RFC 5104






Network Working Group                                          S. Wenger
Request for Comments: 5104                                    U. Chandra
Category: Standards Track                                          Nokia
                                                           M. Westerlund
                                                               B. Burman
                                                                Ericsson
                                                           February 2008


                     Codec Control Messages in the
             RTP Audio-Visual Profile with Feedback (AVPF)

Status of This Memo



   This document specifies an Internet standards track protocol for the
   Internet community, and requests discussion and suggestions for
   improvements.  Please refer to the current edition of the "Internet
   Official Protocol Standards" (STD 1) for the standardization state
   and status of this protocol.  Distribution of this memo is unlimited.

Abstract



   This document specifies a few extensions to the messages defined in
   the Audio-Visual Profile with Feedback (AVPF).  They are helpful
   primarily in conversational multimedia scenarios where centralized
   multipoint functionalities are in use.  However, some are also usable
   in smaller multicast environments and point-to-point calls.

   The extensions discussed are messages related to the ITU-T Rec. H.271
   Video Back Channel, Full Intra Request, Temporary Maximum Media
   Stream Bit Rate, and Temporal-Spatial Trade-off.




















Wenger, et al.              Standards Track                     [Page 1]

RFC 5104             Codec Control Messages in AVPF        February 2008


Table of Contents



   1. Introduction ....................................................4
   2. Definitions .....................................................5
      2.1. Glossary ...................................................5
      2.2. Terminology ................................................5
      2.3. Topologies .................................................8
   3. Motivation ......................................................8
      3.1. Use Cases ..................................................9
      3.2. Using the Media Path ......................................11
      3.3. Using AVPF ................................................11
           3.3.1. Reliability ........................................12
      3.4. Multicast .................................................12
      3.5. Feedback Messages .........................................12
           3.5.1. Full Intra Request Command .........................12
                  3.5.1.1. Reliability ...............................13
           3.5.2. Temporal-Spatial Trade-off Request and
                  Notification .......................................14
                  3.5.2.1. Point-to-Point ............................15
                  3.5.2.2. Point-to-Multipoint Using
                           Multicast or Translators ..................15
                  3.5.2.3. Point-to-Multipoint Using RTP Mixer .......15
                  3.5.2.4. Reliability ...............................16
           3.5.3. H.271 Video Back Channel Message ...................16
                  3.5.3.1. Reliability ...............................19
           3.5.4. Temporary Maximum Media Stream Bit Rate
                  Request and Notification ...........................19
                  3.5.4.1. Behavior for Media Receivers Using TMMBR ..21
                  3.5.4.2. Algorithm for Establishing Current
                           Limitations ...............................23
                  3.5.4.3. Use of TMMBR in a Mixer-Based
                           Multipoint Operation ......................29
                  3.5.4.4. Use of TMMBR in Point-to-Multipoint Using
                           Multicast or Translators ..................30
                  3.5.4.5. Use of TMMBR in Point-to-Point Operation ..31
                  3.5.4.6. Reliability ...............................31
   4. RTCP Receiver Report Extensions ................................32
      4.1. Design Principles of the Extension Mechanism ..............32
      4.2. Transport Layer Feedback Messages .........................33
           4.2.1. Temporary Maximum Media Stream Bit Rate
                  Request (TMMBR) ....................................34
                  4.2.1.1. Message Format ............................34
                  4.2.1.2. Semantics .................................35
                  4.2.1.3. Timing Rules ..............................39
                  4.2.1.4. Handling in Translators and Mixers ........39
           4.2.2. Temporary Maximum Media Stream Bit Rate
                  Notification (TMMBN) ...............................39
                  4.2.2.1. Message Format ............................39



Wenger, et al.              Standards Track                     [Page 2]

RFC 5104             Codec Control Messages in AVPF        February 2008


                  4.2.2.2. Semantics .................................40
                  4.2.2.3. Timing Rules ..............................41
                  4.2.2.4. Handling by Translators and Mixers ........41
      4.3. Payload-Specific Feedback Messages ........................41
           4.3.1. Full Intra Request (FIR) ...........................42
                  4.3.1.1. Message Format ............................42
                  4.3.1.2. Semantics .................................43
                  4.3.1.3. Timing Rules ..............................44
                  4.3.1.4. Handling of FIR Message in Mixers and
                           Translators ...............................44
                  4.3.1.5. Remarks ...................................44
           4.3.2. Temporal-Spatial Trade-off Request (TSTR) ..........45
                  4.3.2.1. Message Format ............................46
                  4.3.2.2. Semantics .................................46
                  4.3.2.3. Timing Rules ..............................47
                  4.3.2.4. Handling of Message in Mixers and
                           Translators ...............................47
                  4.3.2.5. Remarks ...................................47
           4.3.3. Temporal-Spatial Trade-off Notification (TSTN) .....48
                  4.3.3.1. Message Format ............................48
                  4.3.3.2. Semantics .................................49
                  4.3.3.3. Timing Rules ..............................49
                  4.3.3.4. Handling of TSTN in Mixers and
                           Translators ...............................49
                  4.3.3.5. Remarks ...................................49
           4.3.4. H.271 Video Back Channel Message (VBCM) ............50
                  4.3.4.1. Message Format ............................50
                  4.3.4.2. Semantics .................................51
                  4.3.4.3. Timing Rules ..............................52
                  4.3.4.4. Handling of Message in Mixers or
                           Translators ...............................52
                  4.3.4.5. Remarks ...................................52
   5. Congestion Control .............................................52
   6. Security Considerations ........................................53
   7. SDP Definitions ................................................54
      7.1. Extension of the rtcp-fb Attribute ........................54
      7.2. Offer-Answer ..............................................55
      7.3. Examples ..................................................56
   8. IANA Considerations ............................................58
   9. Contributors ...................................................60
   10. Acknowledgements ..............................................60
   11. References ....................................................60
      11.1. Normative References .....................................60
      11.2. Informative References ...................................61







Wenger, et al.              Standards Track                     [Page 3]

RFC 5104             Codec Control Messages in AVPF        February 2008


1.  Introduction



   When the Audio-Visual Profile with Feedback (AVPF) [RFC4585] was
   developed, the main emphasis lay in the efficient support of point-
   to-point and small multipoint scenarios without centralized
   multipoint control.  However, in practice, many small multipoint
   conferences operate utilizing devices known as Multipoint Control
   Units (MCUs).  Long-standing experience of the conversational video
   conferencing industry suggests that there is a need for a few
   additional feedback messages, to support centralized multipoint
   conferencing efficiently.  Some of the messages have applications
   beyond centralized multipoint, and this is indicated in the
   description of the message.  This is especially true for the message
   intended to carry ITU-T Rec. H.271 [H.271] bit strings for Video Back
   Channel messages.

   In Real-time Transport Protocol (RTP) [RFC3550] terminology, MCUs
   comprise mixers and translators.  Most MCUs also include signaling
   support.  During the development of this memo, it was noticed that
   there is considerable confusion in the community related to the use
   of terms such as mixer, translator, and MCU.  In response to these
   concerns, a number of topologies have been identified that are of
   practical relevance to the industry, but are not documented in
   sufficient detail in [RFC3550].  These topologies are documented in
   [RFC5117], and understanding this memo requires previous or parallel
   study of [RFC5117].

   Some of the messages defined here are forward only, in that they do
   not require an explicit notification to the message emitter that they
   have been received and/or indicating the message receiver's actions.
   Other messages require a response, leading to a two-way communication
   model that one could view as useful for control purposes.  However,
   it is not the intention of this memo to open up RTP Control Protocol
   (RTCP) to a generalized control protocol.  All mentioned messages
   have relatively strict real-time constraints, in the sense that their
   value diminishes with increased delay.  This makes the use of more
   traditional control protocol means, such as Session Initiation
   Protocol (SIP) [RFC3261], undesirable when used for the same purpose.
   That is why this solution is recommended instead of "XML Schema for
   Media Control" [XML-MC], which uses SIP Info to transfer XML messages
   with similar semantics to what are defined in this memo.
   Furthermore, all messages are of a very simple format that can be
   easily processed by an RTP/RTCP sender/receiver.  Finally, and most
   importantly, all messages relate only to the RTP stream with which
   they are associated, and not to any other property of a communication
   system.  In particular, none of them relate to the properties of the
   access links traversed by the session.




Wenger, et al.              Standards Track                     [Page 4]

RFC 5104             Codec Control Messages in AVPF        February 2008


2.  Definitions



2.1.  Glossary



   AIMD   - Additive Increase Multiplicative Decrease
   AVPF   - The extended RTP profile for RTCP-based feedback
   FCI    - Feedback Control Information [RFC4585]
   FEC    - Forward Error Correction
   FIR    - Full Intra Request
   MCU    - Multipoint Control Unit
   MPEG   - Moving Picture Experts Group
   PLI    - Picture Loss Indication
   PR     - Packet rate
   QP     - Quantizer Parameter
   RTT    - Round trip time
   SSRC   - Synchronization Source
   TMMBN  - Temporary Maximum Media Stream Bit Rate Notification
   TMMBR  - Temporary Maximum Media Stream Bit Rate Request
   TSTN   - Temporal-Spatial Trade-off Notification
   TSTR   - Temporal-Spatial Trade-off Request
   VBCM   - Video Back Channel Message

2.2.  Terminology



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

      Message:
          An RTCP feedback message [RFC4585] defined by this
          specification, of one of the following types:

          Request:
              Message that requires acknowledgement

          Command:
              Message that forces the receiver to an action

          Indication:
              Message that reports a situation

          Notification:
              Message that provides a notification that an event has
              occurred.  Notifications are commonly generated in
              response to a Request.

   Note that, with the exception of "Notification", this terminology is
   in alignment with ITU-T Rec. H.245 [H245].



Wenger, et al.              Standards Track                     [Page 5]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Decoder Refresh Point:
          A bit string, packetized in one or more RTP packets, that
          completely resets the decoder to a known state.

          Examples for "hard" decoder refresh points are Intra pictures
          in H.261, H.263, MPEG-1, MPEG-2, and MPEG-4 part 2, and
          Instantaneous Decoder Refresh (IDR) pictures in H.264.
          "Gradual" decoder refresh points may also be used; see for
          example [AVC].  While both "hard" and "gradual" decoder
          refresh points are acceptable in the scope of this
          specification, in most cases the user experience will benefit
          from using a "hard" decoder refresh point.

          A decoder refresh point also contains all header information
          above the picture layer (or equivalent, depending on the video
          compression standard) that is conveyed in-band.  In H.264, for
          example, a decoder refresh point contains parameter set
          Network Adaptation Layer (NAL) units that generate parameter
          sets necessary for the decoding of the following slice/data
          partition NAL units (and that are not conveyed out of band).

   Decoding:
          The operation of reconstructing the media stream.

   Rendering:
          The operation of presenting (parts of) the reconstructed media
          stream to the user.

   Stream thinning:
          The operation of removing some of the packets from a media
          stream.  Stream thinning, preferably, is media-aware, implying
          that media packets are removed in the order of increasing
          relevance to the reproductive quality.  However, even when
          employing media-aware stream thinning, most media streams
          quickly lose quality when subjected to increasing levels of
          thinning.  Media-unaware stream thinning leads to even worse
          quality degradation.  In contrast to transcoding, stream
          thinning is typically seen as a computationally lightweight
          operation.

   Media:
          Often used (sometimes in conjunction with terms like bit rate,
          stream, sender, etc.) to identify the content of the forward
          RTP packet stream (carrying the codec data), to which the
          codec control message applies.






Wenger, et al.              Standards Track                     [Page 6]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Media Stream:
          The stream of RTP packets labeled with a single
          Synchronization Source (SSRC) carrying the media (and also in
          some cases repair information such as retransmission or
          Forward Error Correction (FEC) information).

   Total media bit rate:
          The total bits per second transferred in a media stream,
          measured at an observer-selected protocol layer and averaged
          over a reasonable timescale, the length of which depends on
          the application.  In general, a media sender and a media
          receiver will observe different total media bit rates for the
          same stream, first because they may have selected different
          reference protocol layers, and second, because of changes in
          per-packet overhead along the transmission path.  The goal
          with bit rate averaging is to be able to ignore any burstiness
          on very short timescales (e.g., below 100 ms) introduced by
          scheduling or link layer packetization effects.

   Maximum total media bit rate:
          The upper limit on total media bit rate for a given media
          stream at a particular receiver and for its selected protocol
          layer.  Note that this value cannot be measured on the
          received media stream.  Instead, it needs to be calculated or
          determined through other means, such as quality of service
          (QoS) negotiations or local resource limitations.  Also note
          that this value is an average (on a timescale that is
          reasonable for the application) and that it may be different
          from the instantaneous bit rate seen by packets in the media
          stream.

