Seamoby Working Group                                  Stefano M. Faccin
INTERNET-DRAFT                                             Rajeev Koodli
Date: November 2001                                            Franck Le
Expires: May 2002                                        Jari T. Malinen
                                                            Rene Purnadi


            Dormant Mode Handover Support in Mobile Networks
                      <draft-koodli-paging-01.txt>



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 documents of the Internet Engineering
   Task Force (IETF), its areas, and its working groups.  Note that
   other groups may also distribute working 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
   material 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.


Abstract

   This document defines an IP paging concept that supports IP dormancy
   for hosts and Mobile Nodes.  The concept specifies a generic model
   that can be applied to several mobility mechanisms (including Mobile
   IPv6, Localized Mobility Managament) and is independent of the
   layer-2 access technology and paging capabilities.  The model allows
   for optimizations of the benefits of layer-2 paging when present,
   while minimizing the impact on layer-2 paging.








Faccin, Koodli, Le, Malinen, Purnadi                            [Page i]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


                             Table of Contents


   Status of This Memo . . . . . . . . . . . . . . . . . . . . . . .   i

   Abstract  . . . . . . . . . . . . . . . . . . . . . . . . . . . .   i

   1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . .   1

   2. Terminolgy . . . . . . . . . . . . . . . . . . . . . . . . . .   1

   3. Protocol Overview  . . . . . . . . . . . . . . . . . . . . . .   2

   4. IP Paging Model  . . . . . . . . . . . . . . . . . . . . . . .   3
      4.1. Mobile Node IP Paging States  . . . . . . . . . . . . . .   3
      4.2. IP Paging Illustration  . . . . . . . . . . . . . . . . .   4
      4.3. Advertising Paging Area Information . . . . . . . . . . .   5
      4.4. PF Discovery  . . . . . . . . . . . . . . . . . . . . . .   6
      4.5. Paging Area Update  . . . . . . . . . . . . . . . . . . .   6
      4.6. Detecting IP Dormant MN . . . . . . . . . . . . . . . . .   7
      4.7. Paging Deregistration . . . . . . . . . . . . . . . . . .   7

   5. IP Paging Identity . . . . . . . . . . . . . . . . . . . . . .   7
      5.1. The network configures L3 IP Paging Identity  . . . . . .   8
      5.2. The L3 IP Paging Identity via autoconfiguration . . . . .   9
      5.3. L2 identity for paging  . . . . . . . . . . . . . . . . .   9

   6. IP Paging and L2 paging  . . . . . . . . . . . . . . . . . . .  10

   7. Security Model for IP Paging . . . . . . . . . . . . . . . . .  10
      7.1. Authentication of Paging Request messages . . . . . . . .  10
      7.2. Authentication of Paging Response . . . . . . . . . . . .  12
      7.3. Authentication of Paging Area Update  . . . . . . . . . .  13
      7.4. Filtering . . . . . . . . . . . . . . . . . . . . . . . .  15

   8. Application of IP Paging Model to Different Scenarios  . . . .  16

   9. Comparing Against the Requirements . . . . . . . . . . . . . .  19
      9.1. Impact on Power Consumption . . . . . . . . . . . . . . .  20
      9.2. Scalability . . . . . . . . . . . . . . . . . . . . . . .  20
      9.3. Control of Broadcast/Multicast/Anycast  . . . . . . . . .  21
      9.4. Efficient Signaling for Inactive Mode . . . . . . . . . .  21
      9.5. No Routers  . . . . . . . . . . . . . . . . . . . . . . .  21
      9.6. Multiple Dormant Mode . . . . . . . . . . . . . . . . . .  21
      9.7. Independence of Mobility Protocol . . . . . . . . . . . .  22
      9.8. Support for Existing Mobiltiy Protocols . . . . . . . . .  22
      9.9. Dormant Mode Termination  . . . . . . . . . . . . . . . .  22
      9.10. Network Update . . . . . . . . . . . . . . . . . . . . .  23



Faccin, Koodli, Le, Malinen, Purnadi                           [Page ii]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


      9.11. Efficient Utilization of L2  . . . . . . . . . . . . . .  23
      9.12. Orthogonality of Paging Area Subnets . . . . . . . . . .  24
      9.13. Future L3 Paging Support . . . . . . . . . . . . . . . .  24
      9.14. Robustness Against Failure of Network Elements . . . . .  24
      9.15. Reliability of Packet Delivery . . . . . . . . . . . . .  24
      9.16. Robustness Against Message Loss  . . . . . . . . . . . .  25
      9.17. Flexibility of Administration  . . . . . . . . . . . . .  25
      9.18. Flexibility of Paging Area Design  . . . . . . . . . . .  25
      9.19. Availability of Security Support . . . . . . . . . . . .  25
      9.20. Authentication of Paging Location Registration . . . . .  26
      9.21. Authentication of Paging Area Information  . . . . . . .  26
      9.22. Authentication of Paging Messages  . . . . . . . . . . .  26
      9.23. Paging Volume  . . . . . . . . . . . . . . . . . . . . .  26

   10. References  . . . . . . . . . . . . . . . . . . . . . . . . .  27

   11. Authors' Addresses  . . . . . . . . . . . . . . . . . . . . .  28

   12. Appendix A - Hierarchical Overlapping Paging Area Support . .  29
































Faccin, Koodli, Le, Malinen, Purnadi                          [Page iii]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


1.  Introduction

   The problem of Dormant Mode Host Alerting, otherwise known as IP
   Paging, is well-defined and described in [2].  It is generally
   accepted that an IP Paging solution would offer many advantages
   including, power saving when it is not available in certain Layer 2
   technologies (and enhancing it in those that already offer it),
   reduced IP layer signaling during dormant mode movements, and
   offering location tracking across heterogeneous access technologies,
   among others.  While these advantages are compelling, the solution
   itself has to meet certain requirements.  Specifically, it is
   identified that an IP Paging proposal has to address the requirements
   set forth in RFC 3154.  In this document, we propose a mechanism for
   IP Paging that provides the following advantages.

   1. is independent of any mobility management protocol

   2. can be adopted with several mobility management protocols while
      not only allowing to enjoy the benefit of DMHA in terms of both
      power saving and reduced signaling message exchange, but actually
      optimizing both

   3. allows to optimize the interaction between IP paging and L2 paging
      mechanisms when present

   4. minimizes impact on L2 paging and link layer technologies

   5. provides improvements in terms of power saving and minimization of
      mobility signaling with respect to the presence of L2 paging only

   6. supports both implicit and explicit IP dormancy

   7. allows support of different access technologies in the same IP
      paging area

   8. supports dynamic IP paging area

   Our proposal addresses the requirements set forth in RFC 3154.  This
   is outlined in Section 8.


2.  Terminology

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", "OPTIONAL", and
   "silently ignore" in this document are to be interpreted as described
   in RFC 2119.




