along or
not of unknown information is indeed susceptible to a certain policy,
there I certainly agree with you.


regards,
Yves



> Robert
> 
> --


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 21 04:40:58 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id EAA12751;
	Thu, 21 Oct 1999 04:40:58 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id EAA18931 for nasreq-outgoing; Thu, 21 Oct 1999 04:22:41 -0400 (EDT)
Received: from scallop.baynetworks.com (ns5.baynetworks.com [194.133.90.101]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id EAA18927 for <nasreq@tdmx.rutgers.edu>; Thu, 21 Oct 1999 04:22:39 -0400 (EDT)
Received: from mailhost.BayNetworks.COM (h8754.s84f5.BayNetworks.COM [132.245.135.84])
	by scallop.baynetworks.com (8.9.1/8.9.1) with ESMTP id KAA15671
	for <nasreq@tdmx.rutgers.edu>; Thu, 21 Oct 1999 10:24:57 +0200 (MET DST)
Received: from pobox.engeast.BayNetworks.COM (pobox.engeast.baynetworks.com [192.32.61.6])
	by mailhost.BayNetworks.COM (8.9.1/8.8.8) with ESMTP id EAA04003
	for <nasreq@tdmx.rutgers.edu>; Thu, 21 Oct 1999 04:18:15 -0400 (EDT)
Received: from ltrpluim.corpemea.baynetworks.com ([141.251.166.11])
	by pobox.engeast.BayNetworks.COM (SMI-8.6/BNET-97/04/24-S) with ESMTP
	id EAA10450; Thu, 21 Oct 1999 04:22:13 -0400
	for <nasreq@tdmx.rutgers.edu>
Received: (from rpluim@localhost)
	by ltrpluim.corpemea.baynetworks.com (8.8.7/8.8.7) id KAA03378;
	Thu, 21 Oct 1999 10:17:19 +0200
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <14350.52235.528327.109860@ltrpluim.corpemea.baynetworks.com>
Date: Thu, 21 Oct 1999 10:17:15 +0200 (CEST)
To: nasreq <nasreq@tdmx.rutgers.edu>
Subject: Re: nasreq attribute handling
In-Reply-To: <380EBD68.509EAEE3@alcatel.be>
References: <380C61A7.D868ED24@alcatel.be>
	<14349.28788.240535.140448@ltrpluim.corpemea.baynetworks.com>
	<380EBD68.509EAEE3@alcatel.be>
X-Mailer: VM 6.75 under 21.2  (beta19) "Shinjuku" XEmacs Lucid
X-Attribution: Robert
From: Robert Pluim <rpluim@bigfoot.com>
Reply-To: Robert Pluim <rpluim@bigfoot.com>
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Yves T'Joens writes:
 > Robert Pluim wrote:
 > > Yves T'Joens writes:
 > >  > The protocol MUST allow the sending implementation to indicate to the
 > >  > next hop in the proxy chain to behave as follows on the receipt of an
 > >  > unknown attribute/message
 > >  > (a) silently discard and proceed
 > >  > (b) interrupt the session associated with the sending of this
 > >  > attribute/message
 > > 
 > > ?? What's the difference between (a) and (b). Discarding implies
 > > interruption.
 > 
 > I would not say so. Some AVPs may just be discarded, while the session
 > will survive, only with deteriorated service handling. This is in my
 > opinion how most of the vendor specific additions to any base protocol
 > should be handled.
 >

Sorry, I misunderstood here, I thought you meant discarding the entire
message. You're right, in most cases unrecognised attributes should
just be transparently discarded.

 > >  > (c) send back an error message and proceed unmodified
 > >  > (d) forward the attribute/message unmodified to the next hop (no
 > >  > semantic interpretation required)
 > >  >
 > > 
 > > If I were even a mildly paranoid ISP, I would never allow (c) and
 > 
 > well, (c) would only allow some limited debugging, however this req. may
 > be dropped, I don't care too much...
 >

Hmm, ok. (b) would make it more obvious that things are going wrong,
though.