   Overhead:
          All protocol header information required to convey a packet
          with media data from sender to receiver, from the application
          layer down to a pre-defined protocol level (for example, down
          to, and including, the IP header).  Overhead may include, for
          example, IP, UDP, and RTP headers, any layer 2 headers, any
          Contributing Sources (CSRCs), RTP padding, and RTP header
          extensions.  Overhead excludes any RTP payload headers and the
          payload itself.

   Net media bit rate:
          The bit rate carried by a media stream, net of overhead.  That
          is, the bits per second accounted for by encoded media, any
          applicable payload headers, and any directly associated meta
          payload information placed in the RTP packet.  A typical
          example of the latter is redundancy data provided by the use
          of RFC 2198 [RFC2198].  Note that, unlike the total media bit



Wenger, et al.              Standards Track                     [Page 7]

RFC 5104             Codec Control Messages in AVPF        February 2008


          rate, the net media bit rate will have the same value at the
          media sender and at the media receiver unless any mixing or
          translating of the media has occurred.

          For a given observer, the total media bit rate for a media
          stream is equal to the sum of the net media bit rate and the
          per-packet overhead as defined above multiplied by the packet
          rate.

   Feasible region:
          The set of all combinations of packet rate and net media bit
          rate that do not exceed the restrictions in maximum media bit
          rate placed on a given media sender by the Temporary Maximum
          Media Stream Bit Rate Request (TMMBR) messages it has
          received.  The feasible region will change as new TMMBR
          messages are received.

   Bounding set:
          The set of TMMBR tuples, selected from all those received at a
          given media sender, that define the feasible region for that
          media sender.  The media sender uses an algorithm such as that
          in section 3.5.4.2 to determine or iteratively approximate the
          current bounding set, and reports that set back to the media
          receivers in a Temporary Maximum Media Stream Bit Rate
          Notification (TMMBN) message.

2.3.  Topologies



   Please refer to [RFC5117] for an in-depth discussion.  The topologies
   referred to throughout this memo are labeled (consistently with
   [RFC5117]) as follows:

   Topo-Point-to-Point . . . . . Point-to-point communication
   Topo-Multicast  . . . . . . . Multicast communication
   Topo-Translator . . . . . . . Translator based
   Topo-Mixer  . . . . . . . . . Mixer based
   Topo-RTP-switch-MCU . . . . . RTP stream switching MCU
   Topo-RTCP-terminating-MCU . . Mixer but terminating RTCP

3.  Motivation



   This section discusses the motivation and usage of the different
   video and media control messages.  The video control messages have
   been under discussion for a long time, and a requirement document was
   drawn up [Basso].  That document has expired; however, we quote
   relevant sections of it to provide motivation and requirements.





Wenger, et al.              Standards Track                     [Page 8]

RFC 5104             Codec Control Messages in AVPF        February 2008


3.1.  Use Cases



   There are a number of possible usages for the proposed feedback
   messages.  Let us begin by looking through the use cases Basso et al.
   [Basso] proposed.  Some of the use cases have been reformulated and
   comments have been added.

   1. An RTP video mixer composes multiple encoded video sources into a
      single encoded video stream.  Each time a video source is added,
      the RTP mixer needs to request a decoder refresh point from the
      video source, so as to start an uncorrupted prediction chain on
      the spatial area of the mixed picture occupied by the data from
      the new video source.

   2. An RTP video mixer receives multiple encoded RTP video streams
      from conference participants, and dynamically selects one of the
      streams to be included in its output RTP stream.  At the time of a
      bit stream change (determined through means such as voice
      activation or the user interface), the mixer requests a decoder
      refresh point from the remote source, in order to avoid using
      unrelated content as reference data for inter picture prediction.
      After requesting the decoder refresh point, the video mixer stops
      the delivery of the current RTP stream and monitors the RTP stream
      from the new source until it detects data belonging to the decoder
      refresh point.  At that time, the RTP mixer starts forwarding the
      newly selected stream to the receiver(s).

   3. An application needs to signal to the remote encoder that the
      desired trade-off between temporal and spatial resolution has
      changed.  For example, one user may prefer a higher frame rate and
      a lower spatial quality, and another user may prefer the opposite.
      This choice is also highly content dependent.  Many current video
      conferencing systems offer in the user interface a mechanism to
      make this selection, usually in the form of a slider.  The
      mechanism is helpful in point-to-point, centralized multipoint and
      non-centralized multipoint uses.

   4. Use case 4 of the Basso document applies only to Picture Loss
      Indication (PLI) as defined in AVPF [RFC4585] and is not
      reproduced here.



   5. Use case 5 of the Basso document relates to a mechanism known as
      "freeze picture request".  Sending freeze picture requests over a
      non-reliable forward RTCP channel has been identified as
      problematic.  Therefore, no freeze picture request has been
      included in this memo, and the use case discussion is not
      reproduced here.




Wenger, et al.              Standards Track                     [Page 9]

RFC 5104             Codec Control Messages in AVPF        February 2008


   6. A video mixer dynamically selects one of the received video
      streams to be sent out to participants and tries to provide the
      highest bit rate possible to all participants, while minimizing
      stream trans-rating.  One way of achieving this is to set up
      sessions with endpoints using the maximum bit rate accepted by
      each endpoint, and accepted by the call admission method used by
      the mixer.  By means of commands that reduce the maximum media
      stream bit rate below what has been negotiated during session set
      up, the mixer can reduce the maximum bit rate sent by endpoints to
      the lowest of all the accepted bit rates.  As the lowest accepted
      bit rate changes due to endpoints joining and leaving or due to
      network congestion, the mixer can adjust the limits at which
      endpoints can send their streams to match the new value.  The
      mixer then requests a new maximum bit rate, which is equal to or
      less than the maximum bit rate negotiated at session setup for a
      specific media stream, and the remote endpoint can respond with
      the actual bit rate that it can support.

   The picture Basso, et al., draw up covers most applications we
   foresee.  However, we would like to extend the list with two
   additional use cases:

   7. Currently deployed congestion control algorithms (AIMD and TCP
      Friendly Rate Control (TFRC) [RFC3448]) probe for additional
      available capacity as long as there is something to send.  With
      congestion control algorithms using packet loss as the indication
      for congestion, this probing generally results in reduced media
      quality (often to a point where the distortion is large enough to
      make the media unusable), due to packet loss and increased delay.

      In a number of deployment scenarios, especially cellular ones, the
      bottleneck link is often the last hop link.  That cellular link
      also commonly has some type of QoS negotiation enabling the
      cellular device to learn the maximal bit rate available over this
      last hop.  A media receiver behind this link can, in most (if not
      all) cases, calculate at least an upper bound for the bit rate
      available for each media stream it presently receives.  How this
      is done is an implementation detail and not discussed herein.
      Indicating the maximum available bit rate to the transmitting
      party for the various media streams can be beneficial to prevent
      that party from probing for bandwidth for this stream in excess of
      a known hard limit.  For cellular or other mobile devices, the
      known available bit rate for each stream (deduced from the link
      bit rate) can change quickly, due to handover to another
      transmission technology, QoS renegotiation due to congestion, etc.
      To enable minimal disruption of service, quick convergence is
      necessary, and therefore media path signaling is desirable.




Wenger, et al.              Standards Track                    [Page 10]

RFC 5104             Codec Control Messages in AVPF        February 2008


    8. The use of reference picture selection (RPS) as an error
       resilience tool was introduced in 1997 as NEWPRED [NEWPRED], and
       is now widely deployed.  When RPS is in use, simplistically put,
       the receiver can send a feedback message to the sender,
       indicating a reference picture that should be used for future
       prediction.  ([NEWPRED] mentions other forms of feedback as
       well.)  AVPF contains a mechanism for conveying such a message,
       but did not specify for which codec and according to which syntax
       the message should conform.  Recently, the ITU-T finalized Rec.
       H.271, which (among other message types) also includes a feedback
       message.  It is expected that this feedback message will fairly
       quickly enjoy wide support.  Therefore, a mechanism to convey
       feedback messages according to H.271 appears to be desirable.

3.2.  Using the Media Path



   There are two reasons why we use the media path for the codec control
   messages.

   First, systems employing MCUs often separate the control and media
   processing parts.  As these messages are intended for or generated by
   the media part rather than the signaling part of the MCU, having them
   on the media path avoids transmission across interfaces and
   unnecessary control traffic between signaling and processing.  If the
   MCU is physically decomposed, the use of the media path avoids the
   need for media control protocol extensions (e.g., in media gateway
   control (MEGACO) [RFC3525]).

   Secondly, the signaling path quite commonly contains several
   signaling entities, e.g., SIP proxies and application servers.
   Avoiding going through signaling entities avoids delay for several
   reasons.  Proxies have less stringent delay requirements than media
   processing, and due to their complex and more generic nature may
   result in significant processing delay.  The topological locations of
   the signaling entities are also commonly not optimized for minimal
   delay, but rather towards other architectural goals.  Thus, the
   signaling path can be significantly longer in both geographical and
   delay sense.

3.3.  Using AVPF



   The AVPF feedback message framework [RFC4585] provides the
   appropriate framework to implement the new messages.  AVPF implements
   rules controlling the timing of feedback messages to avoid congestion
   through network flooding by RTCP traffic.  We re-use these rules by
   referencing AVPF.





Wenger, et al.              Standards Track                    [Page 11]

RFC 5104             Codec Control Messages in AVPF        February 2008


   The signaling setup for AVPF allows each individual type of function
   to be configured or negotiated on an RTP session basis.

3.3.1.  Reliability



   The use of RTCP messages implies that each message transfer is
   unreliable, unless the lower layer transport provides reliability.
   The different messages proposed in this specification have different
   requirements in terms of reliability.  However, in all cases, the
   reaction to an (occasional) loss of a feedback message is specified.

3.4.  Multicast



   The codec control messages might be used with multicast.  The RTCP
   timing rules specified in [RFC3550] and [RFC4585] ensure that the
   messages do not cause overload of the RTCP connection.  The use of
   multicast may result in the reception of messages with inconsistent
   semantics.  The reaction to inconsistencies depends on the message
   type, and is discussed for each message type separately.

3.5.  Feedback Messages



   This section describes the semantics of the different feedback
   messages and how they apply to the different use cases.

3.5.1.  Full Intra Request Command



   A Full Intra Request (FIR) Command, when received by the designated
   media sender, requires that the media sender sends a Decoder Refresh
   Point (see section 2.2) at the earliest opportunity.  The evaluation
   of such an opportunity includes the current encoder coding strategy
   and the current available network resources.

   FIR is also known as an "instantaneous decoder refresh request",
   "fast video update request" or "video fast update request".

   Using a decoder refresh point implies refraining from using any
   picture sent prior to that point as a reference for the encoding
   process of any subsequent picture sent in the stream.  For predictive
   media types that are not video, the analogue applies.  For example,
   if in MPEG-4 systems scene updates are used, the decoder refresh
   point consists of the full representation of the scene and is not
   delta-coded relative to previous updates.








Wenger, et al.              Standards Track                    [Page 12]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Decoder refresh points, especially Intra or IDR pictures, are in
   general several times larger in size than predicted pictures.  Thus,
   in scenarios in which the available bit rate is small, the use of a
   decoder refresh point implies a delay that is significantly longer
   than the typical picture duration.

   Usage in multicast is possible; however, aggregation of the commands
   is recommended.  A receiver that receives a request closely after
   sending a decoder refresh point -- within 2 times the longest round
   trip time (RTT) known, plus any AVPF-induced RTCP packet sending
   delays -- should await a second request message to ensure that the
   media receiver has not been served by the previously delivered
   decoder refresh point.  The reason for the specified delay is to
   avoid sending unnecessary decoder refresh points.  A session
   participant may have sent its own request while another participant's
   request was in-flight to them.  Suppressing those requests that may
   have been sent without knowledge about the other request avoids this
   issue.

   Using the FIR command to recover from errors is explicitly
   disallowed, and instead the PLI message defined in AVPF [RFC4585]
   should be used.  The PLI message reports lost pictures and has been
   included in AVPF for precisely that purpose.

   Full Intra Request is applicable in use-cases 1 and 2.

3.5.1.1.  Reliability



   The FIR message results in the delivery of a decoder refresh point,
   unless the message is lost.  Decoder refresh points are easily
   identifiable from the bit stream.  Therefore, there is no need for
   protocol-level notification, and a simple command repetition
   mechanism is sufficient for ensuring the level of reliability
   required.  However, the potential use of repetition does require a
   mechanism to prevent the recipient from responding to messages
   already received and responded to.