Faccin, Koodli, Le, Malinen, Purnadi                            [Page 1]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


3.  Protocol Overview

   We coalesce the functional entities identified in RFC 3154, namely
   the Dormant Mobility Agent, the Tracking Agent, and the Paging Agent,
   into a single logical entity which we call the ``Paging Function'' or
   ``PF''. There are several advantages to such a coalescing of
   different functional elements.  First, it provides a single, trusted
   channel of paging-related communication between a MN and the network.
   For all paging purposes, the MN communicates with a single logical
   entity and secures its communication using an appropriate security
   association it shares with that entity.  Second, since the MN deals
   with a single logical entity, the problem of discovering different
   functionalities is alleviated.  Depending on where the PF is
   realized, the MN deals with its Access Router or visited domain
   Mobility Agent only.  Third, since the MN deals with a single logical
   entity, the number of separate messages necessary to communicate with
   different elements, potentially over an expensive air interface, can
   be reduced by aggregating them together.  For instance, a single
   message can establish the necessary paging state at DMA, TA and the
   PA. It has to be noted that the PF is a logical funtion and for a
   given MN in a given moment there is only one PF. PF is, however, in
   no way a centralized function

   The PF may be realized in different ways in practice.  We provide
   some examples in Section 7.  Note that since the IP dormancy of a MN
   has to be checked before the packet can be delivered to a MN, all the
   incoming packets traverse the PF. When an IP packet arrives at a PF,
   where the MN has last established its presence, the PF has to
   determine how to forward the packet.  We propose two ways by which
   the PF acertains that the MN to which the packet is addressed to has
   undergone dormant mode handover.  First, prior to undergoing the
   dormant mode handover, the MN performs an explicit registration with
   the PF. Second, we allow for an implicit registration, in which the
   network may initiate IP Paging when it discovers that a MN is no
   longer reachable.  The details of these registration procedures are
   presented in Section 4.

   Once the PF determines the need for IP Paging, it initiates a paging
   message addressed to an IP Paging Area (IP-PA). We identify each IP-
   PA by an IP multicast group, whose members are typically all the
   access routers to which a MN could be ``dormantly connected''.  We
   propose sending IP Paging messages over the access network, which is
   typically connected with a higher-speed network compared to the
   access link bandwidth, and make use of Layer 2 paging as much as
   possible over the air interface.  This ensures preserving spectrum
   where it is considered important and allows usage of enhanced IP
   messages over a higher-speed access network.  Each IP-PA could span
   multiple subnets and each subnet could represent a disparate access



Faccin, Koodli, Le, Malinen, Purnadi                            [Page 2]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   technology.  An IP Paging message sent to the IP-PA multicast group
   contains a unique paging identity that is known both to the latest PF
   and the MN. This paging identity serves two purposes.  It acts as
   input to a function that creates a Layer 2 identity to page once an
   IP Paging message is received over the access network.  Where Layer 2
   paging is not available, the unique paging identity is used to send
   an IP Paging message over the access link.  The description of this
   unique paging identity and the IP Paging message sent over the acces
   network itself is described in Section 5.

   One or more of the ARs perform paging (either L2 or IP layer) in
   response to the arrival of an IP Paging message over the access
   network.  In response to this access link paging message, the MN
   wakes up and responds to the access link paging message.  The MN may
   then proceed to perform IP layer registrations, such as a Binding
   Updates.  The access link paging messages must be secure and allow
   for mutual authentication of the network and the MN. These messages
   and the corresponding operations (in the form of a state machine) are
   described in Section 4.  In the subsequent section, we describe the
   relation between the IP paging and Layer 2 paging.

   It is clearly identified that the IP paging mechanism must address
   the security considerations.  These are elaborated in Section 6.  We
   discuss the realization of the PF entity in an AR or a Mobility Agent
   in Section 7.  Finally, we describe some important enahncements to
   the basic IP Paging protocol in Section A.


4.  IP Paging Model


4.1.  Mobile Node IP Paging States

   The MN IP paging states model the MN behavior in terms of MN activity
   and reachability at IP level.  The following states are defined.

   -  MN de-registered:  MN is not reachable, has no IP address
      allocated and is not able to send or receive any packets to the
      network (e.g.  terminal is powered off).

   -  MN Registered-Active:  MN has performed the registration to access
      the network, has a valid IP address, and is capable of sending and
      receiving packets without need for additional signaling.

   -  MN Registered-IP Dormant:  Location tracking is at the IP-PA
      level, i.e., routing information is available to route packets to
      the "paging area" or, better, to the node acting as PF. While IP
      dormant, MN wakes up only to perform appropriate operations (e.g.



Faccin, Koodli, Le, Malinen, Purnadi                            [Page 3]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


      listen for paging, signal when entering a new IP-PA, etc.).  In
      order to forward packets to the MN the network needs to page the
      terminal at the IP level (IP paging).

   When the terminal is not a mobile node but any generic host, we
   define the following states.

   -  Active Host:  the host is connected to the network, has a valid IP
      address is capable of sending and receiving packets without need
      for additional signaling.

   -  IP Dormant Host:  the host is reachable only at the IP-PA level,
      i.e., routing information are available to route packets to the
      "paging area" or, better, to the node acting as PF. While IP
      dormant, the host wakes up only to perform appropriate operations
      (e.g.  listen for paging, paging area updates needed when entering
      a new IP-PA, etc.).  In order to forward packets to the host, the
      network needs to page the host at the IP level(IP paging).


4.2.  IP Paging Illustration

   We illustrate the concepts using Figure 1.

   At time t0 packets destined for the MN arrive at the PF. At t1, the
   PF realizes that the MN is IP dormant.  At t2 the PF sends ``IP
   Paging Request'' message to all ARs within the IP-PA where the MN is
   located.  At t3 the Access Router either sends a L3 paging request
   (e.g., a router advertisement with paging extension) or it converts
   the IP Paging Request message into an appropriate L2 paging message
   and forwards the request to the MN. When L2 paging is used, the L3
   paging id is mapped to a L2 paging id.  At the assigned time slot
   (t4) or the paging channel, the MN wakes up and listens to the page.
   The MN then responds to the paging message with a ``Paging Response''
   message to its AR. The MN (either subsequently or together with the
   Paging Response message) performs the IP mobility (neighbor discovery
   and binding updates) procedures.  At t5, the MN or its new AR
   requests PF to forward the buffered packets and delete the state of
   dormancy of the MN.

   The PF determines that the MN is dormant based on explicit ``Paging
   Registration'' message that the MN sends before entering IP-dormant
   state.  The PF may also determine, implicitly, that the MN is IP-
   dormant based on a MN-specific timer that expires in the event of
   traffic inactivity.  When a MN wakes up in response to paging, the PF
   performs ``Paging De-registration'' in order to update or remove
   state related to MN's dormancy.




Faccin, Koodli, Le, Malinen, Purnadi                            [Page 4]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   The IP Paging Request message is resent if it is not acknowledged.
   The acknowledgements, i.e., the Paging Response message could arrive
   from any of the ARs or directly from the MN. The re-transmissions are
   done up to a configurable number of times.


              |
              | (t0) incoming packet
              V
           +----+
           |    |
           | PF | (t1) PF realizes that MN is dormant
           |    |
           +----+
            (  |
           /   |<-(t2)``IP Paging Request''to all ARs within IP-PA.
          |<-(t5)MN (or AR on its behalf)
          |    | requests PF to forward packets.
+---------|---------------------------+
|     +---|----+--------+--------+    |
|     |   |    |        |        |    |
|     |   |    |        |        |    |
|     V   |    V        V        V    |
|  +----+ | +----+   +----+   +----+  |
|  | AR | | | AR |   | AR |   | AR |  |
|  +----+ \ +----+   +----+   +----+  |
|     |    \  |^        |        |    |
|     |     \ ||        |        |<- (t3) L3/L2 Paging Request to all MN
|     V      \V|        V        V  in each AR
|            +----+                   |
|            | MN | (t4)MN wakes-up and replies with ``Paging Response''
|            +----+  MN performs IP mobility
|                    to connect to AR and CN
| IP paging area                      |
+-------------------------------------+



                    Figure 1: IP Paging Illustration



4.3.  Advertising Paging Area Information

   The MN determines the identity of the IP Paging Area (IP-PA) it
   currently is in based on the IP-PA Identity that is advertised.
   Together with the IP-PA Identity, other IP-PA information, such as
   the IP-PA profile that indicates characteristics of that paging area,



Faccin, Koodli, Le, Malinen, Purnadi                            [Page 5]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   could be broadcast as well.  The IP-PA Identity and profile
   information are advertised by Paging Functions to all the Access
   Routers in the IP-PA by multicasting them to the ARs.  On receipt of
   this information, each AR in the IP-PA needs to advertise the same
   information to the MNs in its subnet.  This is done using extensions
   to Router Advertisements.