 > > scream at anyone that implemented (d). Who knows what that
 > > attribute/message could be controlling?
 > > 
 > Given a single administrative domain, and a proxy situation. Trust of
 > operation is not necessarily an issue.
 >

Given a single administrative domain, I agree. Given multiple
administrative domains, behaviour like (d) is distinctly unpopular
unless explicitly enabled.

 > Assume the NAS to be of vendor X, the proxy of vendor Y and the end AAA
 > server of vendor Z. 
 > Assume a piece of information V that only needs to be send from the
 > NAS_X to the end server_Z, without any semantic interpretation of the
 > proxy. (assume e.g., the proxy is just there for scaleability of the
 > configuration, and possible attribute modification)
 > Assume the proxy has not yet the software upgrade to syntactically and
 > semantically interprete V. 
 > In such a case, my opinion is that requirement (d) might be valuable.
 >

In that one case, yes.

 > note that specific to _interdomain_ proxy operation, the pass-along or
 > not of unknown information is indeed susceptible to a certain policy,
 > there I certainly agree with you.

So perhaps we should specify that the behaviour should be configurable
(in practice I think it would be made so anyway).

Robert


From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 21 08:08:50 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id IAA16587;
	Thu, 21 Oct 1999 08:08:49 -0400 (EDT)
Received: (from majordom@localhost) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) id HAA19697 for nasreq-outgoing; Thu, 21 Oct 1999 07:55:23 -0400 (EDT)
Received: from mercury.Sun.COM (mercury.Sun.COM [192.9.25.1]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id HAA19693 for <nasreq@tdmx.rutgers.edu>; Thu, 21 Oct 1999 07:55:21 -0400 (EDT)
Received: from engmail2.Eng.Sun.COM ([129.146.1.25])
	by mercury.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id EAA19820
	for <nasreq@tdmx.rutgers.edu>; Thu, 21 Oct 1999 04:55:19 -0700 (PDT)
Received: from ha1mpk-mail.eng.sun.com (phys-ha1mpka.Eng.Sun.COM [129.146.53.34])
	by engmail2.Eng.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6) with SMTP id EAA09479
	for <nasreq@tdmx.rutgers.edu>; Thu, 21 Oct 1999 04:55:19 -0700 (PDT)
Received: from ha1mpk-mail.Eng.Sun.COM by ha1mpk-mail.eng.sun.com (SMI-8.6/SMI-SVR4)
	id EAA21123; Thu, 21 Oct 1999 04:55:14 -0700
From: Pat.Calhoun@eng.sun.com (Patrice Calhoun)
Message-Id: <199910211155.EAA21123@ha1mpk-mail.eng.sun.com>
Date: Thu, 21 Oct 1999 04:53:19 -0700
To: <nasreq@tdmx.rutgers.edu>
Reply-To: <pcalhoun@eng.sun.com>
Subject: nasreq on demand accounting
X-Mailer: Sun NetMail 2.3
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

All,

While looking at the current nasreq requirements, the following one jumped out
at me:

>     8.4.1.6.  On-demand Accounting
>
>
>     NAS  operators  need  to  maintain an accurate view onto the status of
>     sessions served by a NAS, even  through  failure  of  the  NA  or  AAA
>     server.   Therefore, the AAA protocol MUST support a means of request-
>     ing current session state and accounting from the NAS on demand.   The
>     intention is to provide for recovery if, for whatever reason, the nor-
>     mal flow of accounting data is interrupted.