   To ensure the best possible reliability, a sender of FIR may repeat
   the FIR until the desired content has been received.  The repetition
   interval is determined by the RTCP timing rules applicable to the
   session.  Upon reception of a complete decoder refresh point or the
   detection of an attempt to send a decoder refresh point (which got
   damaged due to a packet loss), the repetition of the FIR must stop.
   If another FIR is necessary, the request sequence number must be
   increased.  A FIR sender shall not have more than one FIR (different
   request sequence number) outstanding at any time per media sender in
   the session.




Wenger, et al.              Standards Track                    [Page 13]

RFC 5104             Codec Control Messages in AVPF        February 2008


   The receiver of FIR (i.e., the media sender) behaves in complementary
   fashion to ensure delivery of a decoder refresh point.  If it
   receives repetitions of the FIR more than 2*RTT after it has sent a
   decoder refresh point, it shall send a new decoder refresh point.
   Two round trip times allow time for the decoder refresh point to
   arrive back to the requestor and for the end of repetitions of FIR to
   reach and be detected by the media sender.

   An RTP mixer or RTP switching MCU that receive a FIR from a media
   receiver is responsible to ensure that a decoder refresh point is
   delivered to the requesting receiver.  It may be necessary for the
   mixer/MCU to generate FIR commands.  From a reliability perspective,
   the two legs (FIR-requesting endpoint to mixer/MCU, and mixer/MCU to
   decoder refresh point generating endpoint) are handled independently
   from each other.

3.5.2.  Temporal-Spatial Trade-off Request and Notification



   The Temporal-Spatial Trade-off Request (TSTR) instructs the video
   encoder to change its trade-off between temporal and spatial
   resolution.  Index values from 0 to 31 indicate monotonically a
   desire for higher frame rate.  That is, a requester asking for an
   index of 0 prefers a high quality and is willing to accept a low
   frame rate, whereas a requester asking for 31 wishes a high frame
   rate, potentially at the cost of low spatial quality.

   In general, the encoder reaction time may be significantly longer
   than the typical picture duration.  See use case 3 for an example.
   The encoder decides whether and to what extent the request results in
   a change of the trade-off.  It returns a Temporal-Spatial Trade-off
   Notification (TSTN) message to indicate the trade-off that it will
   use henceforth.

   TSTR and TSTN have been introduced primarily because it is believed
   that control protocol mechanisms, e.g., a SIP re-invite, are too
   heavyweight and too slow to allow for a reasonable user experience.
   Consider, for example, a user interface where the remote user selects
   the temporal/spatial trade-off with a slider.  An immediate feedback
   to any slider movement is required for a reasonable user experience.
   A SIP re-INVITE [RFC3261] would require at least two round-trips more
   (compared to the TSTR/TSTN mechanism) and may involve proxies and
   other complex mechanisms.  Even in a well-designed system, it could
   take a second or so until the new trade-off is finally selected.
   Furthermore, the use of RTCP solves the multicast use case very
   efficiently.

   The use of TSTR and TSTN in multipoint scenarios is a non-trivial
   subject, and can be achieved in many implementation-specific ways.



Wenger, et al.              Standards Track                    [Page 14]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Problems stem from the fact that TSTRs will typically arrive
   unsynchronized, and may request different trade-off values for the
   same stream and/or endpoint encoder.  This memo does not specify a
   translator's, mixer's, or endpoint's reaction to the reception of a
   suggested trade-off as conveyed in the TSTR.  We only require the
   receiver of a TSTR message to reply to it by sending a TSTN, carrying
   the new trade-off chosen by its own criteria (which may or may not be
   based on the trade-off conveyed by the TSTR).  In other words, the
   trade-off sent in a TSTR is a non-binding recommendation, nothing
   more.

   Three TSTR/TSTN scenarios need to be distinguished, based on the
   topologies described in [RFC5117].  The scenarios are described in
   the following subsections.

3.5.2.1.  Point-to-Point



   In this most trivial case (Topo-Point-to-Point), the media sender
   typically adjusts its temporal/spatial trade-off based on the
   requested value in TSTR, subject to its own capabilities.  The TSTN
   message conveys back the new trade-off value (which may be identical
   to the old one if, for example, the sender is not capable of
   adjusting its trade-off).

3.5.2.2.  Point-to-Multipoint Using Multicast or Translators



   RTCP Multicast is used either with media multicast according to
   Topo-Multicast, or following RFC 3550's translator model according to
   Topo-Translator.  In these cases, unsynchronized TSTR messages from
   different receivers may be received, possibly with different
   requested trade-offs (because of different user preferences).  This
   memo does not specify how the media sender tunes its trade-off.
   Possible strategies include selecting the mean or median of all
   trade-off requests received, giving priority to certain participants,
   or continuing to use the previously selected trade-off (e.g., when
   the sender is not capable of adjusting it).  Again, all TSTR messages
   need to be acknowledged by TSTN, and the value conveyed back has to
   reflect the decision made.

3.5.2.3.  Point-to-Multipoint Using RTP Mixer



   In this scenario (Topo-Mixer), the RTP mixer receives all TSTR
   messages, and has the opportunity to act on them based on its own
   criteria.  In most cases, the mixer should form a "consensus" of
   potentially conflicting TSTR messages arriving from different
   participants, and initiate its own TSTR message(s) to the media
   sender(s).  As in the previous scenario, the strategy for forming




Wenger, et al.              Standards Track                    [Page 15]

RFC 5104             Codec Control Messages in AVPF        February 2008


   this "consensus" is up to the implementation, and can, for example,
   encompass averaging the participants' request values, giving priority
   to certain participants, or using session default values.

   Even if a mixer or translator performs transcoding, it is very
   difficult to deliver media with the requested trade-off, unless the
   content the mixer or translator receives is already close to that
   trade-off.  Thus, if the mixer changes its trade-off, it needs to
   request the media sender(s) to use the new value, by creating a TSTR
   of its own.  Upon reaching a decision on the used trade-off, it
   includes that value in the acknowledgement to the downstream
   requestors.  Only in cases where the original source has
   substantially higher quality (and bit rate) is it likely that
   transcoding alone can result in the requested trade-off.

3.5.2.4.  Reliability



   A request and reception acknowledgement mechanism is specified.  The
   Temporal-Spatial Trade-off Notification (TSTN) message informs the
   requester that its request has been received, and what trade-off is
   used henceforth.  This acknowledgement mechanism is desirable for at
   least the following reasons:

   o  A change in the trade-off cannot be directly identified from the
      media bit stream.
   o  User feedback cannot be implemented without knowing the chosen
      trade-off value, according to the media sender's constraints.
   o  Repetitive sending of messages requesting an unimplementable
      trade-off can be avoided.

3.5.3.  H.271 Video Back Channel Message



   ITU-T Rec. H.271 defines syntax, semantics, and suggested encoder
   reaction to a Video Back Channel Message.  The structure defined in
   this memo is used to transparently convey such a message from media
   receiver to media sender.  In this memo, we refrain from an in-depth
   discussion of the available code points within H.271 and refer to the
   specification text [H.271] instead.

   However, we note that some H.271 messages bear similarities with
   native messages of AVPF and this memo.  Furthermore, we note that
   some H.271 message are known to require caution in multicast
   environments -- or are plainly not usable in multicast or multipoint
   scenarios.  Table 1 provides a brief, simplified overview of the
   messages currently defined in H.271, their roughly corresponding AVPF
   or Codec Control Messages (CCMs) (the latter as specified in this
   memo), and an indication of our current knowledge of their multicast
   safety.



Wenger, et al.              Standards Track                    [Page 16]

RFC 5104             Codec Control Messages in AVPF        February 2008


   H.271 msg type      AVPF/CCM msg type    multicast-safe
   --------------------------------------------------------------------
   0 (when used for
     reference picture
      selection)        AVPF RPSI       No (positive ACK of pictures)
   1 picture loss       AVPF PLI        Yes
   2 partial loss       AVPF SLI        Yes
   3 one parameter CRC  N/A             Yes (no required sender action)
   4 all parameter CRC  N/A             Yes (no required sender action)
   5 refresh point      CCM FIR         Yes

   Table 1: H.271 messages and their AVPF/CCM equivalents

          Note: H.271 message type 0 is not a strict equivalent to
          AVPF's Reference Picture Selection Indication (RPSI); it is an
          indication of known-as-correct reference picture(s) at the
          decoder.  It does not command an encoder to use a defined
          reference picture (the form of control information envisioned
          to be carried in RPSI).  However, it is believed and intended
          that H.271 message type 0 will be used for the same purpose as
          AVPF's RPSI -- although other use forms are also possible.

   In response to the opaqueness of the H.271 messages, especially with
   respect to the multicast safety, the following guidelines MUST be
   followed when an implementation wishes to employ the H.271 video back
   channel message:

   1. Implementations utilizing the H.271 feedback message MUST stay in
      compliance with congestion control principles, as outlined in
      section 5.

   2. An implementation SHOULD utilize the IETF-native messages as
      defined in [RFC4585] and in this memo instead of similar messages
      defined in [H.271].  Our current understanding of similar messages
      is documented in Table 1 above.  One good reason to divert from
      the SHOULD statement above would be if it is clearly understood
      that, for a given application and video compression standard, the
      aforementioned "similarity" is not given, in contrast to what the
      table indicates.

   3. It has been observed that some of the H.271 code points currently
      in existence are not multicast-safe.  Therefore, the sensible
      thing to do is not to use the H.271 feedback message type in
      multicast environments.  It MAY be used only when all the issues
      mentioned later are fully understood by the implementer, and
      properly taken into account by all endpoints.  In all other cases,
      the H.271 message type MUST NOT be used in conjunction with
      multicast.



Wenger, et al.              Standards Track                    [Page 17]

RFC 5104             Codec Control Messages in AVPF        February 2008


   4. It has been observed that even in centralized multipoint
      environments, where the mixer should theoretically be able to
      resolve issues as documented below, the implementation of such a
      mixer and cooperative endpoints is a very difficult and tedious
      task.  Therefore, H.271 messages MUST NOT be used in centralized
      multipoint scenarios, unless all the issues mentioned below are
      fully understood by the implementer, and properly taken into
      account by both mixer and endpoints.

   Issues to be taken into account when considering the use of H.271 in
   multipoint environments:

   1. Different state on different receivers.  In many environments, it
      cannot be guaranteed that the decoder state of all media receivers
      is identical at any given point in time.  The most obvious reason
      for such a possible misalignment of state is a loss that occurs on
      the path to only one of many media receivers.  However, there are
      other not so obvious reasons, such as recent joins to the
      multipoint conference (be it by joining the multicast group or
      through additional mixer output).  Different states can lead the
      media receivers to issue potentially contradicting H.271 messages
      (or one media receiver issuing an H.271 message that, when
      observed by the media sender, is not helpful for the other media
      receivers).  A naive reaction of the media sender to these
      contradicting messages can lead to unpredictable and annoying
      results.

   2. Combining messages from different media receivers in a media
      sender is a non-trivial task.  As reasons, we note that these
      messages may be contradicting each other, and that their transport
      is unreliable (there may well be other reasons).  In case of many
      H.271 messages (i.e., types 0, 2, 3, and 4), the algorithm for
      combining must be aware both of the network/protocol environment
      (i.e., with respect to congestion) and of the media codec
      employed, as H.271 messages of a given type can have different
      semantics for different media codecs.

   3. The suppression of requests may need to go beyond the basic
      mechanisms described in AVPF (which are driven exclusively by
      timing and transport considerations on the protocol level).  For
      example, a receiver is often required to refrain from (or delay)
      generating requests, based on information it receives from the
      media stream.  For instance, it makes no sense for a receiver to
      issue a FIR when a transmission of an Intra/IDR picture is
      ongoing.






Wenger, et al.              Standards Track                    [Page 18]

RFC 5104             Codec Control Messages in AVPF        February 2008


   4. When using the non-multicast-safe messages (e.g., H.271 type 0
      positive ACK of received pictures/slices) in larger multicast
      groups, the media receiver will likely be forced to delay or even
      omit sending these messages.  For the media sender, this looks
      like data has not been properly received (although it was received
      properly), and a naively implemented media sender reacts to these
      perceived problems where it should not.

3.5.3.1.  Reliability



   H.271 Video Back Channel Messages do not require reliable
   transmission, and confirmation of the reception of a message can be
   derived from the forward video bit stream.  Therefore, no specific
   reception acknowledgement is specified.

   With respect to re-sending rules, section 3.5.1.1 applies.

