Network Working Group
Request for Comments: 5567
Category: Informational
T. Melanchuk, Ed.
Rain Willow Communications
June 2009

An Architectural Framework for Media Server Control

Status of This Memo

This memo provides information for the Internet community. It does not specify an Internet standard of any kind. Distribution of this memo is unlimited.

Copyright Notice

Copyright © 2009 IETF Trust and the persons identified as the document authors. All rights reserved.

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents in effect on the date of publication of this document (http://trustee.ietf.org/license-info). Please review these documents carefully, as they describe your rights and restrictions with respect to this document.

This document may contain material from IETF Documents or IETF Contributions published or made publicly available before November 10, 2008. The person(s) controlling the copyright in some of this material may not have granted the IETF Trust the right to allow modifications of such material outside the IETF Standards Process. Without obtaining an adequate license from the person(s) controlling the copyright in such materials, this document may not be modified outside the IETF Standards Process, and derivative works of it may not be created outside the IETF Standards Process, except to format it for publication as an RFC or to translate it into languages other than English.

Abstract

This document describes an architectural framework for Media Server control. The primary focus will be to define logical entities that exist within the context of Media Server control, and define the appropriate naming conventions and interactions between them.

Table of Contents

   1. Introduction ....................................................2
   2. Terminology .....................................................3
   3. Architecture Overview ...........................................4
   4. SIP Usage .......................................................7
   5. Media Control for IVR Services .................................10
      5.1. Basic IVR Services ........................................11
      5.2. IVR Services with Mid-Call Controls .......................11
      5.3. Advanced IVR Services .....................................11
   6. Media Control for Conferencing Services ........................12
      6.1. Creating a New Conference .................................14
      6.2. Adding a Participant to a Conference ......................14
      6.3. Media Controls ............................................15
      6.4. Floor Control .............................................16
   7. Security Considerations ........................................21
   8. Acknowledgments ................................................22
   9. Contributors ...................................................22
   10. Informative References ........................................23

1. Introduction

Application Servers host one or more instances of a communications application. Media Servers provide real-time media processing functions. This document presents the core architectural framework to allow Application Servers to control Media Servers. An overview of the architecture describing the core logical entities and their interactions is presented in Section 3. The requirements for Media Server control are defined in [RFC5167].

The Session Initiation Protocol (SIP) [RFC3261] is used as the session establishment protocol within this architecture. Application Servers use it both to terminate media streams on Media Servers and to create and manage control channels for Media Server control between themselves and Media Servers. The detailed model for Media Server control together with a description of SIP usage is presented in Section 4.

Several services are described using the framework defined in this document. Use cases for Interactive Voice Response (IVR) services are described in Section 5, and conferencing use cases are described in Section 6.

2. Terminology

The following terms are defined for use in this document in the context of Media Server control:

   Application Server (AS):  A functional entity that hosts one or more
      instances of a communication application.  The application server
      may include the conference policy server, the focus, and the
      conference notification server, as defined in [RFC4353].  Also, it
      may include communication applications that use IVR or
      announcement services.
   
   Media Functions:  Functions available on a Media Server that are used
      to supply media services to the AS.  Some examples are Dual-Tone
      Multi-Frequency (DTMF) detection, mixing, transcoding, playing
      announcement, recording, etc.
   
   Media Resource Broker (MRB):  A logical entity that is responsible
      for both the collection of appropriate published Media Server (MS)
      information and supplying of appropriate MS information to
      consuming entities.  The MRB is an optional entity and will be
      discussed in a separate document.
   
   Media Server (MS):  The media server includes the mixer as defined in
      [RFC4353].  The media server plays announcements, it processes
      media streams for functions like DTMF detection and transcoding.
      The media server may also record media streams for supporting IVR
      functions like announcing conference participants.  In the
      architecture for the 3GPP IP Multimedia Subsystem (IMS) a Media
      Server is referred to as a Media Resource Function (MRF).
   
   Media Services:  Application service requiring media functions such
      as Interactive Voice Response (IVR) or media conferencing.
   
   Media Session:  From the Session Description Protocol (SDP)
      specification [RFC4566]: "A multimedia session is a set of
      multimedia senders and receivers and the data streams flowing from
      senders to receivers.  A multimedia conference is an example of a
      multimedia session."
   
   MS Control Channel:  A reliable transport connection between the AS
      and MS used to exchange MS Control PDUs.  Implementations must
      support the Transport Control Protocol (TCP) [RFC0793] and may
      support the Stream Control Transmission Protocol (SCTP) [RFC4960].
      Implementations must support TLS [RFC5246] as a transport-level
      security mechanism although its use in deployments is optional.
   
   MS Control Dialog:  A SIP dialog that is used for establishing a
      control channel between the user agent (UA) and the MS.
   
   MS Control Protocol:  The protocol used for by an AS to control an
      MS.  The MS Control Protocol assumes a reliable underlying
      transport protocol for the MS Control Channel.
   
   MS Media Dialog:  A SIP dialog between the AS and MS that is used for
      establishing media sessions between a user device such as a SIP
      phone and the MS.

The definitions for AS, MS, and MRB above are taken from [RFC5167].

3. Architecture Overview

A Media Server (MS) is a network device that processes media streams. Examples of media processing functionality may include:

  • Control of the Real-Time Protocol (RTP) [RFC3550] streams using the Extended RTP Profile for Real-time Transport Control Protocol (RTCP)-Based Feedback (RTP/AVPF) [RFC4585].
  • Mixing of incoming media streams.
  • Media stream source (for multimedia announcements).
  • Media stream processing (e.g., transcoding, DTMF detection).
  • Media stream sink (for multimedia recordings).

An MS supplies one or more media processing functionalities, which may include others than those illustrated above, to an Application Server (AS). An AS is able to send a particular call to a suitable MS, either through discovery of the capabilities that a specific MS provides or through the use of a Media Resource Broker.

The type of processing that a Media Server performs on media streams is specified and controlled by an Application Server. Application Servers are logical entities that are capable of running one or more instances of a communications application. Examples of Application Servers that may interact with a Media Server are an AS acting as a Conference 'Focus' as defined in [RFC4353], or an IVR application using a Media Server to play announcements and detect DTMF key presses.

Application servers use SIP to establish control channels between themselves and MSs. An MS Control Channel implements a reliable transport protocol that is used to carry the MS Control Protocol. A

SIP dialog used to establish a control channel is referred to as an MS Control Dialog.

Application Servers terminate SIP [RFC3261] signaling from SIP User Agents and may terminate other signaling outside the scope of this document. They use SIP Third Party Call Control [RFC3725] (3PCC) to establish, maintain, and tear down media streams from those SIP UAs to a Media Server. A SIP dialog used by an AS to establish a media session on an MS is referred to as an MS Media Dialog.

Media streams go directly between SIP User Agents and Media Servers. Media Servers support multiple types of media. Common supported RTP media types include audio and video, but others such as text and the Binary Floor Control Protocol (BFCP) [RFC4583] are also possible. This basic architecture, showing session establishment signaling between a single AS and MS is shown in Figure 1 below.

           +-------------+                         +--------------+
           |             | SIP (MS Control Dialog) |              |
           | Application |<----------------------->|     Media    |
           |   Server    |                         |    Server    |
           |             |<----------------------->|              |
           +-------------+ SIP (MS Media Dialog)   +--------------+
                       ^                               ^
                        \                              | RTP/SRTP
                         \                             |  audio/
                          \                            | video/etc)
                           \                           |
                            \                          v
                             \                 +--------------+
                              \     SIP        |              |
                               +-------------->|      SIP     |
                                               |  User Agent  |
                                               |              |
                                               +--------------+

Figure 1: Basic Signaling Architecture

The architecture must support a many-to-many relationship between Application Servers and Media Servers. In real world deployments, an Application Server may interact with multiple Media Servers and/or a Media Server may be controlled by more than one Application Server.

Application Servers can use the SIP URI as described in [RFC4240] to request basic functions from Media Servers. Basic functions are characterized as requiring no mid-call interactions between the AS and MS. Examples of these functions are simple announcement-playing or basic conference-mixing where the AS does not need to explicitly control the mixing.

Most services however have interactions between the AS and MS during a call or conference. The type of interactions can be generalized as follows:

  • commands from an AS to an MS to request the application or configuration of a function. The request may apply to a single media stream, multiple media streams associated with multiple SIP dialogs, or to properties of a conference mix.
  • responses from an MS to an AS reporting on the status of particular commands.
  • notifications from an MS to an AS that report results from commands or notify changes to subscribed status.

Commands, responses, and notifications are transported using one or more dedicated control channels between the Application Server and the Media Server. Dedicated control channels provide reliable, sequenced, peer-to-peer transport for Media Server control interactions. Implementations must support the Transport Control Protocol (TCP) [RFC0793] and may support the Stream Control Transmission Protocol (SCTP) [RFC4960]. Because MS control requires sequenced reliable delivery of messages, unreliable protocols such as the User Datagram Protocol (UDP) are not suitable. Implementations must support TLS [RFC5246] as a transport-level security mechanism although its use in deployments is optional. A dedicated control channel is shown in Figure 2 below.

             +-------------+                     +--------------+
             |             |                     |              |
             | Application |   MS ctrl channel   |     Media    |
             |   Server    |<------------------->|    Server    |
             |             |                     |              |
             +-------------+                     +--------------+
                                                         ^ ^ ^
                                                RTP/SRTP | | |
                                                (audio/  | | |
                                              video/etc) | | |
                                                         | | v
                                                     +---|-v-------+
                                                   +-|---v-------+ |
                                                 +-|-----------+ | |
                                                 |             | | |
                                                 |     SIP     | | |
                                                 | User Agent  | |-+
                                                 |             |-+
                                                 +-------------+

Figure 2: Media Server Control Architecture

Both Application Servers and Media Servers may interact with other servers for specific purposes beyond the scope of this document. For example, Application Servers will often communicate with other infrastructure components that are usually based on deployment requirements with links to back-office data stores and applications. Media Servers will often retrieve announcements from external file servers. Also, many Media Servers support IVR dialog services using VoiceXML [W3C.REC-voicexml20-20040316]. In this case, the MS interacts with other servers using HTTP during standard VoiceXML processing. VoiceXML Media Servers may also interact with speech engines (for example, using the Media Resource Control Protocol version 2 (MRCPv2)) for speech recognition and generation purposes.

Some specific types of interactions between Application and Media servers are also out of scope for this document. MS resource reservation is one such interaction. Also, any interactions between Application Servers, or between Media Servers, are also out of scope.

4. SIP Usage

The Session Initiation Protocol (SIP) [RFC3261] was developed by the IETF for the purposes of initiating, managing, and terminating multimedia sessions. The popularity of SIP has grown dramatically since its inception and is now the primary Voice over IP (VoIP) protocol. This includes being selected as the basis for architectures such as the IP Multimedia Subsystem (IMS) in 3GPP and included in many of the early live deployments of VoIP-related systems. Media servers are not a new concept in IP telephony networks and there have been numerous signaling protocols and techniques proposed for their control. The most popular techniques to date have used a combination of SIP and various markup languages to convey media service requests and responses.

As discussed in Section 3 and illustrated in Figure 1, the logical architecture described by this document involves interactions between an Application Server (AS) and a Media Server (MS). The SIP interactions can be broken into "MS media dialogs" that are used between an AS and an MS to establish media sessions between an endpoint and a Media Server, and "MS control dialogs" that are used to establish and maintain MS control channels.

SIP is the primary signaling protocol for session signaling and is used for all media sessions directed towards a Media Server as described in this document. Media Servers may support other signaling protocols but this type of interaction is not considered here. Application Servers may terminate non-SIP signaling protocols but must gateway those requests to SIP when interacting with a Media Server.

SIP will also be used for the creation, management, and termination of the dedicated MS control channel(s). Control channel(s) provide reliable sequenced delivery of MS Control Protocol messages. The Application and Media Servers use the SDP attributes defined in [RFC4145] to allow SIP negotiation of the control channel. A control channel is closed when SIP terminates the corresponding MS control dialog. Further details and example flows are provided in the SIP Control Framework [SIP-CTRL-FW]. The SIP Control Framework also includes basic control message semantics corresponding to the types of interactions identified in Section 3. It uses the concept of "packages" to allow domain-specific protocols to be defined using the Extensible Markup Language (XML) [W3C.REC-xml-20060816] format. The MS Control Protocol is made up of one or more packages for the SIP Control Framework.

Using SIP for both media and control dialogs provides a number of inherent benefits over other potential techniques. These include:

  1. The use of SIP location and rendezvous capabilities, as defined in [RFC3263]. This provides core mechanisms for routing a SIP request based on techniques such as DNS SRV and NAPTR records. The SIP infrastructure makes heavy use of such techniques.
  1. The security and identity properties of SIP; for example, using TLS for reliably and securely connecting to another SIP-based entity. The SIP protocol has a number of identity mechanisms that can be used. [RFC3261] provides an intra-domain digest- based mechanism and [RFC4474] defines a certificate-based inter- domain identity mechanism. SIP with S/MIME provides the ability to secure payloads using encrypted and signed certificate techniques.
  1. SIP has extremely powerful and dynamic media-negotiation properties as defined in [RFC3261] and [RFC3264].
  1. The ability to select an appropriate SIP entity based on capability sets as discussed in [RFC3840]. This provides a powerful function that allows Media Servers to convey a specific capability set. An AS is then free to select an appropriate MS based on its requirements.
  1. Using SIP also provides consistency with IETF protocols and usages. SIP was intended to be used for the creation and management of media sessions, and this provides a correct usage of the protocol.

As mentioned previously in this section, media services using SIP are fairly well understood. Some previous proposals suggested using the SIP INFO [RFC2976] method as the transport vehicle between the AS and MS. Using SIP INFO in this way is not advised for a number of reasons, which include:

  • INFO is an opaque request with no specific semantics. A SIP endpoint that receives an INFO request does not know what to do with it based on SIP signaling.
  • SIP INFO was not created to carry generic session control information along the signaling path, and it should only really be used for optional application information, e.g., carrying mid-call Public Switched Telephone Network (PSTN) signaling messages between PSTN gateways.
  • SIP INFO traverses the signaling path, which is an inefficient use for control messages that can be routed directly between the AS and MS.
  • [RFC3261] contains rules when using an unreliable protocol such as UDP. When a packet reaches a size close to the Maximum Transmission Unit (MTU), the protocol should be changed to TCP. This type of operation is not ideal when constantly dealing with large payloads such as XML-formatted MS control messages.

5. Media Control for IVR Services

One of the functions of a Media Server is to assist an Application Server that is implementing IVR services by performing media processing functions on media streams. Although "IVR" is somewhat generic terminology, the scope of media functions provided by an MS addresses the needs for user interaction dialogs. These functions include media transcoding, basic announcements, user input detection (via DTMF or speech), and media recording.

A particular IVR or user dialog application typically requires the use of several specific media functions, as described above. The range and complexity of IVR dialogs can vary significantly, from a simple single announcement play-back to complex voice mail applications.

As previously discussed, an AS uses SIP [RFC3261] and SDP [RFC4566] to establish and configure media sessions to a Media Server. An AS uses the MS control channel, established using SIP, to invoke IVR requests and to receive responses and notifications. This topology is shown in Figure 3 below.

      +-------------+             SIP              +-------------+
      | Application |<---------------------------->|   Media     |
      |    Server   | (media & MS Control dialogs) |   Server    |
      |             |                              |             |
      |             |  MS Control Protocol (IVR)   |             |
      |             |<---------------------------->| (IVR media  |
      | (App logic) |       (CtrlChannel)          | functions)  |
      +-------------+                              +-------------+
      
             ^                                            ^^
              \                                           ||  R
               \                                          ||  T
                \                                         ||  P
                 \                                        ||  /
                  \                                       ||  S
                   \                                      ||  R
                    \                                     ||  T
                     \                                    ||  P
                      \                                   vv
                       \    call signaling           +-----------+
                        ---------------------------->|   User    |
                              (e.g., SIP)            | Equipment |
                                                     +-----------+

Figure 3: IVR Topology

The variety in complexity of Application Server IVR services requires support for different levels of media functions from the Media Server as described in the following sub-sections.

5.1. Basic IVR Services

For simple basic announcement requests, the MS control channel, as depicted in Figure 3 above, is not required. Simple announcement requests may be invoked on the Media Server using the SIP URI mechanism defined in [RFC4240]. This interface allows no digit detection or collection of user input and no mid-call dialog control. However, many applications only require basic media services, and the processing burden on the Media Server to support more complex interactions with the AS would not be needed in that case.

5.2. IVR Services with Mid-Call Controls

For more complex IVR dialogs, which require mid-call interaction and control between the Application Server and the Media Server, the MS control channel (as shown in Figure 3 above) is used to invoke specific media functions on the Media Server. These functions include, but are not limited to, complex announcements with barge-in facility, user-input detection and reporting (e.g., DTMF) to an Application Server, DTMF and voice-activity controlled recordings, etc. Composite services, such as play-collect and play-record, are also addressed by this model.

Mid-call control also allows Application Servers to subscribe to IVR- related events and for the Media Server to notify the AS when these events occur. Examples of such events are announcement completion events, record completion events, and reporting of collected DTMF digits.

5.3. Advanced IVR Services

Although IVR services with mid-call control, as described above, provide a comprehensive set of media functions expected from a Media Server, the advanced IVR services model allows a higher level of abstraction describing application logic, as provided by VoiceXML, to be executed on the Media Server. Invocation of VoiceXML IVR dialogs may be via the "Prompt and Collect" mechanism of [RFC4240]. Additionally, the IVR control protocol can be extended to allow VoiceXML requests to also be invoked over the MS control channel. VoiceXML IVR services invoked on the Media Server may require an HTTP interface (not shown in Figure 3) between the Media Server and one or more back-end servers that host or generate VoiceXML documents. The back-end server(s) may or may not be physically separate from the Application Server.

6. Media Control for Conferencing Services

[RFC4353] describes the overall architecture and protocol components needed for multipoint conferencing using SIP. The framework for centralized conferencing [RFC5239] extends the framework to include a protocol between the user and the conferencing server. [RFC4353] describes the conferencing server decomposition but leaves the specifics open.

This section describes the decomposition and discusses the functionality of the decomposed functional units. The conferencing factory and the conference focus are part of the Application Server described in this document.

An Application Server uses SIP Third Party Call Control [RFC3725] to establish media sessions from SIP user agents to a Media Server. The same mechanism is used by the Application Server as described in this section to add/remove participants to/from a conference, as well as to handle the involved media streams set up on a per-user basis. Since the XCON framework has been conceived as protocol-agnostic when talking about the Call Signaling Protocol used by users to join a conference, an XCON-compliant Application Server will have to take care of gatewaying non-SIP signaling negotiations. This is in order to set up and make available valid SIP media sessions between itself and the Media Server, while still keeping the non-SIP interaction with the user in a transparent way.

                +------------+             +------------+
                |            | SIP (2m+1c) |            |
                | Application|-------------|   Media    |
                |   Server   |             |   Server   |
                |  (Focus)   |-------------|  (Mixer)   |
                |            | CtrlChannel |            |
                +------------+             +------------+
                    |      \                    .. .
                    |       \\            RTP...   .
                    |         \\           ..      .
                    |     H.323  \\      ...       .
                SIP |             \\ ...           .RTP
                    |              ..\             .
                    |           ...   \\           .
                    |        ...        \\         .
                    |      ..             \\       .
                    |   ...                 \\     .
                    | ..                      \    .
               +-----------+              +-----------+
               |Participant|              |Participant|
               +-----------+              +-----------+

Figure 4: Conference Topology

To complement the functionality provided by 3PCC and by the XCON control protocol, the Application Server makes use of a dedicated Media Server control channel in order to set up and manage media conferences on the Media Server. Figure 4 shows the signaling and media paths for a two-participant conference. The three SIP dialogs between the AS and MS establish one control session (1c) and two media sessions (2m) from the participants (one originally signaled using H.323 and then gatewayed into SIP and one signaled directly in SIP).

As a conference focus, the Application Server is responsible for setting up and managing a media conference on the Media Servers, in order to make sure that all the media streams provided in a conference are available to its participants. This is achieved by using the services of one or more mixer entities (as described in RFC 4353), whose role as part of the Media Server is described in this section. Services required by the Application Server include, but are not limited to, means to set up, handle, and destroy a new media conference, adding and removing participants from a conference, managing media streams in a conference, controlling the layout and the mixing configuration for each involved media, allowing per-user custom media profiles, and so on.

As a mixer entity, in such a multimedia conferencing scenario, the Media Server receives a set of media streams of the same type (after transcoding if needed) and then takes care of combining the received media in a type-specific manner, redistributing the result to each authorized participant. The way each media stream is combined, as well as the media-related policies, is properly configured and handled by the Application Server by means of a dedicated MS control channel.

To summarize, the AS needs to be able to manage Media Servers at a conference and participant level.

6.1. Creating a New Conference

When a new conference is created, as a result of a previous conference scheduling or of the first participant dialing in to a specified URI, the Application Server must take care of appropriately creating a media conference on the Media Server. It does so by sending an explicit request to the Media Server. This can be by means of an MS control channel message. This request may contain detailed information upon the desired settings and policies for the conference (e.g., the media to involve, the mixing configuration for them, the relevant identifiers, etc.). The Media Server validates such a request and takes care of allocating the needed resources to set up the media conference.

Application Servers may use mechanisms other than sending requests over the control channel to establish conferences on a Media Server, and then subsequently use the control channel to control the conference. Examples of other mechanisms to create a conference include using the Request-URI mechanism of [RFC4240] or the procedures defined in [RFC4579].

Once done, the MS informs the Application Server about the result of the request. Each conference will be referred to by a specific identifier, which both the Application Server and the Media Server will include in subsequent transactions related to the same conference (e.g., to modify the settings of an extant conference).

6.2. Adding a Participant to a Conference

As stated before, an Application Server uses SIP 3PCC to establish media sessions from SIP user agents to a Media Server. The URI that the AS uses in the INVITE to the MS may be one associated with the conference on the MS. More likely however, the media sessions are first established to the Media Server using a URI for the Media Server and then subsequently joined to the conference using the MS

Control Protocol. This allows IVR dialogs to be performed prior to joining the conference.

The AS as a 3PCC correlates the media session negotiation between the UA and the MS, in order to appropriately establish all the needed media streams based on the conference policies.

6.3. Media Controls

The XCON Common Data Model [XCON-DM] currently defines some basic media-related controls, which conference-aware participants can take advantage of in several ways, e.g., by means of an XCON conference control protocol or IVR dialogs. These controls include the possibility to modify the participants' own volume for audio in the conference, configure the desired layout for incoming video streams, mute/unmute oneself, and pause/unpause one's own video stream. Such controls are exploited by conference-aware participants through the use of dedicated conference control protocol requests to the Application Server. The Application Server takes care of validating such requests and translates them into the Media Server Control Protocol, before forwarding them over the MS Control Channel to the MS. According to the directives provided by the Application Server, the Media Server manipulates the involved media streams accordingly.

                  +------------+                  +------------+
                  |            | 'Include audio   |            |
                  | Application|  sent by user X  |   Media    |
                  |   Server   |  in conf Y mix'  |   Server   |
                  |  (Focus)   |----------------->|  (Mixer)   |
                  |            |   (MS CtrlChn)   |            |
                  +------^-----+                  +------------+
                         |                          ..
                         |                       ...
                         | 'Unmute me'        ... RTP
                         |   (XCON)        ...
                         |              ...
                         |           ...
                  +-----------+   ...
                  |Participant|...
                  +-----------+

Figure 5: Conferencing Example: Unmuting A Participant

The Media Server may need to inform the AS of events like in-band DTMF tones during the conference.

6.4. Floor Control

The XCON framework introduces "floor control" functionality as an enhancement upon [RFC4575]. Floor control is a means to manage joint or exclusive access to shared resources in a (multiparty) conferencing environment. Floor control is not a mandatory mechanism for a conferencing system implementation, but it provides advanced media input control features for conference-aware participants. Such a mechanism allows for coordinated and moderated access to any set of resources provided by the conferencing system. To do so, a so-called floor is associated to a set of resources, thus representing for participants the right to access and manipulate the related resources themselves. In order to take advantage of the floor control functionality, a specific protocol, the Binary Floor Control Protocol, has been specified [RFC4582]. [RFC4583] provides a way for SIP UAs to set up a BFCP connection towards the Floor Control Server and exploit floor control by means of a Connection-Oriented Media (COMEDIA) [RFC4145] negotiation.

In the context of the AS-MS interaction, floor control constitutes a further means to control participants' media streams. A typical example is a floor associated with the right to access the shared audio channel in a conference. A participant who is granted such a floor is granted by the conferencing system the right to talk, which means that its audio frames are included by the MS in the overall audio conference mix. Similarly, when the floor is revoked, the participant is muted in the conference, and its audio is excluded from the final mix.

The BFCP defines a Floor Control Server (FCS) and the floor chair. It is clear that the floor chair making decisions about floor requests is part of the application logic. This implies that when the role of floor chair in a conference is automated, it will normally be part of the AS.

The example makes it clear that there can be a direct or indirect interaction between the Floor Control Server and the Media Server, in order to correctly bind each floor to its related set of media resources. Besides, a similar interaction is needed between the Floor Control Server and the Application Server as well, since the latter must be aware of all the associations between floors and resources, in order to opportunely orchestrate the related bindings with the element responsible for such resources (e.g., the Media Server when talking about audio and/or video streams) and the operations upon them (e.g., mute/unmute a participant in a conference). For this reason, the Floor Control Server can be co- located with either the Media Server or the Application Server, as long as both elements are allowed to interact with the Floor Control Server by means of some kind of protocol.

In the following text, both the approaches will be described in order to better explain the interactions between the involved components in both the topologies.

When the AS and the FCS are co-located, the scenario is quite straightforward. In fact, it can be considered as a variation of the case depicted in Figure 5. The only relevant difference is that in this case the action the AS commands on the control channel is triggered by a change in the floor control status instead of a specific control requested by a participant himself. The sequence diagram in Figure 6 describes the interaction between the involved parties in a typical scenario. It assumes that a BFCP connection between the UA and the FCS (which we assume is co-located with the AS) has already been negotiated and established, and that the UA has been made aware of all the relevant identifiers and floors-resources- associations (e.g., by means of [RFC4583]). It also assumes that the AS has previously configured the media mixing on the MS using the MS control channel. Every frame the UA might be sending on the related media stream is currently being dropped by the MS, since the UA still isn't authorized to use the resource. For a SIP UA, this state could be consequent to a 'sendonly' field associated to the media stream in a re-INVITE originated by the MS. It is worth pointing out that the AS has to make sure that no user media control mechanisms, such as mentioned in the previous sub-section, can override the floor control.

     UA                                   AS                         MS
     (Floor Participant)                 (FCS)
     |                                     |                          |
     |<===================== One-way RTP stream ======================|
     |                                     |                          |
     | FloorRequest(BFCP)                  |                          |
     |------------------------------------>|                          |
     |                                     |                          |
     |   FloorRequestStatus[PENDING](BFCP) |                          |
     |<------------------------------------|                          |
     |                                     |--+ apply                 |
     |                                     |  | policies              |
     |                                     |<-+ to request            |
     |                                     |                          |
     |  FloorRequestStatus[ACCEPTED](BFCP) |                          |
     |<------------------------------------|                          |
     |                                     |                          |
     .                                     .                          .
     .                                     .                          .
     |                                     |                          |
     |   FloorRequestStatus[GRANTED](BFCP) |                          |
     |<------------------------------------|                          |
     |                                     | 'Unmute UA' (CtrlChn)    |
     |                                     |------------------------->|
     |                                     |                          |
     |<==================== Bidirectional RTP stream ================>|
     |                                     |                          |
     .                                     .                          .
     .                                     .                          .

Figure 6: Conferencing Example: Floor Control Call Flow

A UA, which also acts as a floor participant, sends a "FloorRequest" to the floor control server (FCS, which is co-located with the AS), stating his will to be granted the floor associated with the audio stream in the conference. The AS answers the UA with a "FloorRequestStatus" message with a PENDING status, meaning that a decision on the request has not been made yet. The AS, according to the BFCP policies for this conference, makes a decision on the request, i.e., accepting it. Note that this decision might be relayed to another participant in case he has previously been assigned as chair of the floor. Assuming the request has been accepted, the AS notifies the UA about the decision with a new "FloorRequestStatus", this time with an ACCEPTED status in it. The ACCEPTED status of course only means that the request has been accepted, which doesn't mean the floor has been granted yet. Once the queue management in the FCS, according to the specified algorithms for scheduling, states that the floor request previously made by the UA can be granted, the AS sends a new "FloorRequestStatus" to the UA with a GRANTED status, and takes care of unmuting the participant in the conference by sending a directive to the MS through the control channel. Once the UA receives the notification stating his request has been granted, he can start sending its media, aware of the fact that now his media stream won't be dropped by the MS. In case the session has been previously updated with a 'sendonly' associated to the media stream, the MS must originate a further re-INVITE stating that the media stream flow is now bidirectional ('sendrecv').

As mentioned before, this scenario envisages an automated floor chair role, where it's the AS, according to some policies, which makes decisions on floor requests. The case of a chair role performed by a real person is exactly the same, with the difference that the incoming request is not directly handled by the AS according to its policies, but it is instead forwarded to the floor control participant that the chair UA is exploiting. The decision on the request is then communicated by the chair UA to the AS-FCS by means of a 'ChairAction' message.

The rest of this section will instead explore the other scenario, which assumes that the interaction between AS-FCS happens through the MS control channel. This scenario is compliant with the H.248.19 document related to conferencing in 3GPP. The following sequence diagram describes the interaction between the involved parties in the same use-case scenario that has been explored for the previous topology: consequently, the diagram makes exactly the same assumptions that have been made for the previously described scenario. This means that the scenario again assumes that a BFCP connection between the UA and the FCS has already been negotiated and established, and that the UA has been made aware of all the relevant identifiers and floors-resources-associations. It also assumes that the AS has previously configured the media mixing on the MS using the MS control channel. This time it includes identifying the BFCP- moderated resources, establishing basic policies and instructions about chair identifiers for each resource, and subscribing to events of interest, because the FCS is not co-located with the AS anymore. Additionally, a BFCP session has been established between the AS (which in this scenario acts as a floor chair) and the FCS (MS). Every frame the UA might be sending on the related media stream is currently being dropped by the MS, since the UA still isn't authorized to use the resource. For a SIP UA, this state could be consequent to a 'sendonly' field associated to the media stream in a re-INVITE originated by the MS. Again, it is worth pointing out that the AS has to make sure that no user media control mechanisms, such as mentioned in the previous sub-section, can override the floor control.

     UA                          AS                                  MS
     (Floor Participant)   (Floor Chair)                          (FCS)
     |                           |                                    |
     |<===================== One-way RTP stream ======================|
     |                           |                                    |
     | FloorRequest(BFCP)        |                                    |
     |--------------------------------------------------------------->|
     |                           |                                    |
     |                           |  FloorRequestStatus[PENDING](BFCP) |
     |<---------------------------------------------------------------|
     |                           |  FloorRequestStatus[PENDING](BFCP) |
     |                           |<-----------------------------------|
     |                           |                                    |
     |                           | ChairAction[ACCEPTED] (BFCP)       |
     |                           |----------------------------------->|
     |                           |       ChairActionAck (BFCP)        |
     |                           |<-----------------------------------|
     |                           |                                    |
     |                           | FloorRequestStatus[ACCEPTED](BFCP) |
     |<---------------------------------------------------------------|
     |                           |                                    |
     .                           .                                    .
     .                           .                                    .
     |                           |                                    |
     |                           |  FloorRequestStatus[GRANTED](BFCP) |
     |<---------------------------------------------------------------|
     |                           | 'Floor has been granted' (CtrlChn) |
     |                           |<-----------------------------------|
     |                           |                                    |
     |<==================== Bidirectional RTP stream ================>|
     |                           |                                    |
     .                           .                                    .
     .                           .                                    .

Figure 7: Conferencing Example: Floor Control Call Flow

A UA, which also acts as a floor participant, sends a "FloorRequest" to the floor control server (FCS, which is co-located with the MS), stating his will to be granted the floor associated with the audio stream in the conference. The MS answers the UA with a "FloorRequestStatus" message with a PENDING status, meaning that a decision on the request has not been made yet. It then notifies the AS, which in this example handles the floor chair role, about the new request by forwarding there the received request. The AS, according to the BFCP policies for this conference, makes a decision on the request, i.e., accepting it. It informs the MS about its decision through a BFCP "ChairAction" message. The MS then acknowledges the 'ChairAction' message and then notifies the UA about the decision with a new "FloorRequestStatus", this time with an ACCEPTED status in it. The ACCEPTED status of course only means that the request has been accepted, which doesn't mean the floor has been granted yet. Once the queue management in the MS, according to the specified algorithms for scheduling, states that the floor request previously made by the UA can be granted, the MS sends a new "FloorRequestStatus" to the UA with a GRANTED status, and takes care of unmuting the participant in the conference. Once the UA receives the notification stating his request has been granted, he can start sending its media, aware of the fact that now his media stream won't be dropped by the MS. In case the session has been previously updated with a 'sendonly' associated to the media stream, the MS must originate a further re-INVITE stating that the media stream flow is now bidirectional ('sendrecv').

This scenario envisages an automated floor chair role, where it's the AS, according to some policies, which makes decisions on floor requests. Again, the case of a chair role performed by a real person is exactly the same, with the difference that the incoming request is not forwarded to the AS but to the floor control participant that the chair UA is exploiting. The decision on the request is communicated by means of a 'ChairAction' message in the same way.

Another typical scenario is a BFCP-moderated conference with no chair to manage floor requests. In such a scenario, the MS has to take care of incoming requests according to some predefined policies, e.g., always accepting new requests. In this case, no decisions are required by external entities, since all are instantly decided by means of policies in the MS.

As stated before, the case of the FCS co-located with the AS is much simpler to understand and exploit. When the AS has full control upon the FCS, including its queue management, the AS directly instructs the MS according to the floor status changes, e.g., by instructing the MS through the control channel to unmute a participant who has been granted the floor associated to the audio media stream.

7. Security Considerations

This document describes the architectural framework to be used for Media Server control. Its focus is the interactions between Application Servers and Media Servers. User agents interact with Application Servers by means of signaling protocols such as SIP. These interactions are beyond the scope of this document. Application Servers are responsible for utilizing the security mechanisms of their signaling protocols, combined with application- specific policy, to ensure they grant service only to authorized users. Media interactions between user agents and Media Servers are also outside the scope of this document. Those interactions are at the behest of Application Servers, which must ensure that appropriate security mechanisms are used. For example, if the MS is acting as the FCS, then the BFCP connection between the user agent and the MS is established to the MS by the AS using SIP and the SDP mechanisms described in [RFC4583]. BFCP [RFC4582] strongly imposes the use of TLS for BFCP.

Media Servers are valuable network resources and need to be protected against unauthorized access. Application Servers use SIP and related standards both to establish control channels to Media Servers and to establish media sessions, including BFCP sessions, between an MS and end users. Media servers use the security mechanisms of SIP to authenticate requests from Application servers and to ensure the integrity of those requests. Leveraging the security mechanisms of SIP ensures that only authorized Application Servers are allowed to establish sessions to an MS and to access MS resources through those sessions.

Control channels between an AS and MS carry the MS control protocol, which affects both the service seen by end users and the resources used on a Media Server. TLS [RFC5246] must be implemented as the transport-level security mechanism for control channels to guarantee the integrity of MS control interactions.

The resources of an MS can be shared by more than one AS. Media Servers must prevent one AS from accessing and manipulating the resources that have been assigned to another AS. This may be achieved by an MS associating ownership of a resource to the AS that originally allocates it, and then insuring that future requests involving that resource correlate to the AS that owns and is responsible for it.

8. Acknowledgments

The authors would like to thank Spencer Dawkins for detailed reviews and comments, Gary Munson for suggestions, and Xiao Wang for review and feedback.

9. Contributors

This document is a product of the Media Control Architecture Design Team. In addition to the editor, the following individuals constituted the design team and made substantial textual contributions to this document:

      Chris Boulton: cboulton@ubiquity.net
      
      Martin Dolly: mdolly@att.com
      
      Roni Even: roni.even@polycom.co.il
      
      Lorenzo Miniero: lorenzo.miniero@unina.it
      
      Adnan Saleem: Adnan.Saleem@radisys.com

10. Informative References

   [RFC0793]  Postel, J., "Transmission Control Protocol", STD 7,
              RFC 793, September 1981.
   
   [RFC2976]  Donovan, S., "The SIP INFO Method", RFC 2976,
              October 2000.
   
   [RFC3261]  Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston,
              A., Peterson, J., Sparks, R., Handley, M., and E.
              Schooler, "SIP: Session Initiation Protocol", RFC 3261,
              June 2002.
   
   [RFC3263]  Rosenberg, J. and H. Schulzrinne, "Session Initiation
              Protocol (SIP): Locating SIP Servers", RFC 3263,
              June 2002.
   
   [RFC3264]  Rosenberg, J. and H. Schulzrinne, "An Offer/Answer Model
              with Session Description Protocol (SDP)", RFC 3264,
              June 2002.
   
   [RFC3550]  Schulzrinne, H., Casner, S., Frederick, R., and V.
              Jacobson, "RTP: A Transport Protocol for Real-Time
              Applications", STD 64, RFC 3550, July 2003.
   
   [RFC3725]  Rosenberg, J., Peterson, J., Schulzrinne, H., and G.
              Camarillo, "Best Current Practices for Third Party Call
              Control (3pcc) in the Session Initiation Protocol (SIP)",
              BCP 85, RFC 3725, April 2004.
   
   [RFC3840]  Rosenberg, J., Schulzrinne, H., and P. Kyzivat,
              "Indicating User Agent Capabilities in the Session
              Initiation Protocol (SIP)", RFC 3840, August 2004.
   
   [RFC4145]  Yon, D. and G. Camarillo, "TCP-Based Media Transport in
              the Session Description Protocol (SDP)", RFC 4145,
              September 2005.
   
   [RFC4240]  Burger, E., Van Dyke, J., and A. Spitzer, "Basic Network
              Media Services with SIP", RFC 4240, December 2005.
   
   [RFC4353]  Rosenberg, J., "A Framework for Conferencing with the
              Session Initiation Protocol (SIP)", RFC 4353,
              February 2006.
   
   [RFC4474]  Peterson, J. and C. Jennings, "Enhancements for
              Authenticated Identity Management in the Session
              Initiation Protocol (SIP)", RFC 4474, August 2006.
   
   [RFC4566]  Handley, M., Jacobson, V., and C. Perkins, "SDP: Session
              Description Protocol", RFC 4566, July 2006.
   
   [RFC4575]  Rosenberg, J., Schulzrinne, H., and O. Levin, "A Session
              Initiation Protocol (SIP) Event Package for Conference
              State", RFC 4575, August 2006.
   
   [RFC4579]  Johnston, A. and O. Levin, "Session Initiation Protocol
              (SIP) Call Control - Conferencing for User Agents",
              BCP 119, RFC 4579, August 2006.
   
   [RFC4582]  Camarillo, G., Ott, J., and K. Drage, "The Binary Floor
              Control Protocol (BFCP)", RFC 4582, November 2006.
   
   [RFC4583]  Camarillo, G., "Session Description Protocol (SDP) Format
              for Binary Floor Control Protocol (BFCP) Streams",
              RFC 4583, November 2006.
   
   [RFC4585]  Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey,
              "Extended RTP Profile for Real-time Transport Control
              Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585,
              July 2006.
   
   [RFC4960]  Stewart, R., "Stream Control Transmission Protocol",
              RFC 4960, September 2007.
   
   [RFC5167]  Dolly, M. and R. Even, "Media Server Control Protocol
              Requirements", RFC 5167, March 2008.
   
   [RFC5239]  Barnes, M., Boulton, C., and O. Levin, "A Framework for
              Centralized Conferencing", RFC 5239, June 2008.
   
   [RFC5246]  Dierks, T. and E. Rescorla, "The Transport Layer Security
              (TLS) Protocol Version 1.2", RFC 5246, August 2008.

[SIP-CTRL-FW]

              Boulton, C., Melanchuk, T., and S. McGlashan, "Media
              Control Channel Framework", Work in Progress,
              February 2009.

[W3C.REC-voicexml20-20040316]

              Carter, J., Tryphonas, S., Danielsen, P., Burnett, D.,
              Rehor, K., McGlashan, S., Ferrans, J., Porter, B., Lucas,
              B., and A. Hunt, "Voice Extensible Markup Language
              (VoiceXML) Version 2.0", World Wide Web Consortium
              Recommendation REC-voicexml20-20040316, March 2004,
              <http://www.w3.org/TR/2004/REC-voicexml20-20040316>.

[W3C.REC-xml-20060816]

              Sperberg-McQueen, C., Paoli, J., Bray, T., Maler, E., and
              F. Yergeau, "Extensible Markup Language (XML) 1.0 (Fourth
              Edition)", World Wide Web Consortium Recommendation REC-
              xml-20060816, August 2006,
              <http://www.w3.org/TR/2006/REC-xml-20060816>.
   
   [XCON-DM]  Novo, O., Camarillo, G., Morgan, D., and J. Urpalainen,
              "Conference Information Data Model for Centralized
              Conferencing (XCON)", Work in Progress, April 2009.

Author's Address

Tim Melanchuk (editor)
Rain Willow Communications

   EMail: tim.melanchuk@gmail.com