I have a question about this one. Although it will probably work in an
intra-domain case, how do we expect this to work in inter-domain cases?
It seems like there would be some scaling issues here.

patC



From owner-nasreq@the-dread-pirate-roberts.rutgers.edu  Thu Oct 21 18:18:11 1999
Received: from the-dread-pirate-roberts.rutgers.edu (the-dread-pirate-roberts.rutgers.edu [128.6.21.44])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA15133;
	Thu, 21 Oct 1999 18:18:10 -0400 (EDT)
Received: (from majordom@localhost)
	by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.8.8) id RAA24090
	for nasreq-outgoing; Thu, 21 Oct 1999 17:57:35 -0400 (EDT)
X-Authentication-Warning: the-dread-pirate-roberts.rutgers.edu: majordom set sender to owner-nasreq@tdmx.rutgers.edu using -f
Received: from viper.columbus.rr.com (viper.columbus.rr.com [204.210.252.254]) by the-dread-pirate-roberts.rutgers.edu (8.8.8+Sun/8.6.12) with ESMTP id RAA24086 for <nasreq@tdmx.rutgers.edu>; Thu, 21 Oct 1999 17:57:34 -0400 (EDT)
Received: from pavilion (dhcp93126062.columbus.rr.com [24.93.126.62])
	by viper.columbus.rr.com (8.9.3/8.9.3) with SMTP id RAA12514
	for <nasreq@tdmx.rutgers.edu>; Thu, 21 Oct 1999 17:54:24 -0400 (EDT)
From: "Mark Anthony Beadles" <mark.beadles@columbus.rr.com>
To: <nasreq@tdmx.rutgers.edu>
Subject: nasreq NASREQ Meetings Scheduled
Date: Thu, 21 Oct 1999 18:00:17 -0400
Message-ID: <000601bf1c0f$a9e35b80$3e7e5d18@pavilion.columbus.rr.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook 8.5, Build 4.71.2173.0
Importance: Normal
X-MimeOLE: Produced By Microsoft MimeOLE V4.72.3110.3
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

We have been scheduled for two slots on the afternoon of Tuesday, November
9th.

TUESDAY, November 9, 1999
1545-1645  Afternoon Sessions III
	APP  ediint	Electronic Data Interchange-Internet Integration WG
	INT  ipcdn	IP Over Cable Data Network WG
*****	OPS  nasreq	Network Access Server Requirements WG **********
	OPS  adslmib	ADSL MIB WG
	RTG  manet	Mobile Ad-hoc Networks WG
	TSV  ecm	Endpoint Congestion Management BOF

1700-1800  Afternoon Sessions IV
*****	OPS  nasreq	Network Access Server Requirements WG **********
	OPS  adslmib	ADSL MIB WG
	RTG  vrrp	Virtual Router Redundancy Protocol WG
	SEC  ipsra	IP Security Remote Access BOF
	TSV  ecm	Endpoint Congestion Management BOF

+        Mark Anthony Beadles         +
+    mark.beadles@columbus.rr.com     +
+ Vox 614.565.1308 Pager 888.660.8427 +



From owner-nasreq@kull.rutgers.edu  Wed Oct 27 14:58:16 1999
Received: from kull.rutgers.edu (kull.rutgers.edu [128.6.21.58])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA00416;
	Wed, 27 Oct 1999 14:58:09 -0400 (EDT)
Received: (from majordom@localhost)
	by kull.rutgers.edu (8.8.8+Sun/8.8.8) id OAA20088
	for nasreq-outgoing; Wed, 27 Oct 1999 14:35:48 -0400 (EDT)
Received: from viper.columbus.rr.com (viper.columbus.rr.com [204.210.252.254])
	by kull.rutgers.edu (8.8.8+Sun/8.8.8) with ESMTP id OAA20080
	for <nasreq@tdmx.rutgers.edu>; Wed, 27 Oct 1999 14:35:42 -0400 (EDT)
Received: from pavilion (dhcp93126062.columbus.rr.com [24.93.126.62])
	by viper.columbus.rr.com (8.9.3/8.9.3) with SMTP id OAA04297;
	Wed, 27 Oct 1999 14:28:21 -0400 (EDT)
From: "Mark Anthony Beadles" <mark.beadles@columbus.rr.com>
To: <aboba@internaut.com>, <jari@arkko.com>, <aaa-wg@merit.edu>,
        <nasreq@tdmx.rutgers.edu>
Cc: <jari.arkko@ericsson.com>
Subject: nasreq RE: Some questions about the requirement matrix
Date: Wed, 27 Oct 1999 14:34:10 -0400
Message-ID: <001801bf20a9$dcaf48a0$3e7e5d18@pavilion.columbus.rr.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook 8.5, Build 4.71.2173.0
X-MimeOLE: Produced By Microsoft MimeOLE V4.72.3110.3
Importance: Normal
In-Reply-To: <009a01bf1ebb$f2dc3a60$478939cc@internaut.com>
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