3.5.4.  Temporary Maximum Media Stream Bit Rate Request and Notification



   A receiver, translator, or mixer uses the Temporary Maximum Media
   Stream Bit Rate Request (TMMBR, "timber") to request a sender to
   limit the maximum bit rate for a media stream (see section 2.2) to,
   or below, the provided value.  The Temporary Maximum Media Stream Bit
   Rate Notification (TMMBN) contains the media sender's current view of
   the most limiting subset of the TMMBR-defined limits it has received,
   to help the participants to suppress TMMBRs that would not further
   restrict the media sender.  The primary usage for the TMMBR/TMMBN
   messages is in a scenario with an MCU or mixer (use case 6),
   corresponding to Topo-Translator or Topo-Mixer, but also to Topo-
   Point-to-Point.

   Each temporary limitation on the media stream is expressed as a
   tuple.  The first component of the tuple is the maximum total media
   bit rate (as defined in section 2.2) that the media receiver is
   currently prepared to accept for this media stream.  The second
   component is the per-packet overhead that the media receiver has
   observed for this media stream at its chosen reference protocol
   layer.

   As indicated in section 2.2, the overhead as observed by the sender
   of the TMMBR (i.e., the media receiver) may differ from the overhead
   observed at the receiver of the TMMBR (i.e., the media sender) due to
   use of a different reference protocol layer at the other end or due
   to the intervention of translators or mixers that affect the amount
   of per packet overhead.  For example, a gateway in between the two
   that converts between IPv4 and IPv6 affects the per-packet overhead
   by 20 bytes.  Other mechanisms that change the overhead include
   tunnels.  The problem with varying overhead is also discussed in



Wenger, et al.              Standards Track                    [Page 19]

RFC 5104             Codec Control Messages in AVPF        February 2008


   [RFC3890].  As will be seen in the description of the algorithm for
   use of TMMBR, the difference in perceived overhead between the
   sending and receiving ends presents no difficulty because
   calculations are carried out in terms of variables that have the same
   value at the sender as at the receiver -- for example, packet rate
   and net media rate.

   Reporting both maximum total media bit rate and per-packet overhead
   allows different receivers to provide bit rate and overhead values
   for different protocol layers, for example, at the IP level, at the
   outer part of a tunnel protocol, or at the link layer.  The protocol
   level a peer reports on depends on the level of integration the peer
   has, as it needs to be able to extract the information from that
   protocol level.  For example, an application with no knowledge of the
   IP version it is running over cannot meaningfully determine the
   overhead of the IP header, and hence will not want to include IP
   overhead in the overhead or maximum total media bit rate calculation.

   It is expected that most peers will be able to report values at least
   for the IP layer.  In certain implementations, it may be advantageous
   to also include information pertaining to the link layer, which in
   turn allows for a more precise overhead calculation and a better
   optimization of connectivity resources.

   The Temporary Maximum Media Stream Bit Rate messages are generic
   messages that can be applied to any RTP packet stream.  This
   separates them from the other codec control messages defined in this
   specification, which apply only to specific media types or payload
   formats.  The TMMBR functionality applies to the transport, and the
   requirements the transport places on the media encoding.

   The reasoning below assumes that the participants have negotiated a
   session maximum bit rate, using a signaling protocol.  This value can
   be global, for example, in case of point-to-point, multicast, or
   translators.  It may also be local between the participant and the
   peer or mixer.  In either case, the bit rate negotiated in signaling
   is the one that the participant guarantees to be able to handle
   (depacketize and decode).  In practice, the connectivity of the
   participant also influences the negotiated value -- it does not make
   much sense to negotiate a total media bit rate that one's network
   interface does not support.

   It is also beneficial to have negotiated a maximum packet rate for
   the session or sender.  RFC 3890 provides an SDP [RFC4566] attribute
   that can be used for this purpose; however, that attribute is not
   usable in RTP sessions established using offer/answer [RFC3264].
   Therefore, an optional maximum packet rate signaling parameter is
   specified in this memo.



Wenger, et al.              Standards Track                    [Page 20]

RFC 5104             Codec Control Messages in AVPF        February 2008


   An already established maximum total media bit rate may be changed at
   any time, subject to the timing rules governing the sending of
   feedback messages.  The limit may change to any value between zero
   and the session maximum, as negotiated during session establishment
   signaling.  However, even if a sender has received a TMMBR message
   allowing an increase in the bit rate, all increases must be governed
   by a congestion control mechanism.  TMMBR indicates known limitations
   only, usually in the local environment, and does not provide any
   guarantees about the full path.  Furthermore, any increases in
   TMMBR-established bit rate limits are to be executed only after a
   certain delay from the sending of the TMMBN message that notifies the
   world about the increase in limit.  The delay is specified as at
   least twice the longest RTT as known by the media sender, plus the
   media sender's calculation of the required wait time for the sending
   of another TMMBR message for this session based on AVPF timing rules.
   This delay is introduced to allow other session participants to make
   known their bit rate limit requirements, which may be lower.

   If it is likely that the new value indicated by TMMBR will be valid
   for the remainder of the session, the TMMBR sender is expected to
   perform a renegotiation of the session upper limit using the session
   signaling protocol.

3.5.4.1.  Behavior for Media Receivers Using TMMBR



   This section is an informal description of behaviour described more
   precisely in section 4.2.

   A media sender begins the session limited by the maximum media bit
   rate and maximum packet rate negotiated in session signaling, if any.
   Note that this value may be negotiated for another protocol layer
   than the one the participant uses in its TMMBR messages.  Each media
   receiver selects a reference protocol layer, forms an estimate of the
   overhead it is observing (or estimating it if no packets has been
   seen yet) at that reference level, and determines the maximum total
   media bit rate it can accept, taking into account its own limitations
   and any transport path limitations of which it may be aware.  In case
   the current limitations are more restricting than what was agreed on
   in the session signaling, the media receiver reports its initial
   estimate of these two quantities to the media sender using a TMMBR
   message.  Overall message traffic is reduced by the possibility of
   including tuples for multiple media senders in the same TMMBR
   message.

   The media sender applies an algorithm such as that specified in
   section 3.5.4.2 to select which of the tuples it has received are
   most limiting (i.e., the bounding set as defined in section 2.2).  It
   modifies its operation to stay within the feasible region (as defined



Wenger, et al.              Standards Track                    [Page 21]

RFC 5104             Codec Control Messages in AVPF        February 2008


   in section 2.2), and also sends out a TMMBN to the media receivers
   indicating the selected bounding set.  That notification also
   indicates who was responsible for the tuples in the bounding set,
   i.e., the "owner"(s) of the limitation.  A session participant that
   owns no tuple in the bounding set is called a "non-owner".

   If a media receiver does not own one of the tuples in the bounding
   set reported by the TMMBN, it applies the same algorithm as the media
   sender to determine if its current estimated (maximum total media bit
   rate, overhead) tuple would enter the bounding set if known to the
   media sender.  If so, it issues a TMMBR reporting the tuple value to
   the sender.  Otherwise, it takes no action for the moment.
   Periodically, its estimated tuple values may change or it may receive
   a new TMMBN.  If so, it reapplies the algorithm to decide whether it
   needs to issue a TMMBR.

   If, alternatively, a media receiver owns one of the tuples in the
   reported bounding set, it takes no action until such time as its
   estimate of its own tuple values changes.  At that time, it sends a
   TMMBR to the media sender to report the changed values.

   A media receiver may change status between owner and non-owner of a
   bounding tuple between one TMMBN message and the next.  Thus, it must
   check the contents of each TMMBN to determine its subsequent actions.

   Implementations may use other algorithms of their choosing, as long
   as the bit rate limitations resulting from the exchange of TMMBR and
   TMMBN messages are at least as strict (at least as low, in the bit
   rate dimension) as the ones resulting from the use of the
   aforementioned algorithm.

   Obviously, in point-to-point cases, when there is only one media
   receiver, this receiver becomes "owner" once it receives the first
   TMMBN in response to its own TMMBR, and stays "owner" for the rest of
   the session.  Therefore, when it is known that there will always be
   only a single media receiver, the above algorithm is not required.
   Media receivers that are aware they are the only ones in a session
   can send TMMBR messages with bit rate limits both higher and lower
   than the previously notified limit, at any time (subject to the AVPF
   [RFC4585] RTCP RR send timing rules).  However, it may be difficult
   for a session participant to determine if it is the only receiver in
   the session.  Because of this, any implementation of TMMBR is
   required to include the algorithm described in the next section or a
   stricter equivalent.







Wenger, et al.              Standards Track                    [Page 22]

RFC 5104             Codec Control Messages in AVPF        February 2008


3.5.4.2.  Algorithm for Establishing Current Limitations



   This section introduces an example algorithm for the calculation of a
   session limit.  Other algorithms can be employed, as long as the
   result of the calculation is at least as restrictive as the result
   that is obtained by this algorithm.

   First, it is important to consider the implications of using a tuple
   for limiting the media sender's behavior.  The bit rate and the
   overhead value result in a two-dimensional solution space for the
   calculation of the bit rate of media streams.  Fortunately, the two
   variables are linked.  Specifically, the bit rate available for RTP
   payloads is equal to the TMMBR reported bit rate minus the packet
   rate used, multiplied by the TMMBR reported overhead converted to
   bits.  As a result, when different bit rate/overhead combinations
   need to be considered, the packet rate determines the correct
   limitation.  This is perhaps best explained by an example:

   Example:

   Receiver A: TMMBR_max total BR = 35 kbps, TMMBR_OH = 40 bytes
   Receiver B: TMMBR_max total BR = 40 kbps, TMMBR_OH = 60 bytes

   For a given packet rate (PR), the bit rate available for media
   payloads in RTP will be:

   Max_net media_BR_A =
       TMMBR_max total BR_A - PR * TMMBR_OH_A * 8 ... (1)

   Max_net media_BR_B =
       TMMBR_max total BR_B - PR * TMMBR_OH_B * 8 ... (2)

   For a PR = 20, these calculations will yield a Max_net media_BR_A =
   28600 bps and Max_net media_BR_B = 30400 bps, which suggests that
   receiver A is the limiting one for this packet rate.  However, at a
   certain PR there is a switchover point at which receiver B becomes
   the limiting one.  The switchover point can be identified by setting
   Max_media_BR_A equal to Max_media_BR_B and breaking out PR:

         TMMBR_max total BR_A - TMMBR_max total BR_B
   PR =  ------------------------------------------- ... (3)
                8*(TMMBR_OH_A - TMMBR_OH_B)

   which, for the numbers above, yields 31.25 as the switchover point
   between the two limits.  That is, for packet rates below 31.25 per
   second, receiver A is the limiting receiver, and for higher packet
   rates, receiver B is more limiting.  The implications of this
   behavior have to be considered by implementations that are going to



Wenger, et al.              Standards Track                    [Page 23]

RFC 5104             Codec Control Messages in AVPF        February 2008


   control media encoding and its packetization.  As exemplified above,
   multiple TMMBR limits may apply to the trade-off between net media
   bit rate and packet rate.  Which limitation applies depends on the
   packet rate being considered.

   This also has implications for how the TMMBR mechanism needs to work.
   First, there is the possibility that multiple TMMBR tuples are
   providing limitations on the media sender.  Secondly, there is a need
   for any session participant (media sender and receivers) to be able
   to determine if a given tuple will become a limitation upon the media
   sender, or if the set of already given limitations is stricter than
   the given values.  In the absence of the ability to make this
   determination, the suppression of TMMBRs would not work.

   The basic idea of the algorithm is as follows.  Each TMMBR tuple can
   be viewed as the equation of a straight line (cf. equations (1) and
   (2)) in a space where packet rate lies along the X-axis and net bit
   rate along the Y-axis.  The lower envelope of the set of lines
   corresponding to the complete set of TMMBR tuples, together with the
   X and Y axes, defines a polygon.  Points lying within this polygon
   are combinations of packet rate and bit rate that meet all of the
   TMMBR constraints.  The highest feasible packet rate within this
   region is the minimum of the rate at which the bounding polygon meets
   the X-axis or the session maximum packet rate (SMAXPR, measured in
   packets per second) provided by signaling, if any.  Typically, a
   media sender will prefer to operate at a lower rate than this
   theoretical maximum, so as to increase the rate at which actual media
   content reaches the receivers.  The purpose of the algorithm is to
   distinguish the TMMBR tuples constituting the bounding set and thus
   delineate the feasible region, so that the media sender can select
   its preferred operating point within that region

   Figure 1 below shows a bounding polygon formed by TMMBR tuples A and
   B.  A third tuple C lies outside the bounding polygon and is
   therefore irrelevant in determining feasible trade-offs between media
   rate and packet rate.  The line labeled ss..s represents the limit on
   packet rate imposed by the session maximum packet rate (SMAXPR)
   obtained by signaling during session setup.  In Figure 1, the limit
   determined by tuple B happens to be more restrictive than SMAXPR.
   The situation could easily be the reverse, meaning that the bounding
   polygon is terminated on the right by the vertical line representing
   the SMAXPR constraint.









Wenger, et al.              Standards Track                    [Page 24]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Net  ^
   Media|a   c   b             s
   Bit  |  a   c  b            s
   Rate |    a   c b           s
        |      a   cb          s
        |        a   c         s
        |          a  bc       s
        |            a b c     s
        |              ab  c   s
        |  Feasible      b   c s
        |   region        ba   s
        |                  b a s c
        |                   b  s   c
        |                    b s a
        |                     bs
        +------------------------------>

              Packet rate

    Figure 1 - Geometric Interpretation of TMMBR Tuples

   Note that the slopes of the lines making up the bounding polygon are
   increasingly negative as one moves in the direction of increasing
   packet rate.  Note also that with slight rearrangement, equations (1)
   and (2) have the canonical form:

          y = mx + b

   where
     m is the slope and has value equal to the negative of the tuple
     overhead (in bits),
   and
     b is the y-intercept and has value equal to the tuple maximum
     total media bit rate.

   These observations lead to the conclusion that when processing the
   TMMBR tuples to select the initial bounding set, one should sort and
   process the tuples by order of increasing overhead.  Once a
   particular tuple has been added to the bounding set, all tuples not
   already selected and having lower overhead can be eliminated, because
   the next side of the bounding polygon has to be steeper (i.e., the
   corresponding TMMBR must have higher overhead) than the latest added
   tuple.

   Line cc..c in Figure 1 illustrates another principle.  This line is
   parallel to line aa..a, but has a higher Y-intercept.  That is, the
   corresponding TMMBR tuple contains a higher maximum total media bit
   rate value.  Since line cc..c is outside the bounding polygon, it



Wenger, et al.              Standards Track                    [Page 25]

RFC 5104             Codec Control Messages in AVPF        February 2008


   illustrates the conclusion that if two TMMBR tuples have the same
   overhead value, the one with higher maximum total media bit rate
   value cannot be part of the bounding set and can be set aside.

   Two further observations complete the algorithm.  Obviously, moving
   from the left, the successive corners of the bounding polygon (i.e.,
   the intersection points between successive pairs of sides) lie at
   successively higher packet rates.  On the other hand, again moving
   from the left, each successive line making up the bounding set
   crosses the X-axis at a lower packet rate.

   The complete algorithm can now be specified.  The algorithm works
   with two lists of TMMBR tuples, the candidate list X and the selected
   list Y, both ordered by increasing overhead value.  The algorithm
   terminates when all members of X have been discarded or removed for
   processing.  Membership of the selected list Y is probationary until
   the algorithm is complete.  Each member of the selected list is
   associated with an intersection value, which is the packet rate at
   which the line corresponding to that TMMBR tuple intersects with the
   line corresponding to the previous TMMBR tuple in the selected list.
   Each member of the selected list is also associated with a maximum
   packet rate value, which is the lesser of the session maximum packet
   rate SMAXPR (if any) and the packet rate at which the line
   corresponding to that tuple crosses the X-axis.

   When the algorithm terminates, the selected list is equal to the
   bounding set as defined in section 2.2.

   Initial Algorithm

   This algorithm is used by the media sender when it has received one
   or more TMMBRs and before it has determined a bounding set for the
   first time.

   1. Sort the TMMBR tuples by order of increasing overhead.  This is
      the initial candidate list X.

   2. When multiple tuples in the candidate list have the same overhead
      value, discard all but the one with the lowest maximum total media
      bit rate value.

   3. Select and remove from the candidate list the TMMBR tuple with the
      lowest maximum total media bit rate value.  If there is more than
      one tuple with that value, choose the one with the highest
      overhead value.  This is the first member of the selected list Y.
      Set its intersection value equal to zero.  Calculate its maximum





Wenger, et al.              Standards Track                    [Page 26]

RFC 5104             Codec Control Messages in AVPF        February 2008


      packet rate as the minimum of SMAXPR (if available) and the value
      obtained from the following formula, which is the packet rate at
      which the corresponding line crosses the X-axis.

          Max PR = TMMBR max total BR / (8 * TMMBR OH) ... (4)

   4. Discard from the candidate list all tuples with a lower overhead
      value than the selected tuple.

   5. Remove the first remaining tuple from the candidate list for
      processing.  Call this the current candidate.



   6. Calculate the packet rate PR at the intersection of the line
      generated by the current candidate with the line generated by the
      last tuple in the selected list Y, using equation (3).

   7. If the calculated value PR is equal to or lower than the
      intersection value stored for the last tuple of the selected list,
      discard the last tuple of the selected list and go back to step 6
      (retaining the same current candidate).

      Note that the choice of the initial member of the selected list Y
      in step 3 guarantees that the selected list will never be emptied
      by this process, meaning that the algorithm must eventually (if
      not immediately) fall through to step 8.

   8. (This step is reached when the calculated PR value of the current
      candidate is greater than the intersection value of the current
      last member of the selected list Y.)  If the calculated value PR
      of the current candidate is lower than the maximum packet rate
      associated with the last tuple in the selected list, add the
      current candidate tuple to the end of the selected list.  Store PR
      as its intersection value.  Calculate its maximum packet rate as
      the lesser of SMAXPR (if available) and the maximum packet rate
      calculated using equation (4).

   9. If any tuples remain in the candidate list, go back to step 5.



   Incremental Algorithm

   The previous algorithm covered the initial case, where no selected
   list had previously been created.  It also applied only to the media
   sender.  When a previously created selected list is available at
   either the media sender or media receiver, two other cases can be
   considered:

        o when a TMMBR tuple not currently in the selected list is a
          candidate for addition;



Wenger, et al.              Standards Track                    [Page 27]

RFC 5104             Codec Control Messages in AVPF        February 2008


        o when the values change in a TMMBR tuple currently in the
          selected list.

   At the media receiver, these cases correspond, respectively, to those
   of the non-owner and owner of a tuple in the TMMBN-reported bounding
   set.

   In either case, the process of updating the selected list to take
   account of the new/changed tuple can use the basic algorithm
   described above, with the modification that the initial candidate set
   consists only of the existing selected list and the new or changed
   tuple.  Some further optimization is possible (beyond starting with a
   reduced candidate set) by taking advantage of the following
   observations.

   The first observation is that if the new/changed candidate becomes
   part of the new selected list, the result may be to cause zero or
   more other tuples to be dropped from the list.  However, if more than
   one other tuple is dropped, the dropped tuples will be consecutive.
   This can be confirmed geometrically by visualizing a new line that
   cuts off a series of segments from the previously existing bounding
   polygon.  The cut-off segments are connected one to the next, the
   geometric equivalent of consecutive tuples in a list ordered by
   overhead value.  Beyond the dropped set in either direction all of
   the tuples that were in the earlier selected list will be in the
   updated one.  The second observation is that, leaving aside the new
   candidate, the order of tuples remaining in the updated selected list
   is unchanged because their overhead values have not changed.

   The consequence of these two observations is that, once the placement
   of the new candidate and the extent of the dropped set of tuples (if
   any) has been determined, the remaining tuples can be copied directly
   from the candidate list into the selected list, preserving their
   order.  This conclusion suggests the following modified algorithm:

       o Run steps 1-4 of the basic algorithm.

       o If the new candidate has survived steps 2 and 4 and has become
          the new first member of the selected list, run steps 5-9 on
          subsequent candidates until another candidate is added to the
          selected list.  Then move all remaining candidates to the
          selected list, preserving their order.

       o If the new candidate has survived steps 2 and 4 and has not
          become the new first member of the selected list, start by
          moving all tuples in the candidate list with lower overhead
          values than that of the new candidate to the selected list,
          preserving their order.  Run steps 5-9 for the new candidate,



Wenger, et al.              Standards Track                    [Page 28]

RFC 5104             Codec Control Messages in AVPF        February 2008


          with the modification that the intersection values and maximum
          packet rates for the tuples on the selected list have to be
          calculated on the fly because they were not previously stored.
          Continue processing only until a subsequent tuple has been
          added to the selected list, then move all remaining candidates
          to the selected list, preserving their order.

          Note that the new candidate could be added to the selected
          list only to be dropped again when the next tuple is
          processed.  It can easily be seen that in this case the new
          candidate does not displace any of the earlier tuples in the
          selected list.  The limitations of ASCII art make this
          difficult to show in a figure.  Line cc..c in Figure 1 would
          be an example if it had a steeper slope (tuple C had a higher
          overhead value), but still intersected line aa..a beyond where
          line aa..a intersects line bb..b.

   The algorithm just described is approximate, because it does not take
   account of tuples outside the selected list.  To see how such tuples
   can become relevant, consider Figure 1 and suppose that the maximum
   total media bit rate in tuple A increases to the point that line
   aa..a moves outside line cc..c.  Tuple A will remain in the bounding
   set calculated by the media sender.  However, once it issues a new
   TMMBN, media receiver C will apply the algorithm and discover that
   its tuple C should now enter the bounding set.  It will issue a TMMBR
   to the media sender, which will repeat its calculation and come to
   the appropriate conclusion.

   The rules of section 4.2 require that the media sender refrain from
   raising its sending rate until media receivers have had a chance to
   respond to the TMMBN.  In the example just given, this delay ensures
   that the relaxation of tuple A does not actually result in an attempt
   to send media at a rate exceeding the capacity at C.

3.5.4.3.  Use of TMMBR in a Mixer-Based Multipoint Operation



   Assume a small mixer-based multiparty conference is ongoing, as
   depicted in Topo-Mixer of [RFC5117].  All participants have
   negotiated a common maximum bit rate that this session can use.  The
   conference operates over a number of unicast paths between the
   participants and the mixer.  The congestion situation on each of
   these paths can be monitored by the participant in question and by
   the mixer, utilizing, for example, RTCP receiver reports (RRs) or the
   transport protocol, e.g., Datagram Congestion Control Protocol (DCCP)
   [RFC4340].  However, any given participant has no knowledge of the
   congestion situation of the connections to the other participants.
   Worse, without mechanisms similar to the ones discussed in this
   document, the mixer (which is aware of the congestion situation on



Wenger, et al.              Standards Track                    [Page 29]

RFC 5104             Codec Control Messages in AVPF        February 2008


   all connections it manages) has no standardized means to inform media
   senders to slow down, short of forging its own receiver reports
   (which is undesirable).  In principle, a mixer confronted with such a
   situation is obliged to thin or transcode streams intended for
   connections that detected congestion.

   In practice, unfortunately, media-aware streaming thinning is a very
   difficult and cumbersome operation and adds undesirable delay.  If
   media-unaware, it leads very quickly to unacceptable reproduced media
   quality.  Hence, a means to slow down senders even in the absence of
   congestion on their connections to the mixer is desirable.

   To allow the mixer to throttle traffic on the individual links,
   without performing transcoding, there is a need for a mechanism that
   enables the mixer to ask a participant's media encoders to limit the
   media stream bit rate they are currently generating.  TMMBR provides
   the required mechanism.  When the mixer detects congestion between
   itself and a given participant, it executes the following procedure:

   1. It starts thinning the media traffic to the congested participant
      to the supported bit rate.

   2. It uses TMMBR to request the media sender(s) to reduce the total
      media bit rate sent by them to the mixer, to a value that is in
      compliance with congestion control principles for the slowest
      link.  Slow refers here to the available bandwidth / bit rate /
      capacity and packet rate after congestion control.

   3. As soon as the bit rate has been reduced by the sending part, the
      mixer stops stream thinning implicitly, because there is no need
      for it once the stream is in compliance with congestion control.

   This use of stream thinning as an immediate reaction tool followed up
   by a quick control mechanism appears to be a reasonable compromise
   between media quality and the need to combat congestion.

3.5.4.4.  Use of TMMBR in Point-to-Multipoint Using Multicast or
          Translators



   In these topologies, corresponding to Topo-Multicast or Topo-
   Translator, RTCP RRs are transmitted globally.  This allows all
   participants to detect transmission problems such as congestion, on a
   medium timescale.  As all media senders are aware of the congestion
   situation of all media receivers, the rationale for the use of TMMBR
   in the previous section does not apply.  However, even in this case
   the congestion control response can be improved when the unicast





Wenger, et al.              Standards Track                    [Page 30]

RFC 5104             Codec Control Messages in AVPF        February 2008


   links are using congestion controlled transport protocols (such as
   TCP or DCCP).  A peer may also report local limitations to the media
   sender.

3.5.4.5.  Use of TMMBR in Point-to-Point Operation



   In use case 7, it is possible to use TMMBR to improve the performance
   when the known upper limit of the bit rate changes.  In this use
   case, the signaling protocol has established an upper limit for the
   session and total media bit rates.  However, at the time of transport
   link bit rate reduction, a receiver can avoid serious congestion by
   sending a TMMBR to the sending side.  Thus, TMMBR is useful for
   putting restrictions on the application and thus placing the
   congestion control mechanism in the right ballpark.  However, TMMBR
   is usually unable to provide the continuously quick feedback loop
   required for real congestion control.  Nor do its semantics match
   those of congestion control given its different purpose.  For these
   reasons, TMMBR SHALL NOT be used as a substitute for congestion
   control.

3.5.4.6.  Reliability



   The reaction of a media sender to the reception of a TMMBR message is
   not immediately identifiable through inspection of the media stream.
   Therefore, a more explicit mechanism is needed to avoid unnecessary
   re-sending of TMMBR messages.  Using a statistically based
   retransmission scheme would only provide statistical guarantees of
   the request being received.  It would also not avoid the
   retransmission of already received messages.  In addition, it would
   not allow for easy suppression of other participants' requests.  For
   these reasons, a mechanism based on explicit notification is used.

   Upon the reception of a TMMBR, a media sender sends a TMMBN
   containing the current bounding set, and indicating which session
   participants own that limit.  In multicast scenarios, that allows all
   other participants to suppress any request they may have, if their
   limitations are less strict than the current ones (i.e., define lines
   lying outside the feasible region as defined in section 2.2).
   Keeping and notifying only the bounding set of tuples allows for
   small message sizes and media sender states.  A media sender only
   keeps state for the SSRCs of the current owners of the bounding set
   of tuples; all other requests and their sources are not saved.  Once
   the bounding set has been established, new TMMBR messages should be
   generated only by owners of the bounding tuples and by other entities
   that determine (by applying the algorithm of section 3.5.4.2 or its
   equivalent) that their limitations should now be part of the bounding
   set.




Wenger, et al.              Standards Track                    [Page 31]

RFC 5104             Codec Control Messages in AVPF        February 2008


4.  RTCP Receiver Report Extensions

   This memo specifies six new feedback messages.  The Full Intra
   Request (FIR), Temporal-Spatial Trade-off Request (TSTR), Temporal-
   Spatial Trade-off Notification (TSTN), and Video Back Channel Message
   (VBCM) are "Payload Specific Feedback Messages" as defined in section
   6.3 of AVPF [RFC4585].  The Temporary Maximum Media Stream Bit Rate
   Request (TMMBR) and Temporary Maximum Media Stream Bit Rate
   Notification (TMMBN) are "Transport Layer Feedback Messages" as
   defined in section 6.2 of AVPF.

   The new feedback messages are defined in the following subsections,
   following a similar structure to that in sections 6.2 and 6.3 of the
   AVPF specification [RFC4585].

4.1.  Design Principles of the Extension Mechanism



   RTCP was originally introduced as a channel to convey presence,
   reception quality statistics and hints on the desired media coding.
   A limited set of media control mechanisms was introduced in early RTP
   payload formats for video formats, for example, in RFC 2032 [RFC2032]
   (which was obsoleted by RFC 4587 [RFC4587]).  However, this
   specification, for the first time, suggests a two-way handshake for
   some of its messages.  There is danger that this introduction could
   be misunderstood as a precedent for the use of RTCP as an RTP session
   control protocol.  To prevent such a misunderstanding, this
   subsection attempts to clarify the scope of the extensions specified
   in this memo, and it strongly suggests that future extensions follow
   the rationale spelled out here, or compellingly explain why they
   divert from the rationale.

   In this memo, and in AVPF [RFC4585], only such messages have been
   included as:

   a) have comparatively strict real-time constraints, which prevent the
      use of mechanisms such as a SIP re-invite in most application
      scenarios (the real-time constraints are explained separately for
      each message where necessary);

   b) are multicast-safe in that the reaction to potentially
      contradicting feedback messages is specified, as necessary for
      each message; and

   c) are directly related to activities of a certain media codec, class
      of media codecs (e.g., video codecs), or a given RTP packet
      stream.





Wenger, et al.              Standards Track                    [Page 32]

RFC 5104             Codec Control Messages in AVPF        February 2008


   In this memo, a two-way handshake is introduced only for messages for
   which:

   a) a notification or acknowledgement is required due to their nature.
      An analysis to determine whether this requirement exists has been
      performed separately for each message.

   b) the notification or acknowledgement cannot be easily derived from
      the media bit stream.

   All messages in AVPF [RFC4585] and in this memo present their
   contents in a simple, fixed binary format.  This accommodates media
   receivers that have not implemented higher control protocol
   functionalities (SDP, XML parsers, and such) in their media path.

   Messages that do not conform to the design principles just described
   are not an appropriate use of RTCP or of the Codec Control Framework
   defined in this document.

4.2.  Transport Layer Feedback Messages



   As specified in section 6.1 of RFC 4585 [RFC4585], transport layer
   feedback messages are identified by the RTCP packet type value RTPFB
   (205).

   In AVPF, one message of this category had been defined.  This memo
   specifies two more such messages.  They are identified by means of
   the feedback message type (FMT) parameter as follows:

   Assigned in AVPF [RFC4585]:

      1:    Generic NACK
      31:   reserved for future expansion of the identifier number space

   Assigned in this memo:

      2:    reserved (see note below)
      3:    Temporary Maximum Media Stream Bit Rate Request (TMMBR)
      4:    Temporary Maximum Media Stream Bit Rate Notification (TMMBN)

          Note: early versions of AVPF [RFC4585] reserved FMT=2 for a
          code point that has later been removed.  It has been pointed
          out that there may be implementations in the field using this
          value in accordance with the expired document.  As there is
          sufficient numbering space available, we mark FMT=2 as
          reserved so to avoid possible interoperability problems with
          any such early implementations.




Wenger, et al.              Standards Track                    [Page 33]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Available for assignment:

      0:    unassigned
      5-30: unassigned



   The following subsection defines the formats of the Feedback Control
   Information (FCI) entries for the TMMBR and TMMBN messages,
   respectively, and specifies the associated behaviour at the media
   sender and receiver.

4.2.1.  Temporary Maximum Media Stream Bit Rate Request (TMMBR)



   The Temporary Maximum Media Stream Bit Rate Request is identified by
   RTCP packet type value PT=RTPFB and FMT=3.

   The FCI field of a Temporary Maximum Media Stream Bit Rate Request
   (TMMBR) message SHALL contain one or more FCI entries.

4.2.1.1.  Message Format



   The Feedback Control Information (FCI) consists of one or more TMMBR
   FCI entries with the following syntax:

    0                   1                   2                   3
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                              SSRC                             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   | MxTBR Exp |  MxTBR Mantissa                 |Measured Overhead|
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

        Figure 2 - Syntax of an FCI Entry in the TMMBR Message

     SSRC (32 bits): The SSRC value of the media sender that is
              requested to obey the new maximum bit rate.

     MxTBR Exp (6 bits): The exponential scaling of the mantissa for the
              maximum total media bit rate value.  The value is an
              unsigned integer [0..63].

     MxTBR Mantissa (17 bits): The mantissa of the maximum total media
              bit rate value as an unsigned integer.

     Measured Overhead (9 bits): The measured average packet overhead
              value in bytes.  The measurement SHALL be done according
              to the description in section 4.2.1.2. The value is an
              unsigned integer [0..511].




Wenger, et al.              Standards Track                    [Page 34]

RFC 5104             Codec Control Messages in AVPF        February 2008


   The maximum total media bit rate (MxTBR) value in bits per second is
   calculated from the MxTBR exponent (exp) and mantissa in the
   following way:

      MxTBR = mantissa * 2^exp

   This allows for 17 bits of resolution in the range 0 to 131072*2^63
   (approximately 1.2*10^24).

   The length of the TMMBR feedback message SHALL be set to 2+2*N where
   N is the number of TMMBR FCI entries.

4.2.1.2.  Semantics



   Behaviour at the Media Receiver (Sender of the TMMBR)

   TMMBR is used to indicate a transport-related limitation at the
   reporting entity acting as a media receiver.  TMMBR has the form of a
   tuple containing two components.  The first value is the highest bit
   rate per sender of a media stream, available at a receiver-chosen
   protocol layer, which the receiver currently supports in this RTP
   session.  The second value is the measured header overhead in bytes
   as defined in section 2.2 and measured at the chosen protocol layer
   in the packets received for the stream.  The measurement of the
   overhead is a running average that is updated for each packet
   received for this particular media source (SSRC), using the following
   formula:

       avg_OH (new) = 15/16*avg_OH (old) + 1/16*pckt_OH,

   where avg_OH is the running (exponentially smoothed) average and
   pckt_OH is the overhead observed in the latest packet.

   If a maximum bit rate has been negotiated through signaling, the
   maximum total media bit rate that the receiver reports in a TMMBR
   message MUST NOT exceed the negotiated value converted to a common
   basis (i.e., with overheads adjusted to bring it to the same
   reference protocol layer).

   Within the common packet header for feedback messages (as defined in
   section 6.1 of [RFC4585]), the "SSRC of packet sender" field
   indicates the source of the request, and the "SSRC of media source"
   is not used and SHALL be set to 0.  Within a particular TMMBR FCI
   entry, the "SSRC of media source" in the FCI field denotes the media
   sender that the tuple applies to.  This is useful in the multicast or
   translator topologies where the reporting entity may address all of
   the media senders in a single TMMBR message using multiple FCI
   entries.



Wenger, et al.              Standards Track                    [Page 35]

RFC 5104             Codec Control Messages in AVPF        February 2008


   The media receiver SHALL save the contents of the latest TMMBN
   message received from each media sender.

   The media receiver MAY send a TMMBR FCI entry to a particular media
   sender under the following circumstances:

     o   before any TMMBN message has been received from that media
         sender;

     o   when the media receiver has been identified as the source of a
         bounding tuple within the latest TMMBN message received from
         that media sender, and the value of the maximum total media bit
         rate or the overhead relating to that media sender has changed;

     o   when the media receiver has not been identified as the source
         of a bounding tuple within the latest TMMBN message received
         from that media sender, and, after the media receiver applies
         the incremental algorithm from section 3.5.4.2 or a stricter
         equivalent, the media receiver's tuple relating to that media
         sender is determined to belong to the bounding set.

   A TMMBR FCI entry MAY be repeated in subsequent TMMBR messages if no
   Temporary Maximum Media Stream Bit Rate Notification (TMMBN) FCI has
   been received from the media sender at the time of transmission of
   the next RTCP packet.  The bit rate value of a TMMBR FCI entry MAY be
   changed from one TMMBR message to the next.  The overhead measurement
   SHALL be updated to the current value of avg_OH each time the entry
   is sent.

   If the value set by a TMMBR message is expected to be permanent, the
   TMMBR setting party SHOULD renegotiate the session parameters to
   reflect that using session setup signaling, e.g., a SIP re-invite.

   Behaviour at the Media Sender (Receiver of the TMMBR)

   When it receives a TMMBR message containing an FCI entry relating to
   it, the media sender SHALL use an initial or incremental algorithm as
   applicable to determine the bounding set of tuples based on the new
   information.  The algorithm used SHALL be at least as strict as the
   corresponding algorithm defined in section 3.5.4.2.  The media sender
   MAY accumulate TMMBRs over a small interval (relative to the RTCP
   sending interval) before making this calculation.

   Once it has determined the bounding set of tuples, the media sender
   MAY use any combination of packet rate and net media bit rate within
   the feasible region that these tuples describe to produce a lower





Wenger, et al.              Standards Track                    [Page 36]

RFC 5104             Codec Control Messages in AVPF        February 2008


   total media stream bit rate, as it may need to address a congestion
   situation or other limiting factors.  See section 5 (congestion
   control) for more discussion.

   If the media sender concludes that it can increase the maximum total
   media bit rate value, it SHALL wait before actually doing so, for a
   period long enough to allow a media receiver to respond to the TMMBN
   if it determines that its tuple belongs in the bounding set.  This
   delay period is estimated by the formula:

      2 * RTT + T_Dither_Max,

   where RTT is the longest round trip time known to the media sender
   and T_Dither_Max is defined in section 3.4 of [RFC4585].  Even in
   point-to-point sessions, a media sender MUST obey the aforementioned
   rule, as it is not guaranteed that a participant is able to determine
   correctly whether all the sources are co-located in a single node,
   and are coordinated.

   A TMMBN message SHALL be sent by the media sender at the earliest
   possible point in time, in response to any TMMBR messages received
   since the last sending of TMMBN.  The TMMBN message indicates the
   calculated set of bounding tuples and the owners of those tuples at
   the time of the transmission of the message.

   An SSRC may time out according to the default rules for RTP session
   participants, i.e., the media sender has not received any RTP or RTCP
   packets from the owner for the last five regular reporting intervals.
   An SSRC may also explicitly leave the session, with the participant
   indicating this through the transmission of an RTCP BYE packet or
   using an external signaling channel.  If the media sender determines
   that the owner of a tuple in the bounding set has left the session,
   the media sender SHALL transmit a new TMMBN containing the previously
   determined set of bounding tuples but with the tuple belonging to the
   departed owner removed.

   A media sender MAY proactively initiate the equivalent to a TMMBR
   message to itself, when it is aware that its transmission path is
   more restrictive than the current limitations.  As a result, a TMMBN
   indicating the media source itself as the owner of a tuple is being
   sent, thereby avoiding unnecessary TMMBR messages from other
   participants.  However, like any other participant, when the media
   sender becomes aware of changed limitations, it is required to change
   the tuple, and to send a corresponding TMMBN.







Wenger, et al.              Standards Track                    [Page 37]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Discussion

   Due to the unreliable nature of transport of TMMBR and TMMBN, the
   above rules may lead to the sending of TMMBR messages that appear to
   disobey those rules.  Furthermore, in multicast scenarios it can
   happen that more than one "non-owning" session participant may
   determine, rightly or wrongly, that its tuple belongs in the bounding
   set.  This is not critical for a number of reasons:

   a) If a TMMBR message is lost in transmission, either the media
      sender sends a new TMMBN message in response to some other media
      receiver or it does not send a new TMMBN message at all.  In the
      first case, the media receiver applies the incremental algorithm
      and, if it determines that its tuple should be part of the
      bounding set, sends out another TMMBR.  In the second case, it
      repeats the sending of a TMMBR unconditionally.  Either way, the
      media sender eventually gets the information it needs.

   b) Similarly, if a TMMBN message gets lost, the media receiver that
      has sent the corresponding TMMBR does not receive the notification
      and is expected to re-send the request and trigger the
      transmission of another TMMBN.

   c) If multiple competing TMMBR messages are sent by different session
      participants, then the algorithm can be applied taking all of
      these messages into account, and the resulting TMMBN provides the
      participants with an updated view of how their tuples compare with
      the bounded set.

   d) If more than one session participant happens to send TMMBR
      messages at the same time and with the same tuple component
      values, it does not matter which of those tuples is taken into the
      bounding set.  The losing session participant will determine,
      after applying the algorithm, that its tuple does not enter the
      bounding set, and will therefore stop sending its TMMBR.

   It is important to consider the security risks involved with faked
   TMMBRs.  See the security considerations in section 6.

   As indicated already, the feedback messages may be used in both
   multicast and unicast sessions in any of the specified topologies.
   However, for sessions with a large number of participants, using the
   lowest common denominator, as required by this mechanism, may not be
   the most suitable course of action.  Large sessions may need to
   consider other ways to adapt the bit rate to participants'
   capabilities, such as partitioning the session into different quality
   tiers or using some other method of achieving bit rate scalability.




Wenger, et al.              Standards Track                    [Page 38]

RFC 5104             Codec Control Messages in AVPF        February 2008


4.2.1.3.  Timing Rules



   The first transmission of the TMMBR message MAY use early or
   immediate feedback in cases when timeliness is desirable.  Any
   repetition of a request message SHOULD use regular RTCP mode for its
   transmission timing.

4.2.1.4.  Handling in Translators and Mixers



   Media translators and mixers will need to receive and respond to
   TMMBR messages as they are part of the chain that provides a certain
   media stream to the receiver.  The mixer or translator may act
   locally on the TMMBR and thus generate a TMMBN to indicate that it
   has done so.  Alternatively, in the case of a media translator it can
   forward the request, or in the case of a mixer generate one of its
   own and pass it forward.  In the latter case, the mixer will need to
   send a TMMBN back to the original requestor to indicate that it is
   handling the request.

4.2.2.  Temporary Maximum Media Stream Bit Rate Notification (TMMBN)



   The Temporary Maximum Media Stream Bit Rate Notification is
   identified by RTCP packet type value PT=RTPFB and FMT=4.

   The FCI field of the TMMBN feedback message may contain zero, one, or
   more TMMBN FCI entries.

4.2.2.1.  Message Format



   The Feedback Control Information (FCI) consists of zero, one, or more
   TMMBN FCI entries with the following syntax:

    0                   1                   2                   3
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                              SSRC                             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   | MxTBR Exp |  MxTBR Mantissa                 |Measured Overhead|
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

        Figure 3 - Syntax of an FCI Entry in the TMMBN Message

     SSRC (32 bits): The SSRC value of the "owner" of this tuple.

     MxTBR Exp (6 bits): The exponential scaling of the mantissa for the
              maximum total media bit rate value.  The value is an
              unsigned integer [0..63].




Wenger, et al.              Standards Track                    [Page 39]

RFC 5104             Codec Control Messages in AVPF        February 2008


     MxTBR Mantissa (17 bits): The mantissa of the maximum total media
              bit rate value as an unsigned integer.

     Measured Overhead (9 bits): The measured average packet overhead
              value in bytes represented as an unsigned integer
              [0..511].

   Thus, the FCI within the TMMBN message contains entries indicating
   the bounding tuples.  For each tuple, the entry gives the owner by
   the SSRC, followed by the applicable maximum total media bit rate and
   overhead value.

   The length of the TMMBN message SHALL be set to 2+2*N where N is the
   number of TMMBN FCI entries.

4.2.2.2.  Semantics



   This feedback message is used to notify the senders of any TMMBR
   message that one or more TMMBR messages have been received or that an
   owner has left the session.  It indicates to all participants the
   current set of bounding tuples and the "owners" of those tuples.

   Within the common packet header for feedback messages (as defined in
   section 6.1 of [RFC4585]), the "SSRC of packet sender" field
   indicates the source of the notification.  The "SSRC of media source"
   is not used and SHALL be set to 0.

   A TMMBN message SHALL be scheduled for transmission after the
   reception of a TMMBR message with an FCI entry identifying this media
   sender.  Only a single TMMBN SHALL be sent, even if more than one
   TMMBR message is received between the scheduling of the transmission
   and the actual transmission of the TMMBN message.  The TMMBN message
   indicates the bounding tuples and their owners at the time of
   transmitting the message.  The bounding tuples included SHALL be the
   set arrived at through application of the applicable algorithm of
   section 3.5.4.2 or an equivalent, applied to the previous bounding
   set, if any, and tuples received in TMMBR messages since the last
   TMMBN was transmitted.

   The reception of a TMMBR message SHALL still result in the
   transmission of a TMMBN message even if, after application of the
   algorithm, the newly reported TMMBR tuple is not accepted into the
   bounding set.  In such a case, the bounding tuples and their owners
   are not changed, unless the TMMBR was from an owner of a tuple within
   the previously calculated bounding set.  This procedure allows
   session participants that did not see the last TMMBN message to get a
   correct view of this media sender's state.




Wenger, et al.              Standards Track                    [Page 40]

RFC 5104             Codec Control Messages in AVPF        February 2008


   As indicated in section 4.2.1.2, when a media sender determines that
   an "owner" of a bounding tuple has left the session, then that tuple
   is removed from the bounding set, and the media sender SHALL send a
   TMMBN message indicating the remaining bounding tuples.  If there are
   no remaining bounding tuples, a TMMBN without any FCI SHALL be sent
   to indicate this.  Without a remaining bounding tuple, the maximum
   media bit rate and maximum packet rate negotiated in session
   signaling, if any, apply.

     Note: if any media receivers remain in the session, this last will
     be a temporary situation.  The empty TMMBN will cause every
     remaining media receiver to determine that its limitation belongs
     in the bounding set and send a TMMBR in consequence.

   In unicast scenarios (i.e., where a single sender talks to a single
   receiver), the aforementioned algorithm to determine ownership
   degenerates to the media receiver becoming the "owner" of the one
   bounding tuple as soon as the media receiver has issued the first
   TMMBR message.

4.2.2.3.  Timing Rules



   The TMMBN acknowledgement SHOULD be sent as soon as allowed by the
   applied timing rules for the session.  Immediate or early feedback
   mode SHOULD be used for these messages.

4.2.2.4.  Handling by Translators and Mixers



   As discussed in section 4.2.1.4, mixers or translators may need to
   issue TMMBN messages as responses to TMMBR messages for SSRCs handled
   by them.

4.3.  Payload-Specific Feedback Messages



   As specified by section 6.1 of RFC 4585 [RFC4585], Payload-Specific
   FB messages are identified by the RTCP packet type value PSFB (206).

   AVPF [RFC4585] defines three payload-specific feedback messages and
   one application layer feedback message.  This memo specifies four
   additional payload-specific feedback messages.  All are identified by
   means of the FMT parameter as follows:










Wenger, et al.              Standards Track                    [Page 41]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Assigned in [RFC4585]:

     1:     Picture Loss Indication (PLI)
     2:     Slice Lost Indication (SLI)
     3:     Reference Picture Selection Indication (RPSI)
     15:    Application layer FB message
     31:    reserved for future expansion of the number space

   Assigned in this memo:

     4:     Full Intra Request (FIR) Command
     5:     Temporal-Spatial Trade-off Request (TSTR)
     6:     Temporal-Spatial Trade-off Notification (TSTN)
     7:     Video Back Channel Message (VBCM)

   Unassigned:

         0: unassigned
      8-14: unassigned
     16-30: unassigned

   The following subsections define the new FCI formats for the
   payload-specific feedback messages.

4.3.1.  Full Intra Request (FIR)



   The FIR message is identified by RTCP packet type value PT=PSFB and
   FMT=4.

   The FCI field MUST contain one or more FIR entries.  Each entry
   applies to a different media sender, identified by its SSRC.

4.3.1.1.  Message Format



   The Feedback Control Information (FCI) for the Full Intra Request
   consists of one or more FCI entries, the content of which is depicted
   in Figure 4.  The length of the FIR feedback message MUST be set to
   2+2*N, where N is the number of FCI entries.

    0                   1                   2                   3
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                              SSRC                             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   | Seq nr.       |    Reserved                                   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

         Figure 4 - Syntax of an FCI Entry in the FIR Message



Wenger, et al.              Standards Track                    [Page 42]

RFC 5104             Codec Control Messages in AVPF        February 2008


     SSRC (32 bits): The SSRC value of the media sender that is
              requested to send a decoder refresh point.

     Seq nr. (8 bits): Command sequence number.  The sequence number
              space is unique for each pairing of the SSRC of command
              source and the SSRC of the command target.  The sequence
              number SHALL be increased by 1 modulo 256 for each new
              command.  A repetition SHALL NOT increase the sequence
              number.  The initial value is arbitrary.

     Reserved (24 bits): All bits SHALL be set to 0 by the sender and
              SHALL be ignored on reception.

   The semantics of this feedback message is independent of the RTP
   payload type.

4.3.1.2.  Semantics



   Within the common packet header for feedback messages (as defined in
   section 6.1 of [RFC4585]), the "SSRC of packet sender" field
   indicates the source of the request, and the "SSRC of media source"
   is not used and SHALL be set to 0.  The SSRCs of the media senders to
   which the FIR command applies are in the corresponding FCI entries.
   A FIR message MAY contain requests to multiple media senders, using
   one FCI entry per target media sender.

   Upon reception of FIR, the encoder MUST send a decoder refresh point
   (see section 2.2) as soon as possible.

   The sender MUST consider congestion control as outlined in section 5,
   which MAY restrict its ability to send a decoder refresh point
   quickly.

   FIR SHALL NOT be sent as a reaction to picture losses -- it is
   RECOMMENDED to use PLI [RFC4585] instead.  FIR SHOULD be used only in
   situations where not sending a decoder refresh point would render the
   video unusable for the users.

   A typical example where sending FIR is appropriate is when, in a
   multipoint conference, a new user joins the session and no regular
   decoder refresh point interval is established.  Another example would
   be a video switching MCU that changes streams.  Here, normally, the
   MCU issues a FIR to the new sender so to force it to emit a decoder
   refresh point.  The decoder refresh point normally includes a Freeze
   Picture Release (defined outside this specification), which re-starts
   the rendering process of the receivers.  Both techniques mentioned
   are commonly used in MCU-based multipoint conferences.




Wenger, et al.              Standards Track                    [Page 43]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Other RTP payload specifications such as RFC 2032 [RFC2032] already
   define a feedback mechanism for certain codecs.  An application
   supporting both schemes MUST use the feedback mechanism defined in
   this specification when sending feedback.  For backward-compatibility
   reasons, such an application SHOULD also be capable of receiving and
   reacting to the feedback scheme defined in the respective RTP payload
   format, if this is required by that payload format.

4.3.1.3.  Timing Rules



   The timing follows the rules outlined in section 3 of [RFC4585].  FIR
   commands MAY be used with early or immediate feedback.  The FIR
   feedback message MAY be repeated.  If using immediate feedback mode,
   the repetition SHOULD wait at least one RTT before being sent.  In
   early or regular RTCP mode, the repetition is sent in the next
   regular RTCP packet.

4.3.1.4.  Handling of FIR Message in Mixers and Translators



   A media translator or a mixer performing media encoding of the
   content for which the session participant has issued a FIR is
   responsible for acting upon it.  A mixer acting upon a FIR SHOULD NOT
   forward the message unaltered; instead, it SHOULD issue a FIR itself.

4.3.1.5. Remarks



   Currently, video appears to be the only useful application for FIR,
   as it appears to be the only RTP payload widely deployed that relies
   heavily on media prediction across RTP packet boundaries.  However,
   use of FIR could also reasonably be envisioned for other media types
   that share essential properties with compressed video, namely,
   cross-frame prediction (whatever a frame may be for that media type).
   One possible example may be the dynamic updates of MPEG-4 scene
   descriptions.  It is suggested that payload formats for such media
   types refer to FIR and other message types defined in this
   specification and in AVPF [RFC4585], instead of creating similar
   mechanisms in the payload specifications.  The payload specifications
   may have to explain how the payload-specific terminologies map to the
   video-centric terminology used herein.

   In conjunction with video codecs, FIR messages typically trigger the
   sending of full intra or IDR pictures.  Both are several times larger
   than predicted (inter) pictures.  Their size is independent of the
   time they are generated.  In most environments, especially when
   employing bandwidth-limited links, the use of an intra picture
   implies an allowed delay that is a significant multiple of the
   typical frame duration.  An example: if the sending frame rate is 10
   fps, and an intra picture is assumed to be 10 times as big as an



Wenger, et al.              Standards Track                    [Page 44]

RFC 5104             Codec Control Messages in AVPF        February 2008


   inter picture, then a full second of latency has to be accepted.  In
   such an environment, there is no need for a particularly short delay
   in sending the FIR message.  Hence, waiting for the next possible
   time slot allowed by RTCP timing rules as per [RFC4585] should not
   have an overly negative impact on the system performance.

   Mandating a maximum delay for completing the sending of a decoder
   refresh point would be desirable from an application viewpoint, but
   is problematic from a congestion control point of view.  "As soon as
   possible" as mentioned above appears to be a reasonable compromise.

   In environments where the sender has no control over the codec (e.g.,
   when streaming pre-recorded and pre-coded content), the reaction to
   this command cannot be specified.  One suitable reaction of a sender
   would be to skip forward in the video bit stream to the next decoder
   refresh point.  In other scenarios, it may be preferable not to react
   to the command at all, e.g., when streaming to a large multicast
   group.  Other reactions may also be possible.  When deciding on a
   strategy, a sender could take into account factors such as the size
   of the receiving group, the "importance" of the sender of the FIR
   message (however "importance" may be defined in this specific
   application), the frequency of decoder refresh points in the content,
   and so on.  However, a session that predominantly handles pre-coded
   content is not expected to use FIR at all.

   The relationship between the Picture Loss Indication and FIR is as
   follows.  As discussed in section 6.3.1 of AVPF [RFC4585], a Picture
   Loss Indication informs the decoder about the loss of a picture and
   hence the likelihood of misalignment of the reference pictures
   between the encoder and decoder.  Such a scenario is normally related
   to losses in an ongoing connection.  In point-to-point scenarios, and
   without the presence of advanced error resilience tools, one possible
   option for an encoder consists in sending a decoder refresh point.
   However, there are other options.  One example is that the media
   sender ignores the PLI, because the embedded stream redundancy is
   likely to clean up the reproduced picture within a reasonable amount
   of time.  The FIR, in contrast, leaves a (real-time) encoder no
   choice but to send a decoder refresh point.  It does not allow the
   encoder to take into account any considerations such as the ones
   mentioned above.

4.3.2.  Temporal-Spatial Trade-off Request (TSTR)



   The TSTR feedback message is identified by RTCP packet type value
   PT=PSFB and FMT=5.

   The FCI field MUST contain one or more TSTR FCI entries.




Wenger, et al.              Standards Track                    [Page 45]

RFC 5104             Codec Control Messages in AVPF        February 2008


4.3.2.1.  Message Format



   The content of the FCI entry for the Temporal-Spatial Trade-off
   Request is depicted in Figure 5.  The length of the feedback message
   MUST be set to 2+2*N, where N is the number of FCI entries included.

    0                   1                   2                   3
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                              SSRC                             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |  Seq nr.      |  Reserved                           | Index   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

         Figure 5 - Syntax of an FCI Entry in the TSTR Message

     SSRC (32 bits): The SSRC of the media sender that is requested to
              apply the trade-off value given in Index.

     Seq nr. (8 bits): Request sequence number.  The sequence number
              space is unique for pairing of the SSRC of request source
              and the SSRC of the request target.  The sequence number
              SHALL be increased by 1 modulo 256 for each new command.
              A repetition SHALL NOT increase the sequence number.  The
              initial value is arbitrary.

     Reserved (19 bits): All bits SHALL be set to 0 by the sender and
              SHALL be ignored on reception.

     Index (5 bits): An integer value between 0 and 31 that indicates
              the relative trade-off that is requested.  An index value
              of 0 indicates the highest possible spatial quality, while
              31 indicates the highest possible temporal resolution.

4.3.2.2.  Semantics



   A decoder can suggest a temporal-spatial trade-off level by sending a
   TSTR message to an encoder.  If the encoder is capable of adjusting
   its temporal-spatial trade-off, it SHOULD take into account the
   received TSTR message for future coding of pictures.  A value of 0
   suggests a high spatial quality and a value of 31 suggests a high
   frame rate.  The progression of values from 0 to 31 indicates
   monotonically a desire for higher frame rate.  The index values do
   not correspond to precise values of spatial quality or frame rate.







Wenger, et al.              Standards Track                    [Page 46]

RFC 5104             Codec Control Messages in AVPF        February 2008


   The reaction to the reception of more than one TSTR message by a
   media sender from different media receivers is left open to the
   implementation.  The selected trade-off SHALL be communicated to the
   media receivers by means of the TSTN message.

   Within the common packet header for feedback messages (as defined in
   section 6.1 of [RFC4585]), the "SSRC of packet sender" field
   indicates the source of the request, and the "SSRC of media source"
   is not used and SHALL be set to 0.  The SSRCs of the media senders to
   which the TSTR applies are in the corresponding FCI entries.

   A TSTR message MAY contain requests to multiple media senders, using
   one FCI entry per target media sender.

4.3.2.3.  Timing Rules



   The timing follows the rules outlined in section 3 of [RFC4585].
   This request message is not time critical and SHOULD be sent using
   regular RTCP timing.  Only if it is known that the user interface
   requires quick feedback, the message MAY be sent with early or
   immediate feedback timing.

4.3.2.4.  Handling of Message in Mixers and Translators



   A mixer or media translator that encodes content sent to the session
   participant issuing the TSTR SHALL consider the request to determine
   if it can fulfill it by changing its own encoding parameters.  A
   media translator unable to fulfill the request MAY forward the
   request unaltered towards the media sender.  A mixer encoding for
   multiple session participants will need to consider the joint needs
   of these participants before generating a TSTR on its own behalf
   towards the media sender.  See also the discussion in section 3.5.2.

4.3.2.5.  Remarks



   The term "spatial quality" does not necessarily refer to the
   resolution as measured by the number of pixels the reconstructed
   video is using.  In fact, in most scenarios the video resolution
   stays constant during the lifetime of a session.  However, all video
   compression standards have means to adjust the spatial quality at a
   given resolution, often influenced by the Quantizer Parameter or QP.
   A numerically low QP results in a good reconstructed picture quality,
   whereas a numerically high QP yields a coarse picture.  The typical
   reaction of an encoder to this request is to change its rate control
   parameters to use a lower frame rate and a numerically lower (on
   average) QP, or vice versa.  The precise mapping of Index value to





Wenger, et al.              Standards Track                    [Page 47]

RFC 5104             Codec Control Messages in AVPF        February 2008


   frame rate and QP is intentionally left open here, as it depends on
   factors such as the compression standard employed, spatial
   resolution, content, bit rate, and so on.

4.3.3.  Temporal-Spatial Trade-off Notification (TSTN)



   The TSTN message is identified by RTCP packet type value PT=PSFB and
   FMT=6.

   The FCI field SHALL contain one or more TSTN FCI entries.

4.3.3.1.  Message Format



   The content of an FCI entry for the Temporal-Spatial Trade-off
   Notification is depicted in Figure 6.  The length of the TSTN message
   MUST be set to 2+2*N, where N is the number of FCI entries.

    0                   1                   2                   3
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                              SSRC                             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |  Seq nr.      |  Reserved                           | Index   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

                    Figure 6 - Syntax of the TSTN

     SSRC (32 bits): The SSRC of the source of the TSTR that resulted in
              this Notification.

     Seq nr. (8 bits): The sequence number value from the TSTR that is
              being acknowledged.

     Reserved (19 bits): All bits SHALL be set to 0 by the sender and
              SHALL be ignored on reception.

     Index (5 bits): The trade-off value the media sender is using
              henceforth.

      Informative note: The returned trade-off value (Index) may differ
      from the requested one, for example, in cases where a media
      encoder cannot tune its trade-off, or when pre-recorded content is
      used.








Wenger, et al.              Standards Track                    [Page 48]

RFC 5104             Codec Control Messages in AVPF        February 2008


4.3.3.2.  Semantics



   This feedback message is used to acknowledge the reception of a TSTR.
   For each TSTR received targeted at the session participant, a TSTN
   FCI entry SHALL be sent in a TSTN feedback message.  A single TSTN
   message MAY acknowledge multiple requests using multiple FCI entries.
   The index value included SHALL be the same in all FCI entries of the
   TSTN message.  Including a FCI for each requestor allows each
   requesting entity to determine that the media sender received the
   request.  The Notification SHALL also be sent in response to TSTR
   repetitions received.  If the request receiver has received TSTR with
   several different sequence numbers from a single requestor, it SHALL
   only respond to the request with the highest (modulo 256) sequence
   number.  Note that the highest sequence number may be a smaller
   integer value due to the wrapping of the field.  Appendix A.1 of
   [RFC3550] has an algorithm for keeping track of the highest received
   sequence number for RTP packets; it could be adapted for this usage.

   The TSTN SHALL include the Temporal-Spatial Trade-off index that will
   be used as a result of the request.  This is not necessarily the same
   index as requested, as the media sender may need to aggregate
   requests from several requesting session participants.  It may also
   have some other policies or rules that limit the selection.

   Within the common packet header for feedback messages (as defined in
   section 6.1 of [RFC4585]), the "SSRC of packet sender" field
   indicates the source of the Notification, and the "SSRC of media
   source" is not used and SHALL be set to 0.  The SSRCs of the
   requesting entities to which the Notification applies are in the
   corresponding FCI entries.

4.3.3.3.  Timing Rules



   The timing follows the rules outlined in section 3 of [RFC4585].
   This acknowledgement message is not extremely time critical and
   SHOULD be sent using regular RTCP timing.

4.3.3.4.  Handling of TSTN in Mixers and Translators



   A mixer or translator that acts upon a TSTR SHALL also send the
   corresponding TSTN.  In cases where it needs to forward a TSTR
   itself, the notification message MAY need to be delayed until the
   TSTR has been responded to.

4.3.3.5.  Remarks



   None.




Wenger, et al.              Standards Track                    [Page 49]

RFC 5104             Codec Control Messages in AVPF        February 2008


4.3.4.  H.271 Video Back Channel Message (VBCM)



   The VBCM is identified by RTCP packet type value PT=PSFB and FMT=7.

   The FCI field MUST contain one or more VBCM FCI entries.

4.3.4.1.  Message Format



   The syntax of an FCI entry within the VBCM indication is depicted in
   Figure 7.

   0                   1                   2                   3
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                              SSRC                             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   | Seq nr.       |0| Payload Type| Length                        |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                    VBCM Octet String....      |    Padding    |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

            Figure 7 - Syntax of an FCI Entry in the VBCM

   SSRC (32 bits): The SSRC value of the media sender that is requested
          to instruct its encoder to react to the VBCM.

   Seq nr. (8 bits): Command sequence number.  The sequence number space
          is unique for pairing of the SSRC of the command source and
          the SSRC of the command target.  The sequence number SHALL be
          increased by 1 modulo 256 for each new command.  A repetition
          SHALL NOT increase the sequence number.  The initial value is
          arbitrary.

   0: Must be set to 0 by the sender and should not be acted upon by the
          message receiver.

   Payload Type (7 bits): The RTP payload type for which the VBCM bