SIPPING Working Group                                        Mahy/Cisco
Internet Draft                                     Campbell/dynamicsoft
Document: draft-ietf-sipping-cc-framework-00.txt      Johnston/Worldcom
February 2002                                            Petrie/Pingtel
                                                  Rosenberg/dynamicsoft
Expires: August 2002                                 Sparks/dynamicsoft


                A Multi-party Application Framework for SIP


   Status of this Memo

   This document is an Internet-Draft and is in full conformance with
   all provisions of Section 10 of RFC2026 [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.


1 Abstract

   This document defines a framework and requirements for multi-party
   applications in SIP.  To enable discussion of multi-party
   applications we define an abstract call model for describing the
   media relationships required by many of these applications.  The
   model and actions described here are specifically chosen to be
   independent of the SIP signaling and/or mixing approach chosen to
   actually setup the media relationships.  In addition to its dialog
   manipulation aspect, this framework includes requirements for
   communicating related information and events such as conference and
   session state, and session history.  This framework also describes
   other goals which embody the spirit of SIP applications as used on
   the Internet.

2 Conventions used in this document

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED",  "MAY", and "OPTIONAL" this
   document are to be interpreted as described in RFC-2119 [RFC2119].





Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg   Exp:Aug 2002 [Page 1]


                       SIP Multiparty Framework


   Table of Contents
   1   Abstract.......................................................1
   2   Conventions used in this document..............................1
   3   Motivation and Background......................................4
   3.1   Goals........................................................4
   3.2   Example Features.............................................6
   4   Key Concepts...................................................9
   4.1   "Conversation Space" Model...................................9
   4.1.1   Comparison with Related Definitions.......................10
   4.2   Signaling Models............................................11
   4.3   Mixing Models...............................................12
   4.3.1   (Single) End System Mixing................................12
   4.3.2   Centralized Mixing........................................12
   4.3.3   Multicast and Multi-unicast conferences...................14
   4.4   Conveying Information and Events............................15
   4.5   Componentization and Decomposition..........................16
   4.5.1   Media Intermediaries......................................17
   4.5.2   Queue Server..............................................18
   4.5.3   Parking Place.............................................18
   4.5.4   Announcements and Voice Dialogs...........................19
   4.6   Use of URIs.................................................21
   4.6.1   Naming Users in SIP.......................................21
   4.6.2   Naming Services with SIP URIs.............................23
   4.7   Invoker Independence........................................26
   4.8   Billing issues..............................................26
   5  Catalog of call control actions and sample features............26
   5.1   Early Dialog Actions........................................27
   5.1.1   Remote Answer.............................................27
   5.1.2   Remote Forward or Put.....................................27
   5.1.3   Remote Busy or Error Out..................................27
   5.2   Single Dialog Actions.......................................27
   5.2.1   Remote Dial...............................................28
   5.2.2   Remote On and Off Hold....................................28
   5.2.3   Remote Hangup.............................................28
   5.3   Multi-dialog actions........................................28
   5.3.1   Transfer..................................................28
   5.3.2   Take......................................................29
   5.3.3   Add.......................................................29
   5.3.4   Local Join................................................30
   5.3.5   Insert....................................................30
   5.3.6   Split.....................................................31
   5.3.7   Near-fork.................................................31
   5.3.8   Far fork..................................................31
   6   Putting it all together.......................................33
   6.1   Feature Solutions...........................................34
   6.1.1   Call Park.................................................34
   6.1.2   Call Pickup...............................................34
   6.1.3   Music on Hold.............................................35
   6.1.4   Call Monitoring...........................................35
   6.1.5   Barge-in..................................................35
   6.1.6   Intercom..................................................35
   6.1.7   Speakerphone paging.......................................36
   6.1.8   Distinctive ring..........................................36

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg   Exp:Aug 2002 [Page 2]


                       SIP Multiparty Framework

   6.1.9   Voice message screening...................................36
   6.1.10  Single Line Extension.....................................36
   6.1.11  Click-to-dial.............................................36
   6.1.12  Pre-paid calling..........................................37
   6.1.13  Voice Portal..............................................37
   7   Security Considerations.......................................38
   8   References....................................................39
   9   Acknowledgments...............................................41
   10   Author's Addresses...........................................41














































Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg   Exp:Aug 2002 [Page 3]


                       SIP Multiparty Framework


3 Motivation and Background


   The Session Initiation Protocol [SIP] was defined for the
   initiation, maintenance, and termination of sessions or calls
   between one or more users.  However, despite its origins as a large-
   scale multiparty conferencing protocol, SIP is used today primarily
   for point to point calls.  This two-party configuration is the focus
   of the SIP specification and most of its extensions.

   This document defines a framework and requirements for multi-party
   applications in SIP.  Most multi-party applications manipulate SIP
   dialogs (also known as call legs) to cause participants in a
   conversation to perceive specific media relationships.  In other
   protocols that deal with the concept of calls, this manipulation is
   known as call control.  In addition to its dialog manipulation
   aspect, "call control" also includes communicating information and
   events related to manipulating calls, including information and
   events dealing with session state and history, conference state,
   user state, and even message state.

3.1 Goals
   Based on input from the SIP community, the authors compiled the
   following set of goals for SIP call control:

   - Define Primitives, Not Services.  Allow for a handful of robust
   yet simple mechanisms which can be combined to deliver features and
   services.  Throughout this document we refer to these simple
   mechanisms as "primitives".  Primitives should be sufficiently
   robust that when they are combined they can be used to build lots of
   services.  However, the goal is not to define a provably complete
   set of primitives. Note that while the IETF will NOT standardize
   behavior or services, it may define example services for
   informational purposes, as in [service examples].

   - Participant oriented.  The primitives should be designed to
   provide services which are oriented around the experience of the
   participants.  The authors observe that end users of features and
   services usually don't care how a media relationship is setup.
   Their ultimate experience is based only on the resulting media and
   other externally visible characteristics.

   - Signaling Model independent: Support both a central control and a
   peer-to-peer feature invocation model (and combinations of the two).
   baseline SIP already supports a centralized control model described
   in [3pcc], and the SIP community has expressed a great deal of
   interest in peer-to-peer or distributed call control.  Some such
   primitives are already defined in [REFER] and [Replaces].

   - Mixing Model independent: The bulk of interesting multiparty
   applications involve mixing or combining media from multiple
   participants.  This mixing can be performed by one or more of the
   participants, or by a centralized mixing resource.  The experience

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg   Exp:Aug 2002 [Page 4]


                       SIP Multiparty Framework

   of the participants should not depend on the mixing model used.
   While most examples in this document refer to audio mixing, the
   framework applies to any media type.  In this context a "mixer"
   refers to combining media in an appropriate, media-specific way.

   - Invoker oriented. Only the user who invokes a feature or a service
   needs to know exactly which service is invoked or why.  This is good
   because it allows new services to be created without requiring new
   primitives from all the participants; and it allows for much simpler
   feature authorization policies, for example, when participation
   spans organizational boundaries.  As discussed in section 4.7, this
   also avoids exponential state explosion when combining features.
   The invoker only has to manage a user interface or API to prevent
   local feature interactions.  All the other participants simply need
   to manage the feature interactions of a much smaller number of
   primitives.

   - Primitives make full use of URIs.  URIs are a very powerful
   mechanism for describing users and services.  They represent a
   plentiful resource which can be extremely expressive and easily
   routed, translated, and manipulated--even across organizational
   boundaries.  URIs can contain special parameters and informational
   headers which need only be relevant to the owner of the namespace
   (domain) of the URI.  Just as a user who selects an http: URL need
   not understand the significance and organization of the web site it
   references, a user may encounter a SIP URL which translates into an
   email-style group alias, which plays a pre-recorded message, or runs
   some complex call-handling logic.

   - Make use of SIP headers and SIP event packages to provide SIP
   entities with information about their environment.  These should
   include information about the status / handling of dialogs on other
   user agents, information about the history of other contacts
   attempted prior to the current contact, the status of participants,
   the status of conferences, user presence information, and the status
   of messages.

   - Encourage service decomposition, and design to make use of
   standard components using well-defined, simple interfaces.  Sample
   components include a media mixer, recording service, announcement
   server, and voice dialog server.  (This is not an exhaustive list).

   - Include authentication, authorization, policy, logging, and
   accounting mechanisms to allow these primitives to be used safely
   among mutually untrusted participants.  Some of these mechanisms may
   be used to assist in billing, but no specific billing system will be
   endorsed.

   - Permit graceful fallback to baseline SIP.  Definitions for new SIP
   call control extensions/primitives MUST describe a graceful way to
   fallback to baseline SIP behavior. Support for one primitive MUST
   NOT imply support for another primitive.



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg   Exp:Aug 2002 [Page 5]


                       SIP Multiparty Framework

   - Do not reinvent traditional models, such as the model used the
   H.450 family of protocols, JTAPI, or the CSTA call model.  In the
   opinion of the authors, these models share more characteristics of
   the traditional telephone network than with SIP.  As these other
   models do not share the design goals presented in this document, it
   would be a disservice to these other protocols and SIP to try to
   shoehorn our new design goals into an existing model.


3.2 Example Features

   Primitives are defined in terms of their ability to provide
   features.  These example features should require an amply robust set
   of services to demonstrate a useful set of primitives.  They are
   described here briefly. Note that the descriptions of these features
   are non-normative.  Some of these features are used as examples in
   section 6 to demonstrate how some features may require certain media
   relationships.  Note also that this document describes a mixture of
   both features originating in the world of telephones, and features
   which are clearly Internet oriented.

   Example Features:

   Call Waiting - Alice is in a call, then receives another call.
   Alice can place the first call on hold, and talk with the other
   caller.  She can typically switch back and forth between the
   callers.

   Blind Transfer - Alice is in a conversation with Bob.  Alice asks
   Bob to contact Carol, but makes no attempt to contact Craol
   independently.  In many implementations, Alice does not verify Bob's
   success or failure in contacting Carol.

   Attended Transfer - The transferring party establishes a session
   with the transfer target before completing the transfer.

   Consultative transfer - the transferring party establishes a session
   with the target and mixes both sessions together so that all three
   parties can participate, then disconnects leaving the transferee and
   transfer target with an active session.

   Conference Call - Three or more active, visible participants in the
   same conversation space.

   Call Park - A call participant parks a call (essentially puts the
   call on hold), and then retrieves it at a later time (typically from
   another location).

   Call Pickup - A party picks up a call that was ringing at another
   location.  One variation allows the caller to choose which location,
   another variation just picks up any call in that user's "pickup
   group".



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg   Exp:Aug 2002 [Page 6]


                       SIP Multiparty Framework

   Music on Hold - When Alice places a call with Bob on hold, it
   replaces its audio with streaming content such as music,
   announcements, or advertisements.

   Call Monitoring - A call center supervisor joins an in-progress call
   for monitoring purposes.

   Barge-in - Carol interrupts Alice who has a call in-progress call
   with Bob.  In some variations, Alice forcibly joins a new
   conversation with Carol, in other variations, all three parties are
   placed in the same conversation (basically a 3-way conference).

   Hotline - Alice picks up a phone and is immediately connected to the
   technical support hotline, for example.

   Autoanswer - Calls to a certain address or location answer
   immediately via a speakerphone.

   Intercom - Alice typically presses a button on a phone which
   immediately connects to another user or phone and casues that phone
   to play her voice over its speaker.  Some variations immediately
   setup two-way communications, other variations require another
   button to be pressed to enable a two-way conversation.

   Speakerphone paging - Alice calls the paging address and speaks.
   Her voice is played on the speaker of every idle phone in a
   preconfigured group of phones.

   Speed dial - Alice dials an abbreviated number, or enters an alias,
   or presses a special speed dial button representing Bob.  Her action
   is interpreted as if she specified the full address of Bob.

   Call Return - Alice calls Bob.  Bob misses the call or is
   disconnected before he is finished talking to Alice.  Bob invokes
   Call return which calls Alice, even if Alice did not provide her
   real identity or location to Bob.

   Inbound Call Screening - Alice doesn't want to receive calls from
   Matt.  Inbound Screening prevents Matt from disturbing Alice.  In
   some variations this works even if Matt hides his identity.

   Outbound Call Screening - Alice is paged and unknowingly calls a
   PSTN pay-service telephone number in the Carribean, but local policy
   blocks her call, and possibly informs her why.

   Call Forwarding - Before a call-leg is accepted it is redirected to
   another location, for example, because the originally intended
   recipient is busy, does not answer, is disconnected from the
   network, configured all requests to go soemwhere else.

   Message Waiting - Bob calls Alice when she steps away from her
   phone, when she returns a visible or audible indicator conveys that
   someone has left her a voicemail message.  The message waiting


Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg   Exp:Aug 2002 [Page 7]


                       SIP Multiparty Framework

   indication may also convey how many messages are waiting, from whom,
   what time, and other useful pieces of information.

   Do Not Disturb - Alice selects the Do Not Disturb option.  Calls to
   her either ring briefly or not at all and are forwarded elsewhere.
   Some variations allow specially authorized callers to override this
   feature and ring Alice anyway.

   Distinctive ring - Incoming calls have different ring cadences or
   sample sounds depending on the From party, the To party, or other
   factors.

   Automatic Callback: Alice calls Bob, but Bob is busy.  Alice would
   like Bob to call her automatically when he is available.  When Bob
   hangs up, alice's phone rings. When Alice answers, Bob's phone
   rings.  Bob answers and they talk.

   Find-Me - Alice sets up complicated rules for how she can be reached
   (possibly using [CPL], [presence] or other factors).  When Bob calls
   Alice, his call is eventually routed to a temporary Contact where
   Alice happens to be available.

   Whispered call waiting - Alice is in a conversation with Bob.  Carol
   calls Alice.  Either Carol can "whisper" to Alice directly ("Can you
   get lunch in 15 minutes?"), or an automaton whispers to Alice
   informing her that Carol is trying to reach her.

   Voice message screening - Bob calls Alice.  Alice is screening her
   calls, so Bob hears Alice's voicemail greeting.  Alice can hear Bob
   leave his message.  If she decides to talk to Bob, she can take the
   call back from the voicemail system, otherwise she can let Bob leave
   a message. This emulates the behavior of a home telephone answering
   machine

   Presence-Enabled Conferencing: Alice wants to set up a conference
   call with Bob and Cathy when they all happen to be available (rather
   than scheduling a predefined time).  The server providing the
   application monitors their status, and calls all three when they are
   all "online", not idle, and not in another call.

   IM Conference Alerts: A user receives an notification as an Instant
   Message whenever someone joins a conference they are also in.

   Single Line Extension -- A group of phones are all treated as
   "extensions" of a single line. A call for one rings them all.  As
   soon as one answers, the others stop ringing.  If any extension is
   actively in a coversation, another extension can "pick up" and
   immediately join the conversation. This emulates the behavior of a
   home telephone line with multiple phones.

   Click-to-dial - Alice looks in her company directory for Bob.  When
   she finds Bob, she clicks on a URL to call him.  Her phone rings (or
   possibly answers automatically), and when she answers, Bob's phone
   rings.

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg   Exp:Aug 2002 [Page 8]


                       SIP Multiparty Framework


   Pre-paid calling - Alice pays for a certain currency or unit amount
   of calling value.  When she places a call, she provides her account
   number somehow.  If her account runs out of calling value during a
   call her call is disconnected or redirected to a service where she
   can purchase more calling value.

   Voice Portal - A service that allows users to access a portal site
   using spoken dialog interaction.  For example, Alice needs to
   schedule a working dinner with her co-worker Carol. Alice uses a
   voice portal to check Carol's flight schedule, find a restauraunt
   near her hotel, make a reservation, get directions there, and page
   Carol with this information.


4 Key Concepts

4.1 "Conversation Space" Model

   This document introduces the concept of an abstract "conversation
   space" (essentially as a set of participants who believe they are
   all communicating among one another).  Each conversation space
   contains one or more participants.

   Participants are SIP User Agents which send original media to or
   terminate and receive media from other members of the conversation
   space.  Logically, every participant in the conversation space has
   access to all the media generated in that space (this is strictly
   true if all participants share a common media type).  A SIP User
   Agent which does not contribute or consume any media is NOT a
   participant; nor is a user agent which merely forwards, transcodes,
   mixes, or selects media originating elsewhere in the conversation
   space.  [Note that a conversation space consists of zero or more SIP
   calls or SIP conferences.  A conversation space is similar to the
   definition of a "call" in some other call models.]

   Participants may represent human users or non-human users (referred
   to as robots or automatons in this document).  Some participants may
   be hidden within a conversation space. Some examples of hidden
   participants include: robots which generate tones, images, or
   announcements during a conference to announce users arriving and
   departing, a human call center supervisor monitoring a conversation
   between a trainee and a customer, and robots which record media for
   training or archival purposes.

   Participants may also be active or passive.  Active participants are
   expected to be intelligent enough to leave a conversation space when
   they no longer desire to participate.  (An attentive human
   participant is obviously active.)  Some robotic participants (such
   as a voice messaging system, an instant messaging agent, or a voice
   dialog system) may be active participants if they can leave the
   conversation space when there is no human interaction.  Other robots
   (for example our tone generating robot from the previous example)
   are passive participants.  A human participant "on-hold" is passive.

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg   Exp:Aug 2002 [Page 9]


                       SIP Multiparty Framework


   An example diagram of a conversation space can be shown as a
   "bubble" or ovals, or as a "set" in curly or square brace notation.
   Each set, oval, or "bubble" represents a conversation space. Hidden
   participants are shown in lowercase letters.


   { A , B }          [ A , B ]

      .-.                 .---.
     /   \               /     \
    /  A  \             / A   b \
   (       )           (         )
    \  B  /             \ C   D /
     \   /               \     /
      '-'                 '---'


4.1.1 Comparison with Related Definitions

   In SIP, a call is "an informal term that refers to some
   communication between peers, generally set up for the purposes of a
   multimedia conversation."  Obviously we cannot discuss normative
   behavior based on such an intentionally vague definition.  The
   concept of a conversation space is needed because the SIP definition
   of call is not sufficiently precise for the purpose of describing
   the user experience of multiparty features.

   Do any other definitions convey the correct meaning?  SIP, and [SDP]
   both define a conference as "a multimedia session identified by a
   common session description."  A session is defined as "a set of
   multimedia senders and receivers and the data streams flowing from
   senders to receivers."  Both of these definitions are heavily
   oriented toward multicast sessions with little differenciation among
   participants.  As such, neither is particularly useful for our
   purposes.  In fact, the definition of "call" in some call models is
   more similar to our definition of a conversation space.

   Some examples of the relationship between conversation spaces, SIP
   call legs, and SIP sessions are listed below.  In each example, a
   human user will perceive that there is a single call.

       A simple two-party call is a single conversation space, a single
       session, and a single call-leg.

       A locally mixed three-way call is two sessions and two call-
       legs.  It is also a single conversation space.

       A simple dial-in audio conference is a single conversation
       space, but is represented by as many call-legs and sessions as
       there are human participants.

       A multicast conference is a single conversation space, a single
       session, and as many call-legs as participants.

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 10]


                       SIP Multiparty Framework


4.2 Signaling Models

   Obviously to make changes to a conversation space, you must be able
   to use SIP signaling to cause these changes.  Specifically there
   must be a way to manipulate SIP dialogs (call legs) to move
   participants into and out of conversation spaces.  Although this is
   not as obvious, there also must be a way to manipulate SIP dialogs
   to include non-participant user agents which are otherwise involved
   in a conversation space (ex: B2BUAs, 3pcc controllers, mixers,
   transcoders, translators, or relays).

   Implementations may setup the media relationships described in the
   conversation space model using the approach described in [3pcc]. The
   3pcc approach relies on only the following 3 primitive operations:

       Create a new call-leg  (INVITE)
       Modify a call-leg      (reINVITE)
       Destroy a call-leg     (BYE)

   The main advantage of the 3pcc approach is that it only requires
   very basic SIP support from end systems to support call control
   features.  As such, third-party call control is a natural way to
   handle protocol conversion and mid-call features.  It also has the
   advantage and disadvantage that new features can/must be implemented
   in one place only (the controller), and neither requires enhanced
   client functionality, nor takes advantage of it.

   In addition, a peer-to-peer approach is discussed at length in this
   draft.  The primary drawback of the peer-to-peer model is additional
   end system complexity.  The benefits of the peer-to-peer model
   include:
   - state remains at the edges
   - call signaling need only go through participants involved
     (there are no additional points of failure)
   - peers can take advantage of end-to-end message integrity or
     encryption
   - setup time is shorter (fewer messages and round trips
     are required)

   The peer-to-peer approach relies on additional "primitive"
   operations, some of which are identified here.

       Replace an existing dialog
       Join a new dialog with an existing dialog [Join]
       Fork a new dialog with an existing dialog
       Locally do media forking (multi-unicast)
       Ask another UA to send a request on your behalf

   Many of the features, primitives, and actions described in this
   document also require some type of media mixing, combining, or
   selection as described in the next section.



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 11]


                       SIP Multiparty Framework

4.3 Mixing Models

   SIP permits a variety of mixing models, which are discussed here
   briefly.  This topic is discussed more thoroughly in [conf-models].
   For brevity, only the two most popular conferencing models are
   significantly discussed in this document (local and centralized
   mixing).  Applications of the conversation spaces model to multicast
   and multi-unicast (full unicast mesh) conferences are left as an
   exercise for the reader.  Note that a distributed full mesh
   conference can be used for basic conferences, but does not easily
   allow for more complex conferencing actions like splitting, joining,
   and forking.

   Call control features should be designed to allow a mixer (local or
   centralized) to decide when to reduce a conference back to a 2-party
   call, or drop all the participants (for example if only two
   automatons are communicating).  The actual heuristics used to
   release calls are beyond the scope of this document, but may depend
   on properties in the conversation space, such as the number of
   active, passive, or hidden participants; and the send-only, receive-
   only, or send-and-receive orientation of various participants.

4.3.1  (Single) End System Mixing

   The first model we call "end system mixing". In this model, user A
   calls user B, and they have a conversation. At some point later, A
   decides to conference in user C. To do this, A calls C, using a
   completely separate SIP call. This call uses a different Call-ID,
   different tags, etc. There is no call set up directly between B and
   C.  No SIP extension or external signaling is needed.  A merely
   decides to locally join two call-legs.

   [diagram]

   A receives media streams from both B and C, and mixes them. A sends
   a stream containing A's and C's streams to B, and a stream
   containing A's and B's streams to C. Basically, user A handles both
   signaling and media mixing. B and C are unaware of the multi-party
   call, from a SIP perspective at least. From an RTP perspective, A is
   a mixer, and so the RTCP reports from A will contain SDES
   information that indicates the existence of an additional party in
   the media stream.

4.3.2 Centralized Mixing

   In a centralized mixing model, all participants have a pairwise SIP
   and media relationship with the mixer.  Three applications of
   centralized mixing are also discussed below.

   [diagram]

4.3.2.1 Dial-In Conference Servers



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 12]


                       SIP Multiparty Framework

   Dial-In conference servers closely mirror dial-in conference bridges
   in the traditional PSTN. A dial-in conference server acts as a
   normal SIP UA. Users call it, and the server maintains point to
   point SIP relationships with each user that calls in. The server
   takes the media from the users who dial into the same conference,
   mixes them, and sends out the appropriate mixed stream to each
   participant separately. The model is depicted in Figure 3. Note that
   each UA (A,B,C,D) has a point to point SIP and RTP relationship with
   the conference server. Each call has a different Call-ID. Each user
   sends their own media to the server. The media delivered to user A
   by the server is the media mixed from users B, C and D. The media
   delivered to user B by the server is the media mixed from users A, C
   and D. The media delivered to user C by the server is the media
   mixed from users A, B and D. The media delivered to user D is the
   media mixed from users A, B and C (this is also known as a mix-minus
   configuration).

   As in other applications of centralized mixing, the conference is
   identified by the request URI of the calls from each participant.
   This provides numerous advantages from a services and routing point
   of view [9]. For example, one conference on the server might be
   known as sip:conference34@servers.com. All users who call
   sip:conference34@servers.com are mixed together. Dial-In conference
   servers are usually associated with pre-arranged conferences.
   However, the same model applies to ad-hoc conferences. An ad-hoc
   conference server creates the conference state when the first user
   joins, and destroys it when the last one leaves. The SIP and RTP
   interfaces are identical to the pre-arranged case.

4.3.2.2 Ad-hoc Centralized Conferences

   In an ad-hoc centralized conference, two users A and B start with a
   normal SIP call. At some point later, they decide to add a third
   party. Instead of using end system mixing, they would prefer to use
   a conference server. Initially, A calls B. At some point, B decides
   to add user C to the call, and begins the transition to a conference
   server. The first step in this process is the discovery of a
   conference server that supports ad-hoc conferences. This can be done
   through static configuration, or through any of a number of standard
   service discovery protocols, such as the Service Location Protocol
   [SLP]. Once the server is discovered, a conference ID is chosen.
   This ID must be globally unique. The conference ID is then prepended
   to the server, and a SIP URL for the ad-hoc conference is formed.
   For example, if the server "a.servers.com" is used, and the unique
   ID is "a7hytaskp09878a", the SIP URL for this conference is
   sip:a7hytaskp09878a@a.servers.com. The first participant to send an
   INVITE to this URL creates the initial conference state in the
   server.  SIP dialogs are manipulated (using any combination of 3pcc
   or peer-to-peer signaling) so that each participant is sending media
   to the conference server. It is also possible to transition from a
   end system mixed conference (even one with a complex connection
   topology), to a centralized conference server.



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 13]


                       SIP Multiparty Framework

4.3.2.3 Dial-Out Conferences

   Dial-out conferences are a simple variation on dial-in conferences.
   Instead of the users joining the conference by sending an INVITE to
   the server, the server chooses the users who are to be members of
   the conference, and then sends them the INVITE. Typically dial out
   conferences are pre-arranged, with specific start times and an
   initial group membership list. However, there are other means for
   the dial-out server to determine the list of participants, including
   user presence [13]. The model in no way limits the means by which
   the server determines the set of users. Once the users accept or
   reject the call from the dial out server, the behavior of this
   system is identical to the dial-in server case of Section 4. Thus, a
   dial-out conference server will generally need to support dial-in
   access for the same conference, if it wishes to allow joining after
   the conference begins. Note that, from the participants perspective,
   they will learn the conference identity (the URL) from the From
   field in the INVITE messages received from the server.

4.3.3 Multicast and Multi-unicast conferences

   In these models, all endpoints send media to all other endpoints.
   Consequently every endpoint mixes their own media from all the other
   sources, and sends their own media to every other participant.

   [diagrams]

4.3.3.1 Large-Scale Multicast Conferences

   Large-scale multicast conferences were the original motivation for
   both the Session Description Protocol [SDP] and SIP. In a large-
   scale multicast conference, one or more multicast addresses are
   allocated to the conference (more than one may be needed if layered
   encodings are in use). Each participant joins that multicast groups,
   and sends their media to those groups. Signaling is not sent to the
   multicast groups. The sole purpose of the signaling is to inform
   participants of which multicast groups to join. Large-scale
   multicast conferences are usually pre-arranged, with specific start
   and stop times (which is why this information exists in SDP).
   Protocols such as the Session Announcement Protocol [SAP] are used
   to announce these conferences. However, multicast conferences do not
   need to be pre-arranged, so long as a mechanism exists to
   dynamically obtain a multicast address. So, if there are N
   participants, there will be point-to-point SIP relationships with
   pairs of participants. Each participant sends a single media stream
   to the group, and receives up to N-1 streams at any time. Note that
   the number of streams that a user will receive depends on who is
   actually sending at any given time. If the stream is audio, and
   silence suppression is utilized, the number of streams a user will
   receive at any given time is equal to the number of users talking at
   any given time. Even for very large conferences, this is usually
   just a small number of users.



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 14]


                       SIP Multiparty Framework

4.3.3.2 Centralized Signaling, Distributed Media

   In this conferencing model, there is a centralized controller, as in
   the dial-in and dial-out cases. However, the centralized server
   handles signaling only. The media is still sent directly between
   participants, using either multicast or multi-unicast. Multi-unicast
   is when a user sends multiple packets (one for each recipient,
   addressed to that recipient). This is referred to as a
   "Decentralized Multipoint Conference" in [H.323].

4.3.3.3 Full Distributed Unicast Conferencing

   In this conferencing model, each participant has both a pairwise
   media relationship and a pairwise SIP relationship with every other
   participant (a full mesh).  This model requires a mechanism to
   maintain a consistent view of distributed state across the group.
   This is a classic hard problem in computer science.  Also, this
   model does not scale well for large numbers of participants.
   bascause for <n> participants the number of media and SIP
   relationships is approximately n-squared.  As a result, this model
   is not generally available in commercial implementations; to the
   contrary it is primarily the topic of research or experimental
   implementations.  Note that this model assumes peer-to-peer
   signaling.

4.4 Conveying Information and Events

   Participants should have access to information about the other
   participants in a conversation space, so that this information can
   be rendered to a human user or processed by an automaton.  Although
   some of this information may be available from the Request-URI or
   To, From, Contact, or other SIP headers, another mechanism of
   reporting this information is necessary.  Note that the data
   reported by RTCP is insufficient for these purposes, as deletions
   and additions are not detectable in real-time, and SIP may setup
   session which do not involve RTP media.

   Many applications are driven by knowledge about the progress of
   calls and conferences.  In general these types of events allow for
   the construction of distributed applications, where the application
   requires information on dialog and conference state, but is not
   necessarily co-resident with an endpoint user agent or conference
   server.  For example, a mixer involved in a conversation space may
   wish to provide URLs for conference status, and/or conference/floor
   control.

   The SIP [Events] architecture defines general mechanisms for
   subscription to and notification of events within SIP networks.  It
   introduces the notion of a package which is a specific
   "instantiation" of the events mechanism for a well-defined set of
   events.

   New event packages should be able to


Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 15]


                       SIP Multiparty Framework

   provide the status of a user's call-legs (dialogs), provide the
   status of conferences and its participants, provide user presence
   information, and provide the status of user's messages.  While this
   is not an exhaustive list, these are sufficient to enable the sample
   features described in this document.

   A conference event package allows users to subscribe to information
   about an entire conference or conversation space.  This conference
   state could be provided by a conference server or mixing component
   (described in Section 4.5) if centralized mixing is used, or
   gathered from relevant peers and merged into a cohesive set of
   state.  Notifications would convey information about the
   pariticipants such as: the SIP URL identifying each user, their
   status in the space (active, declined, departed), URLs to invoke
   other features (such as sidebar conversations), links to other
   relevant information (such as floor control policies), and if floor
   control policies are in place, the user's floor control status.  A
   "call-leg" event package would provide information about all the
   dialogs the target user is maintaining, what conversations the user
   in participating in, and how these are correlated.  A concrete
   proposal for both conference events and call-leg events is described
   in [call-pkg].

   Note that user presence has a close relationship with these two
   proposed event packages. It is fundamental to the presence model
   that the information used to obtain user presence is constructed
   from any number of different input sources. Examples of such sources
   include SIP REGISTER requests and uploads of presence documents.
   These two packages can be considered another mechanism that allows a
   presence agent to determine the presence state of the user.
   Specifically, a user presence server can act as a subscriber for the
   call-leg and conference packages to obtain additional information
   that can be used to construct a presence document.

   The multi-party architecture should also provide a mechanism to get
   information about the status /handling of a dialog (for example,
   information about the history of other contacts attempted prior to
   the current contact).  Finally, the architecture should provide
   ample opportunities to present informational URIs which relate to
   calls, conversations, or dialogs in some way.  For example, consider
   the SIP Call-Info header, or Contact headers returned in a 300-class
   response.  Frequently additional information about a call or dialog
   can be fetched via non-SIP URIs.  For example, consider a web page
   for package tracking when calling a delivery company, or a web page
   with related documentation when joining a dial-in conference.  The
   use of URIs in the multiparty framework is discussed in more detail
   in Section 4.6.


4.5 Componentization and Decomposition

   This framework proposes a decomposed component architecture with a
   very loose coupling of services and components.  This means that a
   service (such as a conferencing server or an auto-attendant) need

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 16]


                       SIP Multiparty Framework

   not be implemented as an actual server.  Rather, these services can
   be built by combining a few basic components in straightforward or
   arbitrarily complex ways.

   Since the components are easily deployed on separate boxes, by
   separate vendors, or even with separate providers, we achieve a
   separation of function that allows each piece to be developed in
   complete isolation.  We can also reuse existing components for new
   applications.  This allows rapid service creation, and the ability
   for services to be distributed across organizational domains
   anywhere in the Internet.

   For many of these components it is also desirable to discover their
   capabilities, for example querying the ability of a mixer to host a
   10 dialog conference, or to reserve resources for a specific time.
   These actions could be provided in the form of URLs, provided there
   is an a priori means of understanding their semantics.  For example
   if there is a published dictionary of operations, a way to query the
   service for the available operations and the associated URLs, the
   URL can be the interface for providing these service operations.
   This concept is described in more detail in the context of dialog
   operations in section 4.6

4.5.1 Media Intermediaries

   Media Intermediaries are not participants in any conversation space,
   although an entity which is also a media translator may also have a
   colocated participant component (for example a mixer which also
   announces the arrival of a new participant; the announcement portion
   is a participant, but the mixer itself is not).  Media
   intermediaries should be as transparent as possible to the end
   users--offering a useful, fundamental service; without getting in
   the way of new features implemented by participants.  Some common
   media intermediaries are desribed below.

4.5.1.1 Mixer

   A mixer is a component that combines media from all call-legs in the
   same conversation in a media specific way.  For example, the default
   combining for an audio conference would be an N-1 configuration.  In
   other words, each user receives a mixed media stream that represents
   the combined audio of all the users except himself or herself.

   For reference, the RTP definition of a mixer is included below.
   Note that SIP multiparty applications may deal with media which is
   not carried by RTP (for example Instant Messages).  A mixer, as
   defined above, can still combine these messages in a media specific
   way and act as a SIP mixing component.

        "Mixer: An intermediate system that recieves RTP packets from
        one or more sources, ... combines the packets in some manner
        and then forwards a new RTP packet.  Since the timing across
        multiple input sources will not generally be syncronized, the
        mixer will make timing adjustments among the streams and

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 17]


                       SIP Multiparty Framework

        generate its own timing for the combined stream.  Thus all data
        packets originating from a mixer will be identified as having
        the mixer as their syncronization source."

   Conventions for specifying a mixing or conferencing service in a SIP
   URI are proposed in [ms-uri].

4.5.1.2 Media Translator

   RTP also defines an entity called a translator.  Like a mixer, this
   concept is useful outside of the context of RTP and can be applied
   to most other media types.

        "Translator: An intermediate system that forwards RTP packets
        with their syncronization source identifier intact.  Examples
        of translators include devices that convert encodings without
        mixing, replicators from multicast to unicast, and application-
        level firewalls."

4.5.1.3 Transcoder

   A transcoder translates media from one encoding to another (for
   example, GSM voice to G.711, or MPEG2 to H.261).  A transcoder for
   RTP media is a type of RTP translator.

4.5.1.4 Media Relay

   A media relay terminates media and simply forwards it to a new
   destination without changing the content in any way.  Sometimes
   media relays are used to provide source IP address anonymity, to
   facilitate middlebox traversal, or to provide a trusted entity where
   media can be forcefully disconnected.  A media relay for RTP is also
   a type of RTP Translator.

4.5.2 Queue Server

   A queue server is a location where calls can be entered into one of
   several FIFO (first-in, first-out) queues.  A queue server would
   subscribe to the presence of groups or individuals who are
   interested in its queues.  When detecting that a user is available
   to service a queue, the server redirects or transfers the last call
   in the relevant queue to the available user.  On a queue-by-queue
   basis, authorized users could also subscribe to the call state
   (dialog information) of calls within a queue.  Authorized users
   could use this information to effectively pluck (take) a call out of
   the queue (for example by sending an INVITE with a Replaces header
   to one of the user agents in the queue).


4.5.3 Parking Place

   A parking place is a location where calls can be terminated
   temporarily and then retrieved later.  While a call is "parked", it
   can receive media "on-hold" such as music, announcements, or

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 18]


                       SIP Multiparty Framework

   advertisements.  Such a service could be further decomposed such
   that announcements or music are handled by a separate component.


4.5.4 Announcements and Voice Dialogs

   An announcement server is a server which can play digitized media
   (frequently audio), such as music or recorded speech.  These servers
   are typically accessible via SIP, HTTP, or RTSP.  An analogous
   service is a recording service which stores digitized media.  A
   convention for specifying announcements in SIP URIs is described in
   [ms-uri].  Likewise the same server could easily provide a service
   which records digitized media.

   A "voice dialog" is a model of spoken interactive behavior between a
   human and an automaton which can include synthesized speech,
   digitized audio, recognition of spoken and DTMF key input, recording
   of spoken input, and interaction with call control. Dialogs
   frequently consist of forms or menus. Forms present information and
   gather input; menus offer choices of what to do next.

   Spoken dialogs are a basic building block of applications which use
   voice.  Consider for example that a voice mail system, the
   conference-id and passcode collection system for a conferencing
   system, and complicated voice portal applications all require a
   voice dialog component.


4.5.4.1. Text-to-Speech and Automatic Speech Recognition

   Text-to-Speech (TTS) is a service which converts text into digitized
   audio.  TTS is frequently integrated into other applications, but
   when separated as a component, it provides greater opportunity for
   broad reuse.  Various interfaces to access standalone TTS services
   via HTTP, RTSP (in [MRCP]), and SIP ([app-components], [ms-uri] and
   [MRCP-SIP]) have been proposed.

   Automatic Speech Recognition (ASR) is a service which attempts to
   decipher digitized speech based on a proposed grammar.  Like TTS,
   ASR services can be embedded, or exposed so that many applications
   can take advantage of such services.  Various IP interfaces to ASR,
   such as MRCP, have been proposed.


4.5.4.2. VoiceXML

   [VoiceXML] is a W3C recommendation that was designed to give authors
   control over the spoken dialog between users and applications. The
   application and user take turns speaking: the application prompts
   the user, and the user in turn responds.  Its major goal is to bring
   the advantages of web-based development and content delivery to
   interactive voice response applications.  We believe that VoiceXML
   represents the ideal partner for SIP in the development of
   distributed IVR servers. VoiceXML is an XML based scripting language

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 19]


                       SIP Multiparty Framework

   for describing IVR services at an abstract level. VoiceXML supports
   DTMF recognition, speech recognition, text-to-speech, and playing
   out of recorded media files. The results of the data collected from
   the user are passed to a controlling entity through an HTTP POST
   operation. The controller can then return another script, or
   terminate the interaction with the IVR server.

   A VoiceXML server also need not be implemented as a monolithic
   server.  Below is a diagram of a VoiceXML browser which is split
   into media and non-media handling parts.  The VoiceXML interpreter
   handles SIP dialog state and state within a VoiceXML document, and
   sends requests to the media component over another protocol (for
   example RTSP).

                       +-------------+
                       |             |
                       | VoiceXML    |
                       | Interpreter |
                       | (signaling) |
                       +-------------+
                         ^          ^
                         |          |
                     SIP |          | [RTSP]
                         |          |
                         |          |
                         v          v
            +-------------+        +-------------+
            |             |        |             |
            |  SIP UA     |   RTP  | RTSP Server |
            |             |<------>|   (media)   |
            |             |        |             |
            +-------------+        +-------------+


               Figure : Decomposed VoiceXML Server


   From a naming perspective, a critical issue when using VoiceXML is
   how a request URI is associated with a script to invoke when the
   call is answered. We see three primary mechanisms: 1) There is a
   one-to-one binding of the address in the request URI to a script to
   execute. These bindings are published by the provider of the IVR
   service. 2) The initial script to execute is actually carried as
   content in the body of the SIP INVITE request. The request URI
   indicates that the desired service is execution of content in the
   request (i.e., sip:executebody@servers.com). 3) The initial script
   to execute is fetched by the VoiceXML server; the URL to fetch it
   from is passed in the SIP INVITE message that initiates the IVR
   session. This can be accomplished either with the application/uri
   MIME type as a body, or using the *-Info headers defined in SIP
   which provide references to content to fetch. We believe that the
   third approach is probably the best one. SIP is not the ideal
   transfer mechanism. Passing a URI allows a far better transfer tool,
   for example HTTP, to be used to actually fetch the script back from

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 20]


                       SIP Multiparty Framework

   the controller. HTTP is then also used to pass back form data from
   the IVR to the controller. The results of the HTTP POST can also
   contain additional VoiceXML scripts to execute.  More details about
   the integration of SIP with VoiceXML are provided in [sip-vxml]

4.6 Use of URIs

   All naming in SIP uses URIs.  URIs in SIP are used in a plethora of
   contexts: the Request-URI; Contact, To, From, and *-Info headers;
   application/uri bodies; and embedded in email, web pages, instant
   messages, and ENUM records.  The request-URI identifies the user or
   service that the call is destined for.

   SIP URIs embedded in informational SIP headers, SIP bodies, and non-
   SIP content can also specify methods, special parameters, headers,
   and even bodies.  For example:

   sip:bob@babylon.biloxi.com;method=BYE?Call-ID=13413098
     &To=<sip:bob@biloxi.com>;tag=879738
     &From=<sip:alice@atlanta.com>;tag=023214

   sip:bob@babylon.biloxi.com;method=REFER?
     Refer-To=<http://www.atlanta.com/~alice>

   Throughout this draft we discuss call control primitive operations.
   One of the biggest problems is defining how these operations may be
   invoked.  There are a number of ways to do this.  One way is to
   define the primitives in the protocol itself such that SIP methods
   (for example REFER) or SIP headers (for example Replaces) indicate a
   specific call control action.  Another way to invoke call control
   primitives is to define a specific Request-URI naming convention.
   Either these conventions must be shared between the client (the
   invoker) and the server, or published by or on behlf of the server.
   The former involves defining URL construction techniques (e.g. URL
   parameters and/or token conventions) as proposed in [ms-uri].  The
   latter technique usually involves discovering the URI via a SIP
   event package, a web page, a business card, or an Instant Message.
   Yet another means to acquire the URLs is to define a dictionary of
   primitives with well-defined semantics and provide a means to query
   the named primitives and corresponding URLs that may be invoked on
   the service or dialogs.

4.6.1 Naming Users in SIP

   An address-of-record, or public SIP address, is a SIP (or SIPS) URI
   that points to a domain with a location server that can map the URI
   to set of Contact URIs where the user might be available.  Typically
   the Contact URIs are populated via registration.

        Address of Record        Contacts

        sip:bob@biloxi.com   ->  sip:bob@babylon.biloxi.com:5060
                                 sip:bbrown@mailbox.provider.net
                                 sip:+1.408.555.6789@mobile.net

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 21]


                       SIP Multiparty Framework


   [Caller-prefs] defines a set of additional parameters to the Contact
   header that define the characteristics of the user agent at the
   specified URI.  For example, there is a mobility parameter which
   indicates whether the UA is fixed or mobile.  When a user agent
   registers, it places these parameters in the Contact headers to
   characterize the URIs it is registering.  This allows a proxy for
   that domain to have information about the contact addresses for that
   user.

   When a caller sends a request, it can optionally include the Accept-
   Contact and Reject-Contact headers which request certain handling by
   the proxy in the target domain.  These headers contain preferences
   that describe the set of desired URIs to which the caller would like
   their request routed.  The proxy in the target domain matches these
   preferences with the Contact characteristics originally registered
   by the target user.  The target user can also choose to run
   arbitrarily complex "Find-me" feature logic on a proxy in the target
   domain.

   There is a strong asymmetry in how preferences for callers and
   callees can be presented to the network. While a caller takes an
   active role by initiating the request, the callee takes a passive
   role in waiting for requests. This motivates the use of callee-
   supplied scripts and caller preferences included in the call
   request.  This asymmetry is also reflected in the appropriate
   relationship between caller and callee preferences. A server for a
   callee should respect the wishes of the caller to avoid certain
   locations, while the preferences among locations has to be the
   callee's choice, as it determines where, for example, the phone
   rings and whether the callee incurs mobile telephone charges for
   incoming calls.

   SIP User Agent implementations are encouraged to make intelligent
   decisions based on the type of participants (active/passive, hidden,
   human/robot) in a conversation space.  This information is conveyed
   in a SIP URI parameter and communicated using an appropriate SIP
   header or event body.  For example, a music on hold service may take
   the sensible approach that if there are two or more unhidden
   participants, it should not provide hold music; or that it will not
   send hold music to robots.

   Multiple participants in the same conversation space may represent
   the same human user.  For example, the user may use one participant
   for video, chat, and whiteboard media on a PC and another for audio
   media on a SIP phone.  In this case, the address-of-record is the
   same for both user agents, but the Contacts are different.  In
   addition, human users may add robot participants which act on their
   behalf (for example a call recording service, or a calendar
   reminder).  Call Control features in SIP should continue to function
   as expected in such an environment.




Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 22]


                       SIP Multiparty Framework

4.6.2 Naming Services with SIP URIs.

   A critical piece of defining a session level service that can be
   accessed by SIP is defining the naming of the resources within that
   service.  This point cannot be overstated.

   In the context of SIP control of application components, we take
   advantage of the fact that the standard SIP URI has a user part.
   Most services may be thought of as user automatons that participate
   in SIP sessions. It naturally follows that the user address, or the
   left-hand-side of the URI, should be utilized as a service
   indicator.

   For example, media servers commonly offer multiple services at a
   single host address.  Use of the user part as a service indicator
   enables service consumers to direct their requests without
   ambiguity.  It has the added benefit of enabling media services to
   register their availability with SIP Registrars just as any "real"
   SIP user would.  This maintains consistency and provides enhanced
   flexibility in the deployment of media services in the network.

   There has been much discussion about the potential for confusion if
   media services URIs are not readily distinguishable from other types
   of SIP UA's.  The use of a service namespace provides a mechanism to
   unambiguously identify standard interfaces while not constraining
   the development of private or experimental services.

   In SIP, the request-URI identifies the user or service that the call
   is destined for.  The great advantage of using URIs (specifically,
   the SIP request URI) as a service identifier comes because of the
   combination of two facts. First, unlike in the PSTN, where the
   namespace (dialable telephone numbers) are limited, URIs come from
   an infinite space. They are plentiful, and they are free. Secondly,
   the primary function of SIP is call routing through manipulations of
   the request URI. In the traditional SIP application, this URI
   represents people. However, the URI can also represent services, as
   we propose here. This means we can apply the routing services SIP
   provides to routing of calls to services. The result - the problem
   of service invocation and service location becomes a routing
   problem, for which SIP provides a scalable and flexible solution.
   Since there is such a vast namespace of services, we can explicitly
   name each service in a finely granular way. This allows the
   distribution of services across the network.

   Consider a conferencing service, where we have separated the names
   of ad-hoc conferences from scheduled conferences, we can program
   proxies to route calls for ad-hoc conferences to one set of servers,
   and calls for scheduled ones to another, possibly even in a
   different provider. In fact, since each conference itself is given a
   URI, we can distribute conferences across servers, and easily
   guarantee that calls for the same conference always get routed to
   the same server. This is in stark contrast to conferences in the
   telephone network, where the equivalent of the URI - the phone
   number - is scarce. An entire conferencing provider generally has

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 23]


                       SIP Multiparty Framework

   one or two numbers. Conference IDs must be obtained through IVR
   interactions with the caller, or through a human attendant. This
   makes it difficult to distribute conferences across servers all over
   the network, since the PSTN routing only knows about the dialed
   number.

   In the case of a dialog server, the voice dialog itself is the
   target for the call. As such, the request URI should contain the
   identifier for this spoken dialog. This is consistent with the
   Request-URI service invocation model of RFC 3087. This URL can be in
   one of two formats. In the first, the VoiceXML script is identified
   directly by an HTTP URL. In the second, the script is not specified.
   Rather, the dialog server uses its configuration to map the incoming
   request to a specific script.

   Since the request URI could indicate a request for a variety of
   different services, of which a dialog server is only one type, this
   example request URI first begins with a service identifier, that
   indicates the basic service required. For VoiceXML scripts, this
   identification information is a URL-encoded version of the URL which
   references the script to execute, or if not present, the dialog
   server uses server-specific configuration to determine which script
   to execute.

      Examples of URLs that invoke VoiceXML dialogs are:
      (line folding for clarity only)

      sip:dialog.vxml.http%3a//dialogs.server.com/script32.vxml
       @vxmlservers.com

      sip:dialog.vxml@vxmlservers.com

   The first of these indicates that the dialog server (located at
   vxmlservers.com) should invoke a VoiceXML script fetched from
   http://dialogs.server.com/script32.vxml. Since the user part of the
   SIP URL cannot contain the : character, this must be escaped to %3a.

   These types of conventions are not limited to application component
   servers.  An ordinary SIP User Agent can have a special URIs as
   well, for example, one which is automatically answered by a
   speakerphone.  Since URIs are so plentiful, using a separate URI for
   this service does not exhaust a valuable resource.  The requested
   service is clear to the user agent receiving the request.  This URI
   can also be included as part of another feature (for example, the
   Intercom feature described in Section 6.1.6).  This feature can be
   specified with a SIP user parameter, since are part of the userpart
   of a SIP URI.

   Likewise a Request URI can fully describe an announcement service
   through the use of the user part of the address and additional URI
   parameters.  In our example, the user portion of the address,
   "annc",  specifies the announcement service on the media server.
   The two URI parameters "play=" and "early=" specify the audio
   resource to play and whether early media is desired.

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 24]


                       SIP Multiparty Framework


       sip:annc@ms2.carrier.net;
        play=http://audio.carrier.net/allcircuitsbusy.au;early=yes

       sip:annc@ms2.carrier.net;
        play=file://fileserver.carrier.net/geminii/yourHoroscope.wav


   In practical applications, it is important that an invoker does not
   necessarily apply semantic rules to various URIs it did not create.
   Instead, it should allow any arbitrary string to be provisioned, and
   map the string to the desired behavior. The administrator of a
   service may choose to provision specific conventions or mnemonic
   strings, but the application should not require it. In any large
   installation, the system owner is likely to have pre-existing rules
   for mnemonic URIs, and any attempt by an application to define its
   own rules may create a conflict.  Implementations should allow an
   arbitrary mix of URLs from these schemes, or any other scheme that
   renders valid SIP URIs to be provisioned, rather than enforce only
   one particular scheme.

   For example, a voicemail application can be built using very
   different sets of URI conventions, as illustrated below:

        URI Identity       Example Scheme 1
                                Example Scheme 2
                                     Example Scheme 3

        Deposit with       sip:sub-rjs-deposit@vm.wcom.com
        standard greeting       sip:677283@vm.wcom.com
                                     sip:rjs@vm.wcom.com;mode=deposit


        Deposit with on    sip:sub-rjs-deposit-busy.vm.wcom.com
        phone greeting          sip:677372@vm.wcom.com
                                     sip:rjs@vm.wcom.com;mode=3991243

        Deposit with       sip:sub-rjs-deposit-sg@vm.wcom.com
        special greeting        sip:677384@vm.wcom.com
                                     sip:rjs@vm.wcom.com;mode=sg

        Retrieve - SIP     sip:sub-rjs-retrieve@vm.wcom.com
        authentication          sip:677405@vm.wcom.com
                                     sip:rjs@vm.wcom.com;mode=retrieve

        Retrieve - prompt  sip:sub-rjs-retrieve-inpin.vm.wcom.com
        for PIN in-band         sip:677415@vm.wcom.com
                                     sip:rjs@vm.wcom.com;mode=inpin

   As we have shown, SIP URIs represent an ideal, flexbile mechanism
   for describing and naming service resources, be they queues,
   conferences, voice dialogs, announcements, voicemail treatments, or
   phone features.


Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 25]


                       SIP Multiparty Framework


4.7 Invoker Independence

   Only the invoker of features in SIP know exactly which feature they
   are invoking.  One of the primary benefits of this approach is that
   combinations of features should work in SIP call control.  For
   example, let us examine the combination of a "transfer" of a call
   which is "conferenced".

   Alice calls Bob.  Alice silently "conferences in" her robotic
   assistant Albert as a hidden party.  Bob transfers Alice to Carol.
   If Bob asks Alice to Replace her leg with a new one to Carol then
   both Alice and Albert should be communicating with Carol
   (transparently).

   Using the peer-to-peer model, this combination of features works
   fine if A is doing local mixing (Alice replaces Bob's call-leg with
   Carol's), or if A is using a central mixer (the mixer replaces Bob's
   call leg with Carol's).  A clever implementation using the 3pcc
   model can generate similar results.

   New extensions to the SIP Call Control Framework should attempt to
   preserve this property.

4.8 Billing issues

   Billing in the PSTN is typically based on who initiated a call.  At
   the moment billing in a SIP network is neither consistent with
   itself, nor with the PSTN.  (A billing model for SIP should allow
   for both PSTN-style billing, and non-PSTN billing.)  The example
   below demonstrates one such inconsistency.

   Alice places a call to Bob.  Alice then blind transfers Bob to Carol
   through a PSTN gateway.  In current usage of REFER and BYE/Also, Bob
   may be billed for a call he did not initiate (his UA originated the
   outgoing call leg however).  This is not necessarily a terrible
   thing, but it demonstrates a security concern (Bob must have
   appropriate local policy to prevent fraud).  Also, Alice may wish to
   pay for Bob's session with Carol.  There should be a way to signal
   this in SIP.

   Likewise a Replacement call may maintain the same billing
   relationship as a Replaced call, so if Alice first calls Carol, then
   asks Bob to Replace this call, Alice may continue to receive a bill.

   Further work in SIP billing should define a way to set or discover
   the direction of billing.


5 Catalog of call control actions and sample features

   Call control actions can be categorized by the dialogs upon which
   they operate.  The actions may involve a single or multiple dialogs.
   These dialogs can be early or established.  Multiple dialogs may be

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 26]


                       SIP Multiparty Framework

   related in a conversation space to form a conference or other
   interesting media topologies.

   It should be noted that it is desirable to provide a means by which
   a party can discover the actions which may be performed on a dialog.
   The interested party may be independent or related to the dialogs.
   One means of accomplishing this is through the ability to define and
   obtain URLs for these actions as described in section 4.6.

   Below are listed several call control "actions" which establish or
   modify dialogs and relate the participants in a conversation space.
   The names of the actions listed are for descriptive purposes only
   (they are not normative).  This list of actions is not meant to be
   exhaustive.

   In the examples, all actions are initiated by the user "Alice"
   represented by UA "A".

5.1 Early Dialog Actions

   The following are a set of actions that may be performed on a single
   early dialog.  These actions can be thought of as a set of remote
   control operations.  For example an automaton might perform the
   operation on behalf of a user.  Alternatively a user might use the
   remote control in the form of an application to perform the action
   on the early dialog of a UA which may be out of reach. All of these
   actions correspond to telling the UA how to respond to a request to
   establish an early dialog. These actions provide useful
   functionality for PDA, PC and server based applications which desire
   the ability to control a UA.

5.1.1 Remote Answer

   A dialog is in some early dialog state such as 180 Ringing.  It may
   be desirable to tell the UA to answer the dialog.  That is tell it
   to send a 200 Ok response to establish the dialog.

5.1.2 Remote Forward or Put

   It may be desirable to tell the UA to respond with a 3xx class
   response to forward an early dialog to another UA.

5.1.3 Remote Busy or Error Out

   It may be desirable to instruct the UA to send an error response
   such as 486 Busy Here.

5.2 Single Dialog Actions

   There is another useful set of actions which operate on a single
   established dialog.  These operations are useful in building
   productivity applications for aiding users to control their phone.
   For example a CRM application which sets up calls for a user


Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 27]


                       SIP Multiparty Framework

   eliminating the need for the user to actually enter an address.
   These operations can also be thought of a remote control actions.

5.2.1 Remote Dial

   This action instructs the UA to initiate a dialog.  This action can
   be performed using the REFER method.

5.2.2 Remote On and Off Hold

   This action instructs the UA to put an established dialog on hold.
   Though this operation can be conceptually be performed with the
   REFER method, there is no semantics defined as to what the referred
   party should do with the SDP. There is no way to distinguish between
   the desire to go on or off hold.

5.2.3 Remote Hangup

   This action instructs the UA to terminate an early or established
   dialog. A REFER request with the following Refer-To URI performs
   this action.  Note: this URL is not properly escaped.

   sip:bob@babylon.biloxi.com;method=BYE?Call-ID=13413098
     &To=<sip:bob@biloxi.com>;tag=879738
     &From=<sip:alice@atlanta.com>;tag=023214


5.3 Multi-dialog actions

   These actions apply to a set of related dialogs.

5.3.1 Transfer

   The conversation space changes as follows:

        before            after
        { A , B }  -->   { C , B }

   A replaces itself with C.

   To make this happen using the peer-to-peer approach, "A" would send
   two SIP requests.  A shorthand for those requests is shown below:
        REFER B  Refer-To:C
        BYE B

   To make this happen instead using the 3pcc approach, the controller
   sends requests represented by the shorthand below:
        INVITE C (w/SDP of B)
        reINVITE B (w/SDP of C)
        BYE A

   Features enabled by this action:
   - blind transfer
   - transfer to a central mixer (some type of conference or forking)

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 28]


                       SIP Multiparty Framework

   - transfer to park server (park)
   - transfer to music on hold or announcement server
   - transfer to a "queue"
   - transfer to a service (such as Voice Dialogs service)
   - transition from local mixer to central mixer

5.3.2 Take

   The conversation space changes as follows:

        { B , C }  -->   { B , A }

   A forcibly replaces C with itself.  In most uses of this primitive,
   A is just "un-replacing" itself.

   Using the peer-to-peer approach, "A" sends:
        INVITE B  Replaces: <call leg between B and C>

   Using the 3pcc approach (all requests sent from controller)
        INVITE A (w/SDP of B)
        reINVITE B (w/SDP of A)
        BYE C

   Features enabled by this action:
   - transferee completes an attended transfer
   - retrieve from central mixer (not recommended)
   - retrieve from music on hold or park
   - retrieve from queue
   - call center take
   - voice portal resuming ownership of a call it originated
   - answering-machine style screening (pickup)
   - pickup of a ringing call (i.e. early dialog)

   Note: that pick up of a ringing call has perhaps some interesting
   additional requirements.  First of all it is an early dialog as
   opposed to an established dialog.  Secondly the party which is to
   pickup the call may only wish to do so only while it is an early
   dialog.  That is in the race condition where the ringing UA accepts
   just before it receives signaling from the party wishing to take the
   call, the taking party wishes to yield or cancel the take.  The goal
   is to avoid yanking an answered call from the called party.

5.3.3 Add

   The conversation space changes as follows:

        { A , B } -->    { A, B, C }

   A adds C to the conversation.

   Using the peer-to-peer approach, adding a party using local mixing
   requires no signaling.  To transition from a 2-party call or a
   locally mixed conference to centrally mixing A could send the
   following requests:

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 29]


                       SIP Multiparty Framework

        REFER B  Refer-To: mixer
        INVITE mixer
        BYE B

   To add a party to a central mixer:
        REFER C  Refer-To: mixer
               or
        REFER mixer  Refer-To: C

   Using the 3pcc approach to transition to centrally mixed, the
   controller would send:
        INVITE mixer leg 1 (w/SDP of A)
        INVITE mixer leg 2 (w/SDP of B)
        INVITE C (late SDP)
        reINVITE A (w/SDP of mixer leg 1)
        reINVITE B (w/SDP of mixer leg 2)
        INVITE mixer leg3 (w/SDP of C)

   To add a party to a central mixer:
        INVITE C (late SDP)
        INVITE mixer (w/SDP of C)

   Features enabled:
   - standard conference feature
   - call recording
   - answering-machine style screening (screening)

5.3.4 Local Join

   The conversation space changes like this:

        { A, B}  , {A, C}  -->  {A, B, C}

               or like this

        { A, B}  , {C, D}  -->  {A, B, C, D}

   A takes two conversation spaces and joins them together into a
   single space.

   Using the peer-to-peer approach, A can mix locally, or REFER the
   participants of both conversation spaces to the same central mixer
   (as in 5.3)

   For the 3pcc approach, the call flows for inserting participants,
   and joining and splitting conversation spaces are tedious yet
   straightforward, so these are left as an exercise for the reader.

   Features enabled:
   - standard conference feature
   - leaving a sidebar to rejoin a larger conference

5.3.5 Insert


Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 30]


                       SIP Multiparty Framework

   The conversation space changes like this:

        { B , C }  -->  {A, B, C }

   A inserts itself into a conversation space.

   A proposed mechanism for signaling this using the peer-to-peer
   approach is to send a new header in an INVITE with "joining"
   semantics.  For example:
        INVITE B  Join: <call id of B and C>

   If B accepted the INVITE, B would accept responsibility to setup the
   call legs and mixing necessary (for example: to mix locally or to
   transfer the participants to a central mixer)

   Features enabled:
   - barge-in
   - call center monitoring
   - call recording

5.3.6 Split
   { A, B, C, D } --> { A, B } , { C, D }

   If using a central mixer with peer-to-peer
   REFER C  Refer-To: mixer (new URI)
   REFER D  Refer-To: mixer (new URI)
   BYE C
   BYE D

   Features enabled:
   - sidebar conversations during a larger conference


5.3.7 Near-fork

   A participates in two conversation spaces simultaneously:

        { A, B } --> { B , A } & { A , C }


   A is a participant in two conversation spaces such that A sends the
   same media to both spaces, and renders media from both spaces,
   presumably by mixing or rendering the media from both.  We can
   define that A is the "anchor" point for both forks, each of which is
   a separate conversation space.

   This action is purely local implementation (it requires no special
   signaling).  Local features such as switching calls between the
   background and foreground are possible using this media
   relationship.

5.3.8 Far fork

   The conversation space diagram...

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 31]


                       SIP Multiparty Framework


        { A, B } --> { A ,  B } & { B , C }

   A requests B to be the "anchor" of two conversation spaces.

   For an example of using 3pcc to setup media forking, see [Media
   forking].  The session descriptions for forking are quite complex.
   Controllers should verify that endpoints can handle forked-media, by
   using some type of Requires header token.

   Two ways to setup this media relationship using peer-to-peer call
   control have been proposed:
   - the anchor receives a REFER with requires forked-media (implicit)
   - the anchor receives an INVITE with Fork-with header (explicit)

   Features enabled:
   - barge-in
   - voice portal services
   - whisper
   - hotword detection
   - sending DTMF somewhere else

   The above notation does not fully describe the media topology. Below
   are the four possible media topologies by which C might want to join
   the  A-B dialog.  For some of the above listed features there is a
   requirement to be able to specify any of these media  topologies as
   part of joining. In addition it is also a requirement that it be
   possible to change the media topology after the initial setup (e.g.
   in a reINVITE).  An example of this is a silent monitored
   conversation which is modified to be a full fledged conference to
   allow a call center supervisor to converse with the customer.

   The media topology can be separated into two perspectives.  The
   topology for the send and receive media streams for C.  For each of
   these streams C needs the ability to specify either point to point
   or mixed media.  This works out to the matrix where the ôsendö
   column indicates what happens with the media from C at B. The
   ôreceiveö column indicates what C wants to receive (mix or only BÆs
   media).  In the greater than 3 party case theoretically this cold be
   generalized to specify the set for the mix, however, from a
   pragmatic perspective the authors feel it is sufficient to constrain
   the description of the sets to all or nothing for now (i.e. point to
   point or max of all).

     Send    Receive
   1 Pt2pt   mix
   2 mix     mix
   3 Pt2pt   Pt2pt
   4 mix     Pt2pt

   For following examples:
   A is the customer
   B is the agent
   C is the supervisor

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 32]


                       SIP Multiparty Framework

   => and <= indicate the direction of media flow

   1. Send: point to point, Receive: mix
   Example application: silent monitoring or coaching
   A <= B  (point to point, only B hears C)
   A => B
   (A+B) => C (C gets mix of A + B)
   B <= C

   2. Send: mix, Receive: mix
   Example application: Normal Conference
   A <= (B+C) (mix, A gets mix of B+C)
   A => B
   (A+B) => C (C gets mix of A + B)
   B <= C

   3. Send: point to point, Receive: point to point
   Example application: Whisper/Sidebar
   A <= B (point to point, only B hears C)
   A => B
   B => C (point to point, C hears only B)
   B <= C

   4. Send: mix, Receive: point to point
   Example application: Recorded Conversation
   C û Voice Recorder
   A <= B (point to point, only B hears C)
   A => B
   (A+B) => C (C gets mix of A + B)
   B <= C

6 Putting it all together

   These example features should require an amply robust set of
   services to demonstrate a useful set of primitives.  A summary of
   these features is listed below.  Implementation of features with an
   asterisk (*) are described briefly in Section 6.1.


   Example Features:
   Call Hold                  [Offer/Answer] for SIP
   Call Waiting               Local Implementation
   Blind Transfer             [cc-transfer]
   Attended Transfer          [cc-transfer]
   Consultative transfer      [cc-transfer]
   Conference Call            [conf-models]
   Call Park                  *[examples]
   Call Pickup                *[examples]
   Music on Hold              *[examples]
   Call Monitoring            *Insert
   Barge-in                   *Insert or Far-Fork
   Hotline                    Local Implementation
   Autoanswer                 Local URI convention
   Speed dial                 Local Implementation

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 33]


                       SIP Multiparty Framework

   Intercom                   *Speed dial + autoanswer
   Speakerphone paging        *Speed dial + autoanswer
   Call Return                Proxy feature
   Inbound Call Screening     Proxy or Local implementation
   Outbound Call Screening    Proxy feature
   Call Forwarding            Proxy or Local implementation
   Message Waiting            [msg-waiting]
   Do Not Disturb             [presence]
   Distinctive ring           *Proxy or Local implementation
   Automatic Callback         2 person presence-based conference
   Find-Me                    Proxy service based on presence
   Whispered call waiting     Local implementation
   Voice message screening    *
   Presence-based Conferencing*call when presence = available
   IM Conference Alerts       subscribe to conference status
   Single Line Extension      *
   Click-to-dial              *
   Pre-paid calling           *
   Voice Portal               *


6.1 Feature Solutions

   The following sections illustrates how some of the primitives can be
   put together to build some powerful and interesting features.

6.1.1 Call Park

   Call park requires the ability to: put a dialog some place,
   advertise it to users in a pickup group and to uniquely identify it
   in a means that can be communicated (including human voice).  The
   dialog can be held locally on the UA parking the dialog or
   alternatively transferred to the park service for the pickup group.
   The parked dialog then needs to be labeled (e.g. orbit 12) in a way
   that can be communicated to the party that is to pick up the call.
   The UAs in the pick up group discovers the parked dialog(s) via
   [call-leg] from the park service.  If the dialog is parked locally
   the park service merely aggregates the parked call states from the
   set of UAs in the pickup up group.

6.1.2 Call Pickup

   There are two different features which are called call pickup.  The
   first is the pickup of a parked dialog.  The UA from which the
   dialog is to be picked up subscribes to the call state [call-leg] of
   the park service or the UA which has locally parked the dialog.
   Dialogs which are parked should be labeled with an identifier.  The
   labels are used by the UA to allow the user to indicate which dialog
   is to be picked up.  The UA picking up the call invoked the URL in
   the call state which is labeled as replace-remote.

   The other call pickup feature involves picking up an early dialog
   (typically ringing).  This feature uses some of the same primitives
   as the pick up of a parked call.  The call state of the UA ringing

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 34]


                       SIP Multiparty Framework

   phone is advertised using [call-leg].  The UA which is to pickup the
   early dialog subscribes either directly to the ringing UA or to a
   service aggregating the states for UAs in the pickup group.  The
   call state identifies early dialogs.  The UA uses the call state(s)
   to help the user choose which early dialog that is to be picked up.
   The UA then invokes the URL in the call state labeled as replace-
   remote.

6.1.3 Music on Hold

   Music on hold can be implemented a number of ways.  One way is to
   transfer the held call to a holding service.  When the UA wishes to
   take the call off hold it basically performs a take on the call from
   the holding service.  This involves subscribing to call state on the
   holding service and then invoking the URL in the call state labeled
   as replace-remote.

   Alternatively music on hold can be performed as a local mixing
   operation.  The UA holding the call can mix in the music from the
   music service via RTP (i.e. an additional dialog) or RTSP or other
   streaming media source.  This approach is simpler (i.e. the held
   dialog does not move so there is less chance of loosing them) from a
   protocol perspective, however it does use more LAN bandwidth and
   resources on the UA.

6.1.4 Call Monitoring

   Call monitoring is a [join] operation.  The monitoring UA sends a
   Join to the dialog it wants to listen to.  It is able to discover
   the dialog via the call state [call-leg] on the monitored UA.  The
   monitoring UA sends SDP in the INVITE which indicates receive only
   media {offer/answer].  IN addition the monitoring UA should indicate
   that it wants to receive a mix (see Error! Reference source not
   found.).  As the UA is monitoring only it does not matter whether
   the UA indicates it wishes the send stream be mix or point to point.

6.1.5 Barge-in

   Barge-in works the same as call monitoring except that it must
   indicate that the send media stream to be mixed so that all of the
   other parties can hear the stream from UA barging in.

6.1.6 Intercom

   The UA initiates a dialog using INVITE in the ordinary way [bis].
   The calling UA then signals the paged UA to answer the call.  The
   calling UA may discover the URL to answer the call via the call
   state [call-leg] of the called UA. The called UA accepts the INVITE
   with a 200 Ok and automatically enables the speakerphone.

   Alternatively this can be a local decision for the UA to answer
   based upon called party identification.



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 35]


                       SIP Multiparty Framework

6.1.7 Speakerphone paging

   Speakerphone paging can be implemented using either multicast or
   through a simple multipoint mixer.  In the multicast solution the
   paging UA sends a multicast INVITE [bis] with send only media in the
   [SDP] (see also [offer/answer]).  The automatic answer and enabling
   of the speakerphone is a locally configured decision on the paged
   UAs.  The paging UA sends RTP via the multicast address indicated in
   the SDP.

   The multipoint solution is accomplished by sending an INVITE to the
   multipoint mixer.  The mixer is configured to automatically answer
   the dialog.  The paging UA then sends [REFER] requests for each of
   the UAs that are to become paging speakers (The UA is likely to send
   out a single REFER which is parallel forked by the proxy server).
   The UAs performing as paging speakers are configured to
   automatically answer based upon caller identification (e.g. To
   field, URI or Referred-To headers).

6.1.8 Distinctive ring

   The target UA either makes a local decision based on information in
   an incoming INVITE (To, From, Contact, Request-URI) or trusts an
   Alert-Info header provded by the caller or inserted by a trusted
   proxy.  In the latter case, the UA fetches the content described in
   the URI (typically via http) and renders it to the user.

6.1.9 Voice message screening

   At first, this is the same as call monitoring.  In this case the
   voicemail service is one of the UAs.  The UA screening the message
   monitors the call on the voicemail service, and also subscribes to
   call-leg information.  If the user screening their messages decides
   to answer, they perform a Take from the voicemail system (for
   example, send an INVITE with Replaces to the UA leaving the message)

6.1.10 Single Line Extension

   Incoming calls ring all the extensions through basic parallel
   forking [bis].  Each extension subscribes to call-leg events from
   each other extension.  While one user has an active call, any other
   UA extension can insert itself into that conversation (it already
   knows the call-leg information)in the same way as barge-in.

6.1.11 Click-to-dial

   The application or server which hosts the click-to-dial application
   captures the URL to be dialed and can setup the call using 3pcc or
   can send a [REFER] request to the UA which is to dial the address.
   As users sometimes change their mind or wish to give up listing to a
   ringing or voicemail answered phone, this application illustrates
   the need to also have the ability to remotely hangup a call.



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 36]


                       SIP Multiparty Framework

6.1.12 Pre-paid calling

   For prepaid calling, the user's media always passes through a device
   which is trusted by the pre-paid provider.  This may be the other
   endpoint (for example a PSTN gateway).  In either case, an
   intermediary proxy or B2BUA can periodically verify the amount of
   time available on the pre-paid account, and use the session-timer
   extension to cause the trusted endpoint (gateway) or intermediary
   (media relay) to send a reINVITE before that time runs out.  During
   the reINVITE, the SIP intermediary can reverify the account and
   insert another session-timer header.

   Note that while most pre-paid systems on the PSTN use an IVR to
   collect the account number and destination, this isn't strictly
   necessary for a SIP-originated prepaid call.  SIP requests and SIP
   URIs are sufficiently expressive to convey the final destination,
   the provider of the prepaid service, the location from which the
   user is calling, and the prepaid account they want to use.  If a
   pre-paid IVR is used, the mechanism described below (Voice Portals)
   can be combined as well.

6.1.13 Voice Portal

   A voice portal is essentially a complex collection of voice dialogs
   used to access interesting content.  One of the most desirable call
   control features of a Voice Portal is the ability to start a new
   outgoing call from within the context of the Portal (to make a
   restauraunt reservation, or return a voicemail message for example).
   Once the new call is over, the user should be able to return to the
   Portal by pressing a special key, using some DTMF sequence (ex: a
   very long pound or hash tone), or by speaking a hotword (ex: "Main
   Menu").

   In order to accomplish this, the Voice Portal starts with the
   following media relationship:

       { User , Voice Portal }

   The user then asks to make an outgoing call.  The Voice Portal asks
   the User to perform a Far-Fork.  In other words the Voice Portal
   wants the following media relationship:

       { Target , User }  &  { User , Voice Portal }

   The Voice Portal is now just listening for a hotword or the
   appropriate DTMF.  As soon as the user indicates they are done, the
   Voice Portal Takes the call from the old Target, and we are back to
   the original media relationship.

   This feature can also be used by the account number and phone number
   collection menu in a pre-paid calling service.  A user can press a
   DTMF sequence which presents them with the a



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 37]


                       SIP Multiparty Framework

7 Security Considerations

   Call Control primitives provide a powerful set of features that can
   be dangerous in the hands of an attacker.  To complicate matters,
   call control primitives are likely to be automatically authorized
   without direct human oversight.

   The class of attacks which are possible using these tools include
   the ability to eavesdrop on calls, disconnect calls, redirect calls,
   render irritating content (including ringing) at a user agent, cause
   an action that has billing consequences, subvert billing (theft-of-
   service), and obtain private information.  Call control extensions
   must take extra care to describe how these attacks will be
   prevented.

   We can also make some general observations about authorization and
   trust with respect to call control.  The security model is
   dramatically dependent on the signaling model chosen (see section
   4.2)

   Let us first examine the security model used in the 3pcc approach.
   All signaling goes through the controller, which is a trusted
   entity.  Traditional SIP authentication and hop-by-hop encrpytion
   and message integrity work fine in this environment, but end-to-end
   encrpytion and message integrity may not be possible.

   When using the peer-to-peer approach, call control actions and
   primitives can be legitimately initiated by a) an existing
   participant in the conversation space, b) a former participant in
   the conversation space, or c) an entity trusted by one of the
   participants.  For example, a participant always initiates a
   transfer; a retrieve from Park (a take) is initiated on behalf of a
   former participant; and a barge-in (insert or far-fork) is initiated
   by a trusted entity (an operator for example).

   Authenticating requests by an existing participant or a trusted
   entity can be done with baseline SIP mechanisms.  In the case of
   features initiated by a former participant, these should be
   protected against replay attacks by using a unique name or
   identifier per invocation.  The Replaces header exhibits this
   behavior as a by-product of its operation (once a Replaces operation
   is successful, the call-leg being Replaced no longer exists).  For
   other requests, a "one-time" Request-URI may be provided to the
   feature invoker.

   To authorize call control primitives that trigger special behavior
   (such as an INVITE with Replace, Join, or Fork semantics), the
   receiving user agent may have trouble finding appropriate
   credentials with which to challenge or authorize the request, as the
   sender may be completely unknown to the receiver, except through the
   introduction of a third party.  These credentials need to be passed
   transitively in some way or fetched in an event body, for example.



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 38]


8 References

   [SIP] M. Handley, E. Schooler, and H. Schulzrinne, "SIP: Session
   Initiation Protocol", RFC2543, Internet Engineering Task Force,
   Nov 1998.

   [RFC2026] S Bradner, "The Internet Standards Process -- Revision 3",
   RFC2026 (BCP), IETF, October 1996.

   [RFC2119] S. Bradner, "Key words for use in RFCs to indicate
   requirement     levels," Request for Comments (Best Current
   Practice) 2119, Internet     Engineering Task Force, Mar. 1997.

   [REFER] R. Sparks, "The Refer Method", Internet Draft <draft-ietf-
   sip-refer-02>, IETF, October 30, 2001, Work in progress.

   [3pcc] J. Rosenberg, J. Peterson, H. Schulzrinne, G. Camarillo,
   "Third Party Call Control in SIP", Internet Draft <draft-rosenberg-
   sip-3pcc-02.txt>, IETF;  March 2001.  Work in progress

   [transfer] R. Sparks, "SIP Call Control - Transfer", Internet Draft
   <draft-ietf-sip-cc-transfer-04.txt>, IETF; Feb. 2001. Work in
   progress.

   [Replaces] B. Biggs, R. Dean, R. Mahy, "The SIP Replaces Header",
   Internet Draft <draft-ietf-sip-replaces-00.txt>, IETF, Nov. 2001.
   Work in progress.

   [conf-models]  J. Rosenberg, H. Schulzrinne, "Models for Multi Party
   Conferencing in SIP", Internet Draft <draft-rosenberg-sip-
   conferencing-models-00.txt>, IETF; Nov. 2000. Work in progress.

   [service examples] A. Johnston, R. Sparks, C. Cunningham, S.
   Donovan, K. Summers, "SIP Service Examples" Internet Draft <draft-
   ietf-sip-service-examples-03.txt>, IETF, June 2002, Work in
   progress.

   [Join] R. Mahy, D. Petrie, "The SIP Join and Fork Headers", Internet
   Draft <draft-mahy-sipping-join-and-fork-00.txt>, IETF, November
   2001, Work in progress.

   [RTP] H. Schulzrinne , S. Casner , R. Frederick , V. Jacobson ,
   "RTP: A Transport Protocol for Real-Time Applications", Request for
   Comments (Standards Track)1889, IETF, January 1996

   [SDP] H. Schulzrinne M. Handley, V. Jacobson, "SDP: Session
   Description Protocol", Request for Comments (Standards Track) 2327,
   Internet Engineering Task Force, April 1998

   [events] A. Roach, "SIP-Specific Event Notification",Internet Draft
   <draft-ietf-sip-events-03.txt>, IETF, February 2002, Work in
   progress.



Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 39]


                       SIP Multiparty Framework

   [offer/answer] J. Rosenberg, H. Schulzrinne, "An Offer/Answer Model
   with SDP", Internet Draft <draft-ietf-mmusic-sdp-offer-answer-
   01.txt>, IETF, February 21, 2002, Work in progress.

   [caller prefs] J. Rosenberg, "SIP Caller Preferences and Callee
   Capabilities",Internet Draft <draft-ietf-sip-callerprefs-05.txt>,
   IETF, November 21, 2001, Work in progress.

   [msg waiting] R. Mahy, I. Slain, "Message Waiting in SIP",Internet
   Draft <draft-mahy-sip-message-waiting-02.txt>, IETF, July 2001, Work
   in progress.

   [Presence] Rosenberg et al., "SIP Extensions for Presence", Internet
   Draft <draft-ietf-simple-presence-04.txt>, IETF, November 21, 2001,
   Work in progress.

   [visited] D. Oran, H. Schulzrinne, "The Visited Header",Internet
   Draft <>, IETF, date, Work in progress.

   [app components] , "",Internet Draft <>, IETF, date, Work in
   progress.

   [ms-uri] J. Van Dyke, E. Burger, "SIP URI Conventions for Media
   Servers",Internet Draft <draft-burger-sipping-msuri-01.txt>, IETF,
   November 21, 2001, Work in progress.

   [call-pkg] J. Rosenberg, H. Schulzrinne, "SIP Event Packages for
   Call Leg and Conference State", Internet Draft <draft-rosenberg-sip-
   call-package-00.txt>, IETF, July 13, 2001, Work in progress.

   [enum] , "",Internet Draft <>, IETF, date, Work in progress.

   [http]  R. Fielding et al, "Hypertext Transfer Protocol --
   HTTP/1.1", Request for Comments (Standards Track) 2616, Internet
   Engineering Task Force, June 1999

   [rtsp] H. Schulzrinne, A. Rao, R. Lanphier, "Real Time Streaming
   Protocol (RTSP)", Request for Comments (Standards Track) 2326,
   Internet Engineering Task Force, April 1998

   [mrcp] S. Shanmugham, P. Monaco, B. Eberman, "MRCP: Media Resource
   Control Protocol", Internet Draft <draft-shanmugham-mrcp-01.txt>,
   IETF, November 20, 2001, Work in progress.

   [VoiceXML] S. McGlashan et al, ôVoice Extensible Markup Language
   (VoiceXML) Version 2.0ö, W3C Working Draft, 23 October 2001, Work in
   progress.

   [H.323]

   [tel URL]

   [caller-prefs]


Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 40]


                       SIP Multiparty Framework

   [session timer]

   [service context]

   [avt tones]

   [GSM]

   [MPEG2]

   [G.711]

   [H.261]

   [H.450]

   [JTAPI]

   [CSTA]

   [mrcp-sip] , "",Internet Draft <draft-robinson-mrcp-sip-00.txt>,
   IETF, date, Work in progress.

   [distributed full mesh conf]

   [Media forking] M. Shankar, "SIP Forked Media", Internet Draft
   <draft-shankar-sip-forked-media-00.txt>, IETF, Feb. 2001.  Work in
   progress.

   [PHONECTL] R. Dean, Belkind, B. Biggs, "PHONECTL: A Protocol for
   Remote Phone Control", Internet Draft <draft-dean-phonectl-03.txt>,
   IETF, Jan. 2001.  Work in progress.


9 To Do

   - Add diagrams to section 4.3.1, 4.3.2, and 4.3.3

   - Fix references

   - Define some semantics for authorization rules.  For example one
   could define a dictionary of primitives and/or perhaps define sets
   or classes of these primitives, then configure who is allowed to use
   them

10    Acknowledgments

   Thanks to all who attended the SIP interim meeting in February 2001
   for their support of the ideas behind this document.


11    Author's Addresses

   Rohan Mahy

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 41]


                       SIP Multiparty Framework

   Cisco Systems
   170 West Tasman Dr, MS: SJC-21/3/3
   Phone: +1 408 526 8570
   Email: rohan@cisco.com

   Ben Campbell
   dynamicsoft
   5100 Tennyson Parkway
   Suite 1200
   Plano, Texas 75024
   Email: bcampbell@dynamicsoft.com

   Alan Johnston
   WorldCom
   100 S. 4th Street
   St. Louis, Missouri 63104
   Email: alan.johnston@wcom.com

   Daniel G. Petrie
   Pingtel Corp.
   400 W. Cummings Park
   Suite 2200
   Woburn, MA 01801
   Phone: +1 781 938 5306
   Email: dpetrie@pingtel.com

   Jonathan Rosenberg
   dynamicsoft
   72 Eagle Rock Avenue
   First Floor
   East Hanover, NJ 07936
   Email: jdrosen@dynamicsoft.com

   Robert J. Sparks
   dynamicsoft
   5100 Tennyson Parkway
   Suite 1200
   Plano, TX  75024
   Email: rsparks@dynamicsoft.com


   Full Copyright Statement

   "Copyright (C) The Internet Society (date). 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 implementation 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
   document itself may not be modified in any way, such as by removing
   the copyright notice or references to the Internet Society or other
   Internet organizations, except as needed for the purpose of
   developing Internet standards in which case the procedures for

Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 42]


                       SIP Multiparty Framework

   copyrights defined in the Internet Standards process must be
   followed, or as required to translate it into languages other than
   English.

   The limited permissions 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 WARRANTY THAT THE USE OF THE INFORMATION
   HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.











































Mahy/Campbell/Johnston/Petrie/Sparks/Rosenberg  Exp:Aug 2002 [Page 43]