From: Bernard Aboba:
>2. Aren't there some mobile IP requirements that might be not specific
>to just mobile IP? Such as "Ability to obtain QoS information from AAA"
>which certainly is usable even for dial-up NASes in various ways (multilink
>limits, DiffServ classes, multicast rights, ...).
>
>>>I was going to task a question about that requirement myself :)
>>>Not clear to me whether is requirement is different from the
>>>discussion in the NASREQ requirements doc or not. Perhaps
>>>the authors can explain this, if not here and now, in DC.

I think that this requirement to "obtain QoS information from AAA" is
certainly also a requirement for dial-up and virtual NAS's also.  I would
love to include more specific requirements for NASREQ concerning QoS.  A
volunteer who is a subject matter expert and who is willing to help me write
some text would be very helpful.

+        Mark Anthony Beadles         +
+    mark.beadles@columbus.rr.com     +
+ Vox 614.565.1308 Pager 888.660.8427 +





From owner-nasreq@kull.rutgers.edu  Wed Oct 27 15:38:41 1999
Received: from kull.rutgers.edu (kull.rutgers.edu [128.6.21.58])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA03132;
	Wed, 27 Oct 1999 15:38:36 -0400 (EDT)
Received: (from majordom@localhost)
	by kull.rutgers.edu (8.8.8+Sun/8.8.8) id PAA27515
	for nasreq-outgoing; Wed, 27 Oct 1999 15:16:01 -0400 (EDT)
Received: from viper.columbus.rr.com (viper.columbus.rr.com [204.210.252.254])
	by kull.rutgers.edu (8.8.8+Sun/8.8.8) with ESMTP id PAA27496
	for <nasreq@tdmx.rutgers.edu>; Wed, 27 Oct 1999 15:15:11 -0400 (EDT)
Received: from pavilion (dhcp93126062.columbus.rr.com [24.93.126.62])
	by viper.columbus.rr.com (8.9.3/8.9.3) with SMTP id PAA08218;
	Wed, 27 Oct 1999 15:11:43 -0400 (EDT)
From: "Mark Anthony Beadles" <mark.beadles@columbus.rr.com>
To: "'D. Harrington'" <dharrington@mediaone.net>, <aaa-wg@merit.edu>,
        <nasreq@tdmx.rutgers.edu>
Cc: <dbh@cabletron.com>
Subject: RE: nasreq-criteria-03 and AAA 
Date: Wed, 27 Oct 1999 15:17:30 -0400
Message-ID: <002501bf20af$ea70f460$3e7e5d18@pavilion.columbus.rr.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook 8.5, Build 4.71.2173.0
X-MimeOLE: Produced By Microsoft MimeOLE V4.72.3110.3
Importance: Normal
In-Reply-To: <3812CB02.97DCF4E3@mediaone.net>
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

From: D. Harrington
Thanks very much for the comments.  I will certainly try to clarify many of
the points in -criteria-04, after D.C.  Here are some responses:

>3) 8.1.2.2 says the transport for accounting MUST be reliable, but it
>doesn't define reliable. Does this mean it must be 100% reliable? no
>packet loss? must use message acknowledgements? must use a mechanism to
>check all data for integrity?

In this requirements document, I have deliberately avoided specifying
protocol implementations.  However, the intent is that it MUST be reliable
in the sense that TCP is reliable, and in the sense that RADIUS Accounting
is NOT.

>5) Most of the requirements specified in 8.1.x.x. have not been
>reflected in the AAA matrix for accounting: load balancing, not a single
>point of failure, fail-over, flow control, must support multiple domains

Yes, an oversight on my part, thanks.

>"Must not use a model that assumes a single domain of authority." - The
>accounting data must be able to contain references to multiple domains of
>authority, but can there be a single domain of authority over the
>accounting data collected? i.e. if SNAFU Communications collects
>accouting data about the resources used in their network, especially
>proprietary data, isn't SNAFU the single domain of authority over that
>data? Can they limit access to the collected data? Can they collect the
>data using a protocol which  "knows" that the collecting entity will be
>the single domain of authority over that collected data?

Certainly the AAA protocol must be able to handle the case where there is
really only a single domain of authority, as a degenerate case of the
multiple-domain scenario.  But I think the multiple domain case is going to
become much more common.  Even in your example, SNAFU may be leasing its
POPs from a reseller, and may be selling services to enterprise customers
who are responsible for authenticating their own users.

>An SNMP agent
>"owns" the data in its virtual database. Can an SNMP-based accounting
>solution ever be considered AAA compliant?

This is a question we must study.  Some of us are going to be writing a
document after D.C. attempting to answer this very question.

>"Years of experience with AAA protocols and NAS's has proven..." Is this
>incontrovertible proof? Who has this proof? Are there references I can
>use to find and review this proof?

"Proven" in the sense of "shown" or "demonstrated", not in the sense of
mathematical proof.  If you have some desire to convert the archives and
outputs of the RADIUS, NASREQ, and Roamops WGs over the years into a formal
calculus, feel free :)

>9) 8.1.3 defines some minimums. I don't understand the source of the
>numbers, and consequently what the requirement is: "minimum of 64K
>Attributes (minimum 16 bits), each with a ninimum length of 64K (minimum
>16 bits)."

Actually, I consider theses requirements to be fairly straightforward.  The
protocol has to support at least 65535 unique attributes.  A minimum 16 bits
are required to enumerate this many attributes.  The protocol must allow
each attribute to contain up to 65535 octets of data.  A minimum 16 bits are
required to encode this length.

>10) 8.1.3.1 AAA MUST provide some support for complex types. Why is this
>a MUST? Are there attributes that MUST be suported that MUST be
>represented using complex types? What if somebody defined a protocol
>that, using simple types in the protocol, could be used to convey all
>the information needed? Why MUST it also support complex types?

The best example that I can think of is
draft-ietf-radius-tunnel-auth-09.txt.  Check out the "tags".  Using complex
tagged types greatly simplifies the description of multiple tunnels.

>11) 8.1.3.2 "an extension to that protocol MUST be defined which
>supports this set." - why not make the requirement that AAA must be able
>to be extended to support the set of RADIUS attributes. We aren't
>defining other required attributes at this time, why are we making these
>required attributes? Can't we define them later, and separately, just
>like it appears is the plan for all other attributes?

An AAA protocol that does not support EVERY bit of functionality that RADIUS
currently does is 100% useless.  We can define them in a separate document,
yes, but unless they are defined, then the requirements for NASREQ are not
met.  I am not interested in a protocol which is merely elegant and
well-designed.  It must also be useful.  ISPs will never move to a new
protocol unless it does what they need it to.

>12) Throughout the industry, years of operational experience with <take
>your pick> protocols has proven that allowing vendor-specific attributes
>and data types can have a negative effect on interoperability.

=Grin= Do you have a copy of this proof that I can peruse?  Seriously, if we
do not provide standard support IN THE PROTOCOL for vendor-specifics, then
vendors will simply come up with their own non-standard extensions anyway.
That will have a very severe effect on interoperability.  Witness the
problem with RADIUS, where prior to VSA's (and for many years after),
certain vendors just picked parts of the standard attribute space and used
them in incompatible ways.

>22) 8.4.2 'the AAA protocol must support delivery of the following types
>of accounting/auditing data: ... All parameters used to authenticate a
>session ... Details of the authorization profile that was applied to the
>session..." - isn't this a potential security hole? The requirement
>should be very clear abut what data should and should NOT be passed in
>the accounting data.

Yes, I should have been more explicit.  Certainly things like user passwords
should not be sent in accounting data.

+        Mark Anthony Beadles         +
+    mark.beadles@columbus.rr.com     +
+ Vox 614.565.1308 Pager 888.660.8427 +



From owner-nasreq@kull.rutgers.edu  Wed Oct 27 15:48:44 1999
Received: from kull.rutgers.edu (kull.rutgers.edu [128.6.21.58])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA03666;
	Wed, 27 Oct 1999 15:48:43 -0400 (EDT)
Received: (from majordom@localhost)
	by kull.rutgers.edu (8.8.8+Sun/8.8.8) id PAA29575
	for nasreq-outgoing; Wed, 27 Oct 1999 15:26:27 -0400 (EDT)
Received: from viper.columbus.rr.com (viper.columbus.rr.com [204.210.252.254])
	by kull.rutgers.edu (8.8.8+Sun/8.8.8) with ESMTP id PAA29565
	for <nasreq@tdmx.rutgers.edu>; Wed, 27 Oct 1999 15:26:26 -0400 (EDT)
Received: from pavilion (dhcp93126062.columbus.rr.com [24.93.126.62])
	by viper.columbus.rr.com (8.9.3/8.9.3) with SMTP id PAA08976;
	Wed, 27 Oct 1999 15:21:21 -0400 (EDT)
From: "Mark Anthony Beadles" <mark.beadles@columbus.rr.com>
To: "'Yves T'Joens'" <yves.tjoens@alcatel.be>,
        "'nasreq'" <nasreq@tdmx.rutgers.edu>
Subject: RE: nasreq comments on criteria-03
Date: Wed, 27 Oct 1999 15:27:09 -0400
Message-ID: <002801bf20b1$4391ebc0$3e7e5d18@pavilion.columbus.rr.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook 8.5, Build 4.71.2173.0
X-MimeOLE: Produced By Microsoft MimeOLE V4.72.3110.3
Importance: Normal
In-Reply-To: <3805DAFD.B160F2E9@alcatel.be>
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk
Content-Transfer-Encoding: 7bit

Thanks very much for the comments, Yves.

From: Yves T'Joens
> section 8.1.2.1 fast fail over.
>
> do you mean here the detection that fast fail over is necessary or the
> fail over mechanism itself ?
> To be more precise, if it would be the detection, would that
> mean that a
> keep alive mechanism must work on such a (pretty fast ;-) )
> time scheme,
> or do you refer more exactly that no whole database synchronization
> should happen between client and server on fail over...

The latter, mostly.  The failover time was a suggestion.  Certainly we can
come up with a better figure, if someone has a better suggestion.

> section 8.1.3 Attribute value protocol
>
> this section only deals with attribute extensibility, should message
> extensibility also be mentioned explicitly ??

This is probably a good idea, too.

> section 8.1.3.1 Attribute data types
>
> this section words the requirement as "... simple attribute
> data types,
> including integer, enumeration, text string...." Does the 'including'
> refer to the fact that many more simple data types may be defined, if
> so, then perhaps something like 'not limited to' might be added to the
> sentence. I was thinking on opaque values to be included in the AVP...

OK, that makes sense.

> section  8.3.2.2 Authorizatoin attribute requirements
>
> -Quality of Service : this requirement seems to be so wide that it is
> hardly applicable, should we somehow more specify it, e.g., should
> support authorization to intserv based resource requests (rsvp) or
> diffserv forwarding behaviour... ?

As I have stated elsewhere, text from any interested party is appreciated,
as I am not an expert in that arena. ;)

+        Mark Anthony Beadles         +
+    mark.beadles@columbus.rr.com     +
+ Vox 614.565.1308 Pager 888.660.8427 +



From owner-nasreq@kull.rutgers.edu  Wed Oct 27 16:03:31 1999
Received: from kull.rutgers.edu (kull.rutgers.edu [128.6.21.58])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA04750;
	Wed, 27 Oct 1999 16:03:16 -0400 (EDT)
Received: (from majordom@localhost)
	by kull.rutgers.edu (8.8.8+Sun/8.8.8) id PAA02400
	for nasreq-outgoing; Wed, 27 Oct 1999 15:39:12 -0400 (EDT)
Received: from dfssl.exchange.microsoft.com (dfssl.exchange.microsoft.com [131.107.88.59])
	by kull.rutgers.edu (8.8.8+Sun/8.8.8) with ESMTP id PAA02392
	for <nasreq@tdmx.rutgers.edu>; Wed, 27 Oct 1999 15:39:11 -0400 (EDT)
Received: by dfssl with Internet Mail Service (5.5.2650.21)
	id <VNQ23QZA>; Wed, 27 Oct 1999 12:39:07 -0700
Message-ID: <078292D50C98D2118D090008C7E9C6A60644A2D9@STAY.platinum.corp.microsoft.com>
From: "Bernard Aboba (Exchange)" <bernarda@Exchange.Microsoft.com>
To: "'Mark Anthony Beadles'" <mark.beadles@columbus.rr.com>,
        "'D. Harrington'" <dharrington@mediaone.net>, aaa-wg@merit.edu,
        nasreq@tdmx.rutgers.edu
Cc: dbh@cabletron.com
Subject: RE: nasreq-criteria-03 and AAA 
Date: Wed, 27 Oct 1999 12:39:06 -0700
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: multipart/alternative;
	boundary="----_=_NextPart_001_01BF20B2.EE7B1718"
Sender: owner-nasreq@tdmx.rutgers.edu
Precedence: bulk

This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.

------_=_NextPart_001_01BF20B2.EE7B1718
Content-Type: text/plain;
	charset="iso-8859-1"

>An SNMP agent
>"owns" the data in its virtual database. Can an SNMP-based accounting
>solution ever be considered AAA compliant?

This issue is partially addressed in draft-aboba-acct-02.txt. The
conclusion in that draft is that SNMP v3 (with the right MIBs) 
*can* meet accounting requirements for a number of scenarios, 
particularly with the NMRG extensions. As a result, it is expected
that SNMP will remain popular in accounting applications for
the foreseeable future. 

However, SNMP v3 cannot meet all the requirements -- particularly 
requirements for low-latency inter-domain event-driven accounting. See the
draft for a detailed explanation of what accounting capabilities
SNMP v3 can and cannot support. 

------_=_NextPart_001_01BF20B2.EE7B1718
Content-Type: text/html;
	charset="iso-8859-1"

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="Generator" CONTENT="MS Exchange Server version 5.5.2650.12">
<TITLE>RE: nasreq-criteria-03 and AAA </TITLE>
</HEAD>
<BODY>

<P><FONT SIZE=2>&gt;An SNMP agent</FONT>
<BR><FONT SIZE=2>&gt;&quot;owns&quot; the data in its virtual database. Can an SNMP-based accounting</FONT>
<BR><FONT SIZE=2>&gt;solution ever be considered AAA compliant?</FONT>
</P>

<P><FONT SIZE=2>This issue is partially addressed in draft-aboba-acct-02.txt. The</FONT>
<BR><FONT SIZE=2>conclusion in that draft is that SNMP v3 (with the right MIBs) </FONT>
<BR><FONT SIZE=2>*can* meet accounting requirements for a number of scenarios, </FONT>
<BR><FONT SIZE=2>particularly with the NMRG extensions. As a result, it is expected</FONT>
<BR><FONT SIZE=2>that SNMP will remain popular in accounting applications for</FONT>
<BR><FONT SIZE=2>the foreseeable future. </FONT>
</P>

<P><FONT SIZE=2>However, SNMP v3 cannot meet all the requirements -- particularly </FONT>
<BR><FONT SIZE=2>requirements for low-latency inter-domain event-driven accounting. See the</FONT>
<BR><FONT SIZE=2>draft for a detailed explanation of what accounting capabilities</FONT>
<BR><FONT SIZE=2>SNMP v3 can and cannot support. </FONT>
</P>

</BODY>
</HTML>
------_=_NextPart_001_01BF20B2.EE7B1718--