   The multicast address for IP Paging Area Advertisements is the same
   used by the PF to send IP Paging Requests in its IP-PA.


4.4.  PF Discovery

   The MN needs to know the identity of the serving PF (or the previous
   serving PF) to send the different Paging messages such as Paging Area
   Update, IP Paging Registration in the case of explicit dormancy, etc.
   In the implicit dormancy case, the MN still needs to know the
   identity of the serving PF: e.g., when it moves to an area served by
   a new PF, the MN needs to provide the new PF with the identity of the
   previous PF so that IP paging state in the previous PF can be
   released and information required at the new PF (e.g.  local paging
   session key) can be retrieved.  As described in the following
   corresponding sections, many mechanisms exist for the MN to discover
   the PF identity.

   The PF discovery can be done in several ways.  The PF address can be
   sent over router advertisements.  Or the MN can send a Paging
   Function Discovery Request to a pre-defined well-known anycast
   address.  And then, one of the Paging Functions, serving the area the
   MN is currently located, would reply in a Paging Function Discovery
   Reply, providing its identity.  Once it discovers the PF, the MN then
   perform all the required procedures with the selected PF.


4.5.  Paging Area Update

   When the MN realizes that it has moved to a different IP-PA, the MN
   initiate a paging area update procedure.  The paging area update will
   inform the the PF the current IP-PA so that paging can be efficiently
   directed to the correct IP-PA. The Paging Area Update message is
   typically sent when a MN learns of a new paging area from paging area
   advertisements. In the implicit dormancy case, the MN still needs to
   know the identity of the serving PF: e.g., when it moves to an area
   served by a new PF, the MN needs to provide the new PF with the
   identity of the previous PF so that IP paging state in the previous
   PF can be released and information required at the new PF (e.g.
   local paging session key) can be retrieved.




Faccin, Koodli, Le, Malinen, Purnadi                            [Page 6]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


4.6.  Detecting IP Dormant MN

   The PF declares that a MN is IP dormant upon receiving an explicit
   message to requet IP dormancy from the MN (explicit dormancy).The
   message requires to carry information such as the indication that MN
   wishes to go IP dormant, authentication data to verify the source of
   request, the Paging Area identity, etc.  On the other hand, if the
   the lack of traffic activity timer expires (implicit dormancy), the
   PF determines that the MN is dormant again.


4.7.  Paging Deregistration

   Paging deregistration is required in the following cases.

   -  when MN goes out of coverage (automatic deregistration): the state
      in the PF must be released, and this will also save unecessary
      paging.  This can be e.g based on a timer with an expiration time
      longer than the timer for IP dormancy.

   -  when MN power off: the state in the PF must be released, and this
      will also save unecessary paging upon receiving signal from the MN

   -  when the MN becomes active, the paging state in the PF must be
      released.

   -  when the MN handovers to a new PF, the paging state in the old PF
      must be released.

   In the latter two cases, the Paging Response message or the Paging
   Area Update message to the new PF, results in releasing the paging
   state.  For instance, when the MN is attached to a new PF, the latter
   sends a message to the old PF to release the state.


5.  IP Paging Identity

   To ensure that IP paging in an IP-PA can trigger L2 paging so that
   the appropriate identity can be used at L2 to page the MN (i.e.  the
   identity at L2 that the MN is expecting to be paged with), an L3 IP
   paging identity is proposed for the MN. The identity is generated
   according to a well-know mechanism and based on a well-known set of
   inputs.  This guarantees all MNs and networks know how to compute it.
   This also allows for the identity to be auto-configurable.  This L3
   IP paging identity needs to be unique only within an IP-PA. This
   allows for a shorter identifier.





Faccin, Koodli, Le, Malinen, Purnadi                            [Page 7]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   The assumption is that when the MN is registered to the network, both
   the MN and the network (in particular the PF) know the L3 IP Paging
   Identity that will be used for paging while MN is within the IP-PA. A
   mapping algorithm is defined for each link layer technology (if
   needed) to convert the L3 IP Paging Identity into a L2 temporary
   identity that can be used to page the MN at L2.  When MN is IP
   dormant and enters an area of coverage of a specific link layer
   technology, it will map the L3 IP Paging Identity to a L2 identity.
   When an IP Paging message is received by an AR, it is pushed to the
   AP where the L3 IP Paging Identity is mapped by the AP to a L2 paging
   identity according to the link layer technology.  The L2 identity is
   then used to page the MN.


5.1.  The network configures L3 IP Paging Identity

   The PF generates a temporary L3 IP paging identity that is valid for
   an IP-PA. If the IP-PA consists of multiple access technologies, the
   L3 IP Paging Identity will have the length equal to the shortest
   length of the L2 paging identity in a particular technology.
   Therefore, for that particular technology that has the shortest
   length L2 paging identity, the L2 paging identity is equal to the L3
   paging identity.  For the other access technology, the L2 paging
   identity is the expansion of the L3 IP paging identity.  To cover the
   case of the distibuted PF, each L3 IP paging identity consists of PF
   unique part and MN unique part.  This scheme allows the uniqueness of
   the L3 paging identity in an IP-PA and uniqueness of L2 paging
   identity in each access technology.

   Let's assume that the IP-PA consists of access technology A that
   requires 32 bits L2 paging identity, access technology B that
   requires 128 bits paging identity and access technology C that
   requires 20 bits L2 paging identity.  The PF will generate a 20 bits
   temporary L3 IP paging identity for the MN. When the MN uses access
   technology C, it uses the temporary L3 IP paging identity as its L2
   paging identity.  If the MN acquires access technology A, the MN will
   expand the 20 bits temporary L3 IP paging identity into 32 bits L2
   paging identity, retaining the uniqueness of the temporary L3 paging
   identity.  In case of distributed PF, the uniqueness of the temporary
   L3 IP Paging identity can be defended by assigning the first `n' bits
   unique to the PF and the remaining of the bits unique to the MN.

   In addition to computing the L2 Paging Identity, both the MN and the
   access point also use the L3 IP Paging identity to calculate the time
   slot/paging channel to be monitored.






Faccin, Koodli, Le, Malinen, Purnadi                            [Page 8]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


5.2.  The L3 IP Paging Identity via autoconfiguration

   Each AR broadcasts PA identifier (PA-Id) that consists of AR specific
   part and paging area specific part.  Upon receiving this PA-Id, the
   MN uses a hash function `F1' to its own identifier (e.g., its NAI,
   its permanent IP Address, etc.)  and generates L3 IP Paging Id that
   consists of the AR specific part that is received from PA-Id and the
   result of `F1' toward its own identifier.  The length of the L3 IP
   Paging Id is maximum equal to the L2 paging id in that particular
   access technology.

   When the MN sends a paging area update, the newly calculated L3 IP
   Paging Id, together with old L3 IP Paging Id and old PF, are included
   in the paging area update request message.  If the newly calculated
   L3 IP Paging Id is valid, an SUCCESS indication is sent by the PF to
   the MN through paging area update response message.  If it is not
   valid, the PF calculates a new L3 IP paging Id that consists of the
   AR specific part and MN specific part, and sends it to the MN via the
   paging area update response.  From the L3 IP Paging Id, the MN uses
   another function `F2' to calculate the L2 Paging Id and function `F3'
   to calculate the time slot/paging channel to monitor.  The access
   point also uses the L3 IP Paging Id to calculate the L2 paging Id
   (using F2 function) and the time slot/paging channel (using F3
   function).


