Skip to main content

Media Server Control
charter-ietf-mediactrl-04

Document Charter Media Server Control WG (mediactrl)
Title Media Server Control
Last updated 2007-02-22
State Approved
WG State Concluded
IESG Responsible AD Richard Barnes
Charter edit AD (None)
Send notices to (None)

charter-ietf-mediactrl-04

Real-time multi-media applications often need the services of media
processing elements. It is true that modern endpoints are capable of
media processing. However, the physics of some media processing
applications dictate that it is much more efficient for the media
processing to occur at a centralized location. By media processing, we
mean media mixing, recording and playing media, and interacting with a
user in the audio or video domains. The commercial market calls these
media processing network elements "media servers."

Some services achieve significant efficiencies when a central node
performs media processing. Because of these efficiencies, media
servers are widely used for conference mixing, multimedia messaging,
content rendering, and speech, voice, key press, and other audio and
video input and output user interface modalities. Given the wide
acceptance of the media server, we need a standard way to control them.

Since the media server is a centralized component, the work group will
not investigate distributed media processing algorithms or control
protocols.

A media server contains media processing components that are able to
manipulate RTP streams. Typical processing includes mixing multiple
streams, transcoding a stream (e.g., from G.711 to MS-GSM), storing or
retrieving a stream (e.g., from RTP to HTTP), detecting tones (e.g.,
DTMF), converting text to speech, and performing speech recognition.
Note that an MRCPv2 server may offer the low-level processing for the
last two services, where the media server is a client to the MRCPv2
server. Also note it is common to call the package of detecting user
input, recording media, and playing media "Interactive Voice
Response," or IVR. Media services offered by the media server are
addressed using SIP mechanisms, such as described in RFC 4240. Media
servers commonly have a built-in VoiceXML interpreter. VoiceXML
describes the elements of the user interaction, and is a proven model
for separating application logic (which run on the clients of the
media server) from the user interface (which the media server
renders). Note this is a fundamentally different interaction model from
MRCPv2, where media processing engines offer raw, low-level speech
services.

The work group will examine protocol extensions between media servers
and their clients. However, modifying existing standard protocols,
such as VoiceXML or SIP towards clients or MRCPv2 towards servers, is
not in the work group's charter. The model of interest to this group
is where the endpoint solely plays audio or video, transmits audio or
video towards the server, and possibly transmits key press information
towards the server. Alternate architectures, where the endpoint
executes user interface commands, is outside the scope of the
work group. For example, WIDEX/BEEP, with its distributed user
interface description, is not in scope.

The only model of user interface processing the work group will
consider is where the media server performs all of the media
processing. A caveat here is the media server, in interpreting a
VoiceXML page, may make requests to a server for speech services.
However, to the media server client and the media end point, the
single point of signaling and media interaction is the media server.

Any protocol developed by this group will meet the requirements for
Internet deployment. This includes addressing Internet security,
privacy, congestion control (or at least congestion safe), operational
and manageability considerations, and scale. The protocol will not
assume a private administrative domain. There is broad market
acceptance of the stimulus/markup application design model for the
application server - media server protocol interface. Thus this work
group will focus on the use of SIP and XML for the protocol suite.

The work product of this group includes the following:

  1. A requirements document. This document will identify and enumerate
    requirements for a suite of media server control protocols. Given that
    one of the common media server clients is a conference application
    server, we will consider the application server - media server
    requirements developed by the XCON work group. Likewise, we will
    consider media server control requirements from other standards
    groups, such as 3GPP SA2 and CT1.

  2. A framework document. This document will describe the different
    network elements, their interrelationship, and the broad set of
    message flows between them.

  3. A protocol suite describing the embodiment of the framework
    document. There may be separate protocol PDU's for audio conference
    control, video conference control, interactive audio (voice) response,
    and interactive video (multimedia) response. The separation and
    negotiation of different PDU's is a working group topic. However,
    there will be one and only one (class) of PDU's defined by the work
    group.

  4. Means for locating, and possibly establishing sessions to, media
    servers with appropriate resources at the request of clients. By
    appropriate, we mean the characteristics of a given media server
    required or desired for handling a given request. The expectation is
    such a means would build upon existing SIP, SNMP, and other protocol
    facilities. Such a means may or may not be an integral part of the
    item 3 deliverables above. This deliverable is an operational protocol
    that may rely on management protocols such as SNMP. We are neither
    creating a new management protocol nor a new provisioning protocol.

Given the above-mentioned conferencing example, the work of this group
is of interest to the XCON work group, as this protocol will describe
the "Protocol used between the conference controller and the mixer
(s)." Thus we expect to work closely with XCON. The protocol suite
also is a possible embodiment of the ISC/Mr interface from the 3GPP
IMS architecture. Thus we expect to gather requirements from, 3GPP,
notably SA2, CT1, and CT4. ATIS and ETSI TISPAN have considered a
functional element known as a media resource broker. The media
resource broker provides the functionality described by deliverable
#4, above. Thus we expect to gather requirements from ATIS and ETSI
TISPAN. The Java Community Process has chartered work on a Java Media
Server Control (JMSC) API, known as JSR 309. We expect to gather
requirements from JCP, as well.

Because of the vast experience with conferencing protocols and
payloads, we expect considerable interaction with AVT and MMUSIC. If
the work group requires extensions to SIP, the work group will forward
those extensions to the SIP work group for consideration and
refinement.