5.3.  L2 identity for paging

   IP Paging must trigger L2 paging for technologies that have it.  L2
   paging is based on temporary identifiers specific to the L2
   technology.  Even if the L2 temporary identifier is known at the
   location where MN went IP dormant, the identifier cannot be used for
   paging in the points of attachment under other ARs since the
   identifier may be already in use.  In addition, if several link layer
   technologies are supported in same IP-PA, IP paging cannot use the L2
   paging identity of one technology to page in the others.

   Link-layer technology needs to know what time-slot/paging channel to
   use to page the MN. The correct time-slot/paging channel is known at
   location where MN went IP dormant, but not necessarily at the point
   of attachment where the MN actually is located.  How do the points of
   attachment select the appropriate time-slot/paging channel to page?
   The time slot/paging channel is decided based on the L3 paging
   identity.  For instance, in the current cellular systems, the time
   slot/paging channel is obtained using a hash function to the IMSI.






Faccin, Koodli, Le, Malinen, Purnadi                            [Page 9]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


6.  IP Paging and L2 paging

   When L2 paging is present:

   -  L3 paging triggers L2 paging for delivery of L3 paging request
      message

   -  when IP dormant, MN can avoid L2 mobility procedure while stays
      within the same IP-PA


7.  Security Model for IP Paging

   IP paging protocol MUST have a strong security mechanism to prevent
   all the threats identified and described in [4] that may affect the
   IP paging protocol performance.  Without an adequate security scheme,
   IP paging may not provide all the identified benefits but results in
   opposite effects, namely, increased signaling load due to IP paging,
   congested network, reduced battery life and even breakdown in
   communication for the MN.

   The different paging messages SHOULD be authenticated.


7.1.  Authentication of Paging Request messages

   The Paging Request messages SHOULD be authenticated.  Otherwise, an
   intruder may send fake Paging Request messages to the dormant MN. The
   MN will unnecessarily wake up and this may prevent that MN from
   switching to dormant mode.  As a result, the MN may quickly run out
   of battery, hence become inaccessible.  In wireless networks, the MN
   will also try to set up the radio connection and this may result in
   the waste of radio resources, which are already rare and expensive.

   Many methods are possible to authenticate the Paging Request
   messages: The MN and the network (specifically the PF) can set up a
   Paging session key `K'. The mechanism to compute and distribute this
   key is out of the scope of this document.  Then the network can use
   this key to authenticate the Paging Request message.

   Paging Request authentication may be performed between the PF and the
   MN or between the AR and the MN. If performed in AR, the PF may store
   this paging session key and distribute it to the different Access
   routers of the paged area in the paging message; or the access router
   may retrieve it from a Key Storage Center. If performed by the AR,
   the paging requet between the PF and the AR can be authenticated with
   keys that are not user specific but established between PF And arS in
   the IP-PA (method is out of the scope of this document)



Faccin, Koodli, Le, Malinen, Purnadi                           [Page 10]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   Anti Replay protection may be provided using timestamps, however,
   time stamp based protocols require the involved nodes to have time
   clocks and for those time clocks to be both synchronized and secured.
   We propose the following mechanism to provide authentication of
   Paging Request message.  See Figure 2.

   1. When an incoming packet destined to a dormant MN arrives at the
      PF, the latter pages different access routers of the Paging Area
      by sending a Paging Request multicast message.

   2. The Paging message may contain the session key shared between the
      MN and the network; or a receiving AR may retrieve it from a Key
      Storage Center.

   3. The Access router generates a random number R, and creates a
      sequence number N1.  This sequence number is user and router
      specific and must only increase in value.

      The Access Router computes a Token based at least on R, N1, K and
      a common algorithm shared with the MN: Token (N1, R, K).

      The access router encrypts N1 using K, K[N1], and sends the Token
      (N1, R, K), the random number, R, and the encrypted N1 to the MN
      for network authentication.

   4. On receipt of the IP paging request, the MN

      - deciphers N1 using K (K[N1), and verifies its validity:  N1 must
      always increase in value; this will ensure the freshness of the
      message

      - verifies the token:  the MN can thus make sure the IP Paging
      Request is coming from the valid network


















Faccin, Koodli, Le, Malinen, Purnadi                           [Page 11]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


                 |
                 | 1)incoming packet
                 V
              +----+
              |    |
              | PF | PF realizes that MN is dormant
              |    | K
              +----+
                |
                | 2)Page to all AR within the IP paging area
                |
   +------------|-----------------------+
   |     +------+ -------+--------+     |
   |     |      |        |        |     |
   |     |(K)   |(K)     |(K)     |(K)  |
   |     V      V        V        V     |
   |  +----+   +----+   +----+   +----+ |
   |  | AR |   | AR |   | AR |   | AR | |
   |  +----+   +----+   +----+   +----+ |
   |     |        |^       |        |   |
   |     |        ||       |        | 3) L3 paging broadcast to
   |     |        ||       |        | all MN in each AR
   |     |        ||       |        |  Token (N1, R, K)
   |     |        ||       |        |  R, K[N1]
   |     V        V|       V        V   |
   |            +----+                  |
   |            | MN | 4) MN deciphers N1
   |            +----+    verifies Token|
   |                      keeps N1 for future
   +------------------------------------+
        IP paging area


                    Figure 2: Securing the Paging Request


   If the authentication data is computed by the PF, anti replay attacks
   may also be provided by sequence numbers shared between the MN and
   the PF. But these sequence numbers must be synchronized.  As an
   example, every time MN changes PF, the sequence number could be re-
   initiated.


7.2.  Authentication of Paging Response

   The Paging Response message SHOULD be authenticated:  this will
   provide user authentication for network access and will ensure the
   validity of the user before forwarding the incoming packets.



Faccin, Koodli, Le, Malinen, Purnadi                           [Page 12]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   Authentication of the Paging Response is based on a local session
   paging key K shared between the network and the MN. The mechanism to
   compute this key is out of the scope of this document.  Based on this
   key, the local challenge, the MN computes a token whose validity is
   verified by the PF.

   In order to expedite the network access authentication, the PF may
   supply the key to all the ARs in the paging area.  The MN, when it
   performs neighbor discovery procedures to connect to the new AR, may
   supply a token using the key and a challenge specific to the AR.
   Since the AR would have already received the session key K, it can
   perform local authentication without having to approach a AAA server,
   for example.  See [idle-mode-ct] for details.

   1. in the Paging Request message, the PF (or the AR) generates and
      includes a Local Challenge.

   2. the MN takes this Local Challenge, the local session paging key K
      and eventually some other information to compute some
      authentication data to be included in the Paging Response

   3. the PF (or the AR) verifies the validity of the Paging Response.
      If the AR verifies the Paging Response, it sends a corresponding
      Paging Response message to the PF.


7.3.  Authentication of Paging Area Update

   Paging Area Update SHOULD be authenticated.  This will limit the
   possible damages of Bogus Paging Information identified and described
   in [4].

   When MN sends a Paging Area Update, the paging information SHOULD be
   authenticated.  This authentication can be based on a local session
   paging key shared between the MN and the PF. The PF can be either the
   current one or a new one.  The MN will e.g., compute a token based on
   the Paging information, the session key and other relevant
   information.  If the PF is the current one, the PF will already have
   the session key and can verify the token.  If the PF is a new one,
   the new PF can forward the token to the previous PF to have it
   verified.  The MN then sets up a new session key with the new PF.

   As another alternative, the new PF may get the session key either
   from the previous PF or from a Key Storage Center, and then perform
   authentication of the paging messages.

   Anti replay attacks may be provided using:




Faccin, Koodli, Le, Malinen, Purnadi                           [Page 13]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


                 |
                 | 1)incoming packet
                 V
              +----+
              |    |
              | PF | PF realizes that MN is dormant
              |    | K
              +----+
                |
                | 2)Page to all AR within the IP paging area
                |
   +------------|-----------------------+
   |     +------+ -------+--------+     |
   |     |      |        |        |     |
   |     |(K)   |(K)     |(K)     |(K)  |
   |     V      V        V        V     |
   |  +----+   +----+   +----+   +----+ |
   |  | AR |   | AR |   | AR |   | AR | |
   |  +----+   +----+   +----+   +----+ |
   |     |        |^       |        |   |
   |     |        ||       |        | 3) L3 paging broadcast to
   |     |        ||       |        | all MN in each AR
   |     |        ||       |        |  Local Challenge
   |     |        ||       |        |   |
   |     V        V| 4) Paging Response |
   |            +----+  authentication data
   |            | MN |  (LC, K)         |
   |            +----+                  |
   |                                    |
   +------------------------------------+
        IP paging area



                    Figure 3: Securing the Paging Response


   -  timestamps, but time stamp based protocols require the involved
      nodes to have time clocks and for those time clocks to be both
      synchronized and secured.

   -  sequence numbers but these sequence numbers must be synchronized.
      As an example, every time MN changes PF, the sequence number could
      be re-initiated.

   -  random numbers generated by the authenticating PF: the MN first
      send a Paging Area Update and the authentication PF will, in
      response, provide this random number.  The MN will resend the



Faccin, Koodli, Le, Malinen, Purnadi                           [Page 14]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


      Paging Are Update containing authentication data computed from the
      Paging information, and the random number.  This method therefore
      requires additional messages.


7.4.  Filtering

   Filtering MAY be applied to reduce the potential DoS Amplification
   damages identified and described in [4] and will allow detection of
   Bogus CN.

   Filtering can be applied on the source or on the types of the
   packets:

   -  The MN specifies sources of packets for which it should woken up.
      Packets from sources not authorized will not generate paging
      request when they reach Paging Function.

   -  The MN can also specify the type of packets it should be woken up
      for.

   The MN can send the filtering information to the network at IP Paging
   Area Update; and MN can update them at paging response or at any
   time.  The network may also have some pre-defined filtering rules of
   its own.

   Filtering allows a flexible handling of incoming packets.  Based on
   the filtering information, the packets can be:

   -  discarded:  packets are discarded and MN will not receive them

   -  buffered:  packets are buffered up to maximum size of per-MN FIFO
      buffer and provided to the MN only when the MN becomes active.
      Buffer overflow can be handled by storing only the recent packets
      or according to other criteria (e.g.  "priority")

   -  grouped:packets are buffered until a certain amount.  When that
      amount is reached, MN is paged and the packets delivered, or

   -  processed based on other rules.

   Efficiency of filtering is strongly related to the ability to verify
   the type or source of packets:  A bogus CN can, e.g., send packets
   using fake source addresses.  So there needs to be a mechanism to
   "verify" the source.  The MN may be able to establish a Security
   Association with CN for which it wishes to be woken up (e.g., a SIP
   server) and provide filtering function with security parameters so
   that it can verify source authenticity for packet from CNs.



Faccin, Koodli, Le, Malinen, Purnadi                           [Page 15]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   Filtering assumes that MN knows in advance what CN will initiate a
   Mobile Terminated communication.  This may be acceptable in some
   specific environments.  MN knows in advance what Mobile Terminated
   services will take place (e.g.  SIP calls from a Call Control
   Function it has previously registered with, IMPP from server it has
   previously provided presence information to, e-mail server it has
   previously "registered" with, etc.); but this may not always be
   applicable and more particularly for push services to the MN from
   unknown sources.


8.  Application of IP Paging Model to Different Scenarios

   The IP paging model can be applied to several different scenarios,
   whether IP mobility is present or not.  In case IP mobility
   mechanisms are present, the IP paging model can be applied to
   different scenarios.  Scenarios for Mobile IPv4 are not described in
   this section, but the IP paging model can be applied to Mobile IPv4
   as well.

   Figure 4 represents the scenario where Mobile IPv6 is adopted in the
   network.  In such a case, the PF can be either in the Access Router
   (AR) or it can be a node in the network not related to mobility. More
   than one PFs for a given IP-PA can be used. For a given MN connected
   to a given AR in a given IP-PA and goes dormant in the IP-PA, only
   one entity acts as the PF for the MN. Other dormant MNs in the IP-PA
   may be served by different PFs
























Faccin, Koodli, Le, Malinen, Purnadi                           [Page 16]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


              +-----------------------------+
             /                               \
            /                                 \
   +----+  +          IP Network               +
   | PF |------->                              |
   +----+  |                                   |
      |    +                                   +
      |     \                                 /
      |      \                               /
      |       +-----------------------------+
      |
      |---------+-------+-----------+-------+-------+
      |         |       |           |       |       |
      |   IP Paging Area|           |IP Paging Area |
   +--|---------|-------|---+  +----|-------|-------|---+
   | +V---+  +--V-+  +--V-+ |  | +--V-+  +--V-+  +--V-+ |
   | | AR |  | AR |  | AR | |  | | AR |  | AR |  | AR | |
   | +----+  +----+  +----+ |  | +----+  +----+  +----+ |
   +------------------------+  +------------------------+



                          Figure 4: MIPv6 scenario.

   Figure 5 represents the scenario where Mobile IPv6 Regional
   Registration is adopted in the network.  In such a case, the PF can
   be either in the Access Router (AR), in one of the Crossover Routers,
   in the GMA or it can be a node in the network not related to mobility
   (e.g.  in case of implicit dormancy).  In the Figure, CR is a Cross-
   over Router. Similarly, more than one PFs for a given IP-PA can be
   used. For a given MN connected to a given AR in a given IP-PA and
   goes dormant in that IP-PA, only one entity acts as the PF for the
   MN. Other dormant MNs in the IP-PA may be served by different PFs


















Faccin, Koodli, Le, Malinen, Purnadi                           [Page 17]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


      +------------------>+----+
      |       +-----------|GMA |------------+
      |      /            +----+             \
      |     /    +----+              +----+   \ IP
   +----+  +     | CR |              | CR |    + Network
   | PF |------->+----+              +----+    |
   +----+  |  +----+  +----+    +----+  +----+ |
      |    +  | CR |  | CR |    | CR |  | CR | +
      |     \ +----+  +----+    +----+  +----+/
      |      \                               /
      |       +-----------------------------+
      |
      |---------+-------+-----------+-------+-------+
      |         |       |           |       |       |
      |   IP Paging Area|           |IP Paging Area |
   +--|---------|-------|---+  +----|-------|-------|---+
   | +V---+  +--V-+  +--V-+ |  | +--V-+  +--V-+  +--V-+ |
   | | AR |  | AR |  | AR | |  | | AR |  | AR |  | AR | |
   | +----+  +----+  +----+ |  | +----+  +----+  +----+ |
   +------------------------+  +------------------------+



                    Figure 5: MIPv6 Regional Registration.

   Figure 6 represents the scenario where Hierarchical Mobile IPv6 is
   adopted in the network.  In such a case, the PF can be either in the
   Access Router (AR), in the MAP, or it can be a node in the network
   not related to mobility.

   In scenarios where PF is distributed to the ARs, each AR acts as a PF
   for a given MN in a given moment.  The simplest approach to implement
   a distributed PF is by defining as PF for a given MN the latest AR
   where the MN is connected before the MN becomes IP dormant
   (explicitly or implicitly).  As shown in Figure 7, the PF forwards IP
   paging messages to all ARs within the IP-PA. Each AR will forward the
   IP paging to all the MNs connected to the AR with the appropriate
   interactions between IP Paging and L2 paging, when L2 paging is
   present.












Faccin, Koodli, Le, Malinen, Purnadi                           [Page 18]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


      +------------------>+----+
      |       +-----------|MAP |------------+
      |      /            +----+             \
      |     /                                 \
   +----+  +         IP Network                +
   | PF |------->                              |
   +----+  |                                   |
      |    +                                   +
      |     \                                 /
      |      \                               /
      |       +-----------------------------+
      |
      |---------+-------+-----------+-------+-------+
      |         |       |           |       |       |
      |   IP Paging Area|           |IP Paging Area |
   +--|---------|-------|---+  +----|-------|-------|---+
   | +V---+  +--V-+  +--V-+ |  | +--V-+  +--V-+  +--V-+ |
   | | AR |  | AR |  | AR | |  | | AR |  | AR |  | AR | |
   | +----+  +----+  +----+ |  | +----+  +----+  +----+ |
   +------------------------+  +------------------------+



                          Figure 6: HMIPv6 Scenario.

   In hierarchical IP mobility solutions such as MIPv6RR, the PF can be
   located in any of the mobile agents.  In particular, the PF can be
   distributed to the ARs, to the CRs or implemented in the GMA. If PF
   is in the GMA, the CR and AR do not maintain any state for a MN when
   MN is IP dormant.  MIPv6RR is used, as currently specified, only when
   MN is not IP dormant.  If the PF is in a CR, IP-PA covers all ARs
   "below" CR and or even larger area.  "Above" the CR, all mobility-
   aware routers (CRs and GMA) maintain the state for the MN (normal
   MIPv6RR). If the PF is in the AR, the scenario is the same as the
   MIPv6 or flat LMM. Similar to the PF in a CR, all mobility-aware
   routers (CRs and GMA) maintain the state for the MN.

   The optimal location of the PF depends on relation with IP mobility
   and LMM and may depend on type of the terminal and the user.


9.  Comparing Against the Requirements

   The proposed paging concept is compared to the draft-ietf-seamoby-
   paging- requirements-01.txt






Faccin, Koodli, Le, Malinen, Purnadi                           [Page 19]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


                  |t4:incoming packet routed
                  |last point of attachment(AR2)
                  |
       +----------|------------------------+  +---------------------+
       |          |   IP-PA 1              |  |       IP-PA 2       |
       |   _____  V t5:IP paging to all ARs in|IP-PA1               |
       |  /     \  /    \      \      \    |  |                     |
       | +---+  +---+  +---+  +---+  +---+ |  | +---+  +---+  +---+ |
       | |AR1|  |AR2|  |AR3|  |AR4|  |AR5| |  | |AR6|  |AR7|  |AR8| |
       | +---+  +---+  +---+  +---+  +---+ |  | +---+  +---+  +---+ |
       +-----------------------------------+  +---------------------+
          |       |      |
          |       |      |
          |       |     t3:moves to AR3, no L3 mobility procedure
          |      t1:moves with L3 mobility procedure (active handoff)
    t0:MN point  t2:goes dormant, AR2 becomes PF
   of attachment
      is AR1



         Figure 7: IP paging applied to the distributed PF scenario.



9.1.  Impact on Power Consumption

   The IP paging protocol MUST minimize impact on the Host's dormant
   mode operation, in order to minimize excessive power drain.

   The proposed paging concept utlilizes IP-PA (IP Paging Area) where
   the MN does not have to wake up and execute L3 procedure while inside
   an IP-PA to make the MN wakes up less frequently.

   It allows further optimizations in terms of power saving and reduced
   signaling message exchange when adopted in conjunction with IP
   mobility mechanisms.

   It also allows to optimize the interaction between IP paging and L2
   paging mechanisms when present.  As an example, with the proposed
   model an IP dormant MN does not need to perform any L2 mobility
   signaling when moving within the same IP paging area


9.2.  Scalability

   The IP paging protocol MUST be scalable to millions of Hosts.
   Distributing and allowing flexible location of the paging function in



Faccin, Koodli, Le, Malinen, Purnadi                           [Page 20]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   flat or hierarchical mobility scenarios as well in non-mobile
   networks in the proposed paging concept provides scalability


9.3.  Control of Broadcast/Multicast/Anycast

   The protocol SHOULD provide a filter mechanism to allow a Host prior
   to entering dormant mode to filter which broadcast/multicast/anycast
   packets active a page.  This prevents the Host from awakening out of
   dormant mode for all broadcast/multicast/anycast traffic.

   The paging function is capable of filtering the paging message (based
   on explicit information provided by the MN)


9.4.  Efficient Signaling for Inactive Mode

   The IP paging protocol SHOULD provide a mechanism for the Tracking
   Agent to determine whether the Host is in inactive mode, to avoid
   paging when a host is completely unreachable.

   The proposed paging concept introduces explicit and implicit
   dormancy.  In the explicit dormancy, the MN explicitly sends a
   message to the tracking agent (as one of the functions in the PF) to
   indicate that it is going to be dormant.  In the implicit dormancy,
   the PF assumes the MN is dormant after the lack of activity from the
   MN goes beyond a certain threshold (e.g.  a timer).


9.5.  No Routers

   Since the basic issues involved in handling mobile routers are not
   well understood and since mobile routers have not exhibited a
   requirement for paging, the IP paging protocol MAY NOT support
   routers.  However, the IP paging protocol MAY support a router acting
   as a Host.

   The proposed paging concept does not have mobile routers.


9.6.  Multiple Dormant Mode

   Recognizing that there are multiple possible dormant modes on the
   Host, the IP paging protocol MUST work with different implementations
   of dormant mode on the Host.

   The solution does not presently describe different dormancy modes
   however, using the ability of a MN to provide information to the PF



Faccin, Koodli, Le, Malinen, Purnadi                           [Page 21]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   regarding its requirements for the dormancy either during the paging
   area update or explicit IP paging registration, such scenarios can be
   easily supported.


9.7.  Independence of Mobility Protocol

   Recognizing that IETF may support multiple mobility protocols in the
   future and that paging may be of value to hosts that do not support a
   mobility protocol, the IP paging protocol MUST be designed so there
   is no dependence on the underlying mobility protocol or on any
   mobility protocol at all.  The protocol SHOULD specify and provide
   support for a mobility protocol, if the Host supports one.

   The proposed paging concept does not depend on any existing mobility
   protocol to function.  On the other hand, it can enhance existing
   mobility protocols.  For instance, when the MN responds to a Paging
   Request message to its AR, it can include Binding Update as
   encapsulation in the Paging Response message.  The Paging Response
   message can also include a request for transfer of network contexts
   from the MN's previous AR. Furthermore, our support for securing
   paging messages can be used for expedited network access
   authentication.


9.8.  Support for Existing Mobility Protocols

   The IP paging protocol MUST specify the binding to the existing IP
   mobility protocols, namely mobile IPv4 [2] and mobile IPv6 [3].  The
   IP paging protocol SHOULD make use of existing registration support.

   The proposed paging concept works with Mobile IP. The PF can be co-
   located in any Mobility Agent, including the Foreign Agent.  The
   Paging Registration message can be constructed as an extension to the
   Registration Request or a Binding Update message.

   Furthermore, when the PF is realized in a Mobility Agent, the
   expiration of lifetime field in the binding cache can serve as a
   trigger for determining the implicit dormancy of the MN. For example,
   when PF is realized in a GMA, the GMA's regional binding cache can
   provide the input for determining the implicit dormancy of the MN.


9.9.  Dormant Mode Termination

   Upon receipt of a page (either with or without an accompanying L3
   packet), the Host MUST execute the steps in its mobility protocol to
   re-establish a routable L3 link with the Internet.



Faccin, Koodli, Le, Malinen, Purnadi                           [Page 22]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   When the MN wakes up, either due to entering a new IP-PA, the need to
   send IP packets or as a response to a page, the MN executes the steps
   in the mobility protocol and re-establishese a routable L3 link with
   the Internet.  For instance, when the MN receives a router
   advertisement with paging extension as an L3 Paging Request message,
   one of the things the MN does is to formulate a new CoA and it could
   include the Binding Update in the Paging Response message as
   encapsulation.


9.10.  Network Update

   Recognizing that locating a dormant mode mobile requires the network
   to have a rough idea of where the Host is located, the IP paging
   protocol SHOULD provide the network a way for the Paging Agent to
   inform a dormant mode Host what paging area it is in and the IP
   paging protocol SHOULD provide a means whereby the Host can inform
   the Target Agent when it changes paging area.  The IP paging protocol
   MAY additionally provide a way for the Host to inform the Tracking
   Agent what paging area it is in at some indeterminate point prior to
   entering dormant mode.

   The proposed paging concept has the L3 procedure to update the paging
   function and all mobility agent 'above' the paging function when the
   MN changes the IP-PA.


9.11.  Efficient Utilization of L2

   Recognizing that many existing wireless link protocols support paging
   at L2 and that these protocols are often intimately tied into the
   Host's dormant mode support, the IP paging protocol SHOULD provide
   support to efficiently utilize an L2 paging protocol if available.

   The proposed concept has been developed in order to work with all the
   currently known L2 paging mechanisms in the different link layer
   technologies.  The concept allows access-independent L3 paging and
   optimized interworking with L2 paging .  As an example, with the
   proposed model an IP dormant MN does not need to perform any L2
   mobility signaling when moving within the same IP paging area.  The
   concept minimizes impact on L2 paging and link layer technologies by
   allowing for IP paging signaling and advertisement of IP paging
   information through optimizations at L2 Moreover, it provides
   improvements in terms of power saving and minimization of mobility
   signaling with respect to the presence of L2 paging only.






Faccin, Koodli, Le, Malinen, Purnadi                           [Page 23]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


9.12.  Orthogonality of Paging Area Subnets

   The IP paging protocol MUST allow an arbitrary mapping between
   subnets and paging areas.

   The proposed paging concept does not associate the paging area and
   subnets.  The proposed paging concept allows an IP-PA to have
   multiple subnet and each AR is allowed to belong to different types
   of IP-PA


9.13.  Future L3 Paging Support

   Recognizing that future dormant mode and wireless link protocols may
   be designed that more efficiently utilize IP, the IP paging protocol
   SHOULD NOT require L2 support for paging.

   The concept does not require L2 support for paging and, thanks to the
   introduction of the IP paging Identity allow for future L3 paging
   only.


9.14.  Robustness Against Failure of Network Elements

   The IP paging protocol MUST be designed to be robust with respect to
   failure of network elements involved in the protocol.  The self-
   healing characteristics SHOULD NOT be any worse than existing routing
   protocols.

   The distributed and flexible location of paging function contributes
   to the robustness against failure of network elements.


9.15.  Reliability of Packet Delivery

   The IP paging protocol MUST be designed so that packet delivery is
   reliable to a high degree of probability.  This does not necessarily
   mean that a reliable transport protocol is required.

   The paging request will be resent if a response is not received after
   PAGE_RESPONSE_RX_TIME expires on the access network.  The re-
   transmission are performed for at most PAGE_RQST_RE_TX number of
   times.  When L2 Paging is initiated on the access link as a result of
   receiving an IP Paging request, we rely on the underlying L2 protocol
   to reliably page the MN.






Faccin, Koodli, Le, Malinen, Purnadi                           [Page 24]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


9.16.  Robustness Against Message Loss

   The IP paging protocol MUST be designed to be robust with respect to
   loss of messages.

   The paging message will be resent if it is not acknowledged.  The
   reduction of the number of message during the IP dormant state
   naturally reduce the possibility of message lost


9.17.  Flexibility of Administration

   The IP paging protocol SHOULD provide a way to flexibly auto-
   configure Paging Agents to reduce the amount of administration
   necessary in maintaining a wireless network with paging.

   The proposed paging concept can configure the paging function to be
   located in any mobility agent (another access router in the flat LMM
   or in any access router, any cross-over router or the GMA in the
   hierarchical MIPv6RR). Moreover, dynamic IP Paging Areas are
   supported, where the size and scope of the areas can be modified
   dynamically.


9.18.  Flexibility of Paging Area Design

   The IP paging protocol MUST be flexible in the support of different
   types of paging areas.  Examples are fixed paging areas, where a
   fixed set of bases stations belong to the paging area for all Hosts,
   and customized paging areas, where the set of base stations is
   customized for each Host.

   The proposed paging concept is flexible enough to accommodate the
   concept of hierarchical overlapping paging area and the paging area
   with multiple access technologies.  Moreover, dynamic IP Paging Areas
   are supported, where the size and scope of the areas can be modified
   dynamically.


9.19.  Availability of Security Support

   The IP paging protocol MUST have available authentication and
   encryption functionality at least equivalent to that provided by
   IPSEC.

   The proposed IP paging security solutions support authentication and
   encryption functionality, equivalent to that provided by IPsec




Faccin, Koodli, Le, Malinen, Purnadi                           [Page 25]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


9.20.  Authentication of Paging Location Registration

   The IP paging protocol MUST provide mutually authenticated paging
   location registration to insulate against replay attacks and to avoid
   the danger of malicious nodes registering for paging.

   The proposed paging concept provides mutual authenticated paging
   location registration (both Paging area update and Paging Area
   Response are authenticated) with anti replay attacks so the network
   can make sure the user is valid and the user can also make sure he is
   communication with the valid network


9.21.  Authentication of Paging Area Information

   The IP paging protocol MUST provide a mechanism for authenticating
   paging area information distributed by the Paging Agent.

   Authentication of PAU and PAR prevent possible attacks from bogus
   paging area information.  If the Paging Area information are
   incorrect, the PAR will tell the MN. In such case, PAR and PAU may
   still be sent unnecessarily over the access link.  Authentication of
   the advertised paging area information would avoid that but there may
   be some issues with anti replay attacks if time stamps are not valid.


9.22.  Authentication of Paging Messages

   The IP paging protocol MUST provide a mechanism for authenticating L3
   paging messages sent by the Paging Agent to dormant mode Hosts.  The
   protocol MUST support the use of L2 security mechanisms so
   implementations that take advantage of L2 paging can also be secured.

   L3 Paging messages sent by the Agent (IP Paging Request, Paging Area
   Response, etc.)  are authenticated; and if L2 security is present, it
   will enhance the security level and make it even more difficult for
   an intruder to perform the desired type of attacks


9.23.  Paging Volume

   The IP paging protocol SHOULD be able to handle large numbers of
   paging requests without denying access to any legitimate Host nor
   degrading its performance.

   The proposed paging concept can handle a large number of paging
   request without denying access to any legitimate Host nor degrading
   its performance.  The limiting factor may be the L2 (wireless link)



Faccin, Koodli, Le, Malinen, Purnadi                           [Page 26]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


10.  References

   [1]       N. Asokan, Patrik Flykt, C. Perkins, and Thomas Eklund.
             AAA for ipv6 Network Access (work in progress).  Internet
             Draft, Internet Engineering Task Force, March 2000.

   [2]       J. Kempf.  Dormant mode host alerting (ip paging) problem
             statement.  Request for Comments (Informational) 3132,
             Internet Engineering Task Force, June 2001.

   [3]       O. Levkowetz and et al.  Problem description:  Reasons for
             doing context transfers between nodes in an ip access
             network (work in progress).  Internet Draft, Internet
             Engineering Task Force, February 2001.

   [4]       P. Mutaf, "IP Paging Security Requirements", Internet
             draft, Internet Engineering Task Force, May 2001


































Faccin, Koodli, Le, Malinen, Purnadi                           [Page 27]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


11.  Authors' Addresses


   Stefano M. Faccin                  Rajeev Koodli
   Mobile Networks Lab                Communications Systems Lab
   Nokia Research Center              Nokia Research Center
   6000 Connection Drive              313 Fairchild Drive
   Irving, Texas 75039                Mountain View, California 94043
   USA                                USA
   Phone:  +1 972 894-4994            Phone:  +1-650 625-2359
   EMail:  stefano.faccin@nokia.com   EMail:  rajeev.koodli@nokia.com
   Fax:  +1 972 894-4589              Fax:  +1 650 625-2502

   Franck Le                          Jari T. Malinen
   Mobile Networks Lab                Communications Systems Lab
   Nokia Research Center              Nokia Research Center
   6000 Connection Drive              313 Fairchild Drive
   Irving, Texas 75039                Mountain View, California 94043
   USA                                USA
   Phone:  +1 972 374-1256            Phone:  +1-650 625-2355
   EMail:  franck.le@nokia.com        EMail:  jmalinen@iprg.nokia.com
   Fax:  +1 972 894-4589              Fax:  +1 650 625-2502

   Rene Purnadi
   Mobile Networks Lab
   Nokia Research Center
   6000 Connection Drive
   Irving, Texas 75039
   USA
   Phone:  +1 972 894-4897
   EMail:  rene.purnadi@nokia.com
   Fax:  +1 972 894-4589



















Faccin, Koodli, Le, Malinen, Purnadi                           [Page 28]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


12.  Appendix A - Hierarchical Overlapping Paging Area Support

   Functional model for overlapping PAs gives no indication on
   implementation or allocation of PFs to any specific node.  According
   to different implementation options, PFs can be located all in the
   same node(i.e.  same node acts as PF for more than one PA and more
   than one type of PA). No relation between PAs at different levels

   Different mobile nodes/users have different traffic patterns and
   mobility patterns, e.g., some MNs generate very low traffic and stay
   dormant for long time (e.g.  basic voice terminals that
   generate/receive few calls a day), others are very heavy on traffic
   (e.g.  data MNs may be always on and receive/send data almost
   continuously).  Based on the different behaviours, the relation
   between mobility procedures and IP Paging is different.  For low
   traffic MN would make sense to use large IP-PA since, even if the MN
   moves a lot inside the IP-PA, the traffic patterns indicates that the
   MN will have to wake up and perform IP mobility procedure not very
   often.  Although in this case IP paging will require multicast of
   paging messages to a larger area, there is however a large saving in
   terms of the IP mobility signalling in the network and on the radio
   link otherwise needed to keep track of the MN point of attachment.

   For always-on terminals that are always in an IP session (i.e.  they
   go dormant very rarely and only for very limited periods of time), it
   would make sense to consider small IP-PA or even rely only on L2
   paging only when available.  If the MN moves often, IP mobility
   procedures will be performed quite often but, when the MN needs to
   wake-up for incoming packets, paging has less impact in terms of
   latency in delivering the packets and in terms of signalling needed
   for paging.

   This concept allows several types of IP-PA defined in the system,
   each type corresponds to a specific set of terminal capabilities/user

















Faccin, Koodli, Le, Malinen, Purnadi                           [Page 29]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   +-----+    +-----+    +-----+           +-----+    +-----+    +-----+
   |     |    |     |    |     |           |     |    |     |    |     |
   | PF1 |    | PF4 |    | PF2 |           | PF6 |    | PF5 |    | PF3 |
   |     |    |     |    |     |           |     |    |     |    |     |
   +-----+    +-----+    +-----+           +-----+    +-----+    +-----+
       \         |          |                 |           |         /
   +----\ -------|----------|-----------------|-----------|--------/---+
   |     \        \         | IP-PA 1 Type 3             /        /    |
   |+-----\ -------\ --------\ ----+  +-----------------/--------/---+ |
   ||      \ IP-PA 1 Type 2   \    |  |        IP-PA 2 Type 2   /    | |
   || +-----\ ------- ---+ +---\ -------------+ +--------------/---+ | |
   || |  IP-PA1 Type 1   | |  IP-PA 2 Type 1  | |  IP-PA 3 Type 1  | | |
   || | +--+  +--+  +--+ | | +--+  +--+  +--+ | | +--+  +--+  +--+ | | |
   || | |AR|  |AR|  |AR| | | |AR|  |AR|  |AR| | | |AR|  |AR|  |AR| | | |
   || | +--+  +--+  +--+ | | +--+  +--+  +--+ | | +--+  +--+  +--+ | | |
   || +------------------+ +------------------+ +------------------+ | |
   |+------------------------------+  +------------------------------+ |
   +-------------------------------------------------------------------+

                          |->|                 |->|              |-->
                           1                    1                  1
                                    |->|                          |-->
                                     2                              2
                                                                   |-->
                                                                     3
   1 = MN "care for" PA Type1, perform "PA Update" in this cases
   2 = MN "care for" PA Type2, perform "PA Update" in this cases
   3 = MN "care for" PA Type3, perform "PA Update" in this case



                Figure 8: Hierarchical Overlapping Paging Area

   requirements/services being used by the user.  For examples, IP-PA
   type 3 is for MNs with low traffic and long period of dormancy, IP-PA
   type 1 is for MNs with very brief periods of dormancy and almost
   continuous traffic.  The choice of the PA Type most appropriate for
   the MN can be done in several ways:

   The visited domain where the MN is roaming determines it based on the
   user profile and terminal capabilities, and indicates it to the MN.
   In such case, the MN may be indicated one or more PA types, and
   decision of which one to use at a certain point in time is based on
   MN status: if MN is inactive (no active sessions at all) MN can
   choose the PA Type that minimizes the need for mobility procedures;
   if MN is in active sessions for services that require continuous or
   frequent exchange of data (e.g.  real-time multimedia services), MN
   choose the PA Type that allows the most precise tracking of the MN.



Faccin, Koodli, Le, Malinen, Purnadi                           [Page 30]


INTERNET-DRAFT        Dormant Mode Handover Support        November 2001


   Or the MN itself makes the selection based on user preferences, user
   profile, services currently used, etc.  Other cases are possible

   Each access router "broadcasts" the IP-PA identifier(s) for the IP-
   PA(s) it belongs to.  Each Access Router may belong to one or more
   IP- PAs, but one AR belongs to only one IP-PA of a specific type.
   E.g.  if the AR belongs to three IP-PAs, it will broadcast three
   different IP-PA identifiers

   The IP-PA identifier(s) must be broadcasted by the AR in such a way
   that the MNs can easily determine what types of IP-PA are supported
   and their identifiers.  Several solutions are available.  Also IP-PA
   types need to be standardized to guarantee inter-operation at
   roaming.  The number of types of paging areas in a given network can
   be dynamic, i.e.  it does not need to be mandated by the standards.
   E.g.  this means that, if the standard defines 10 types of IP-PAs, a
   given network may adopt only 2


































Faccin, Koodli, Le, Malinen, Purnadi                           [Page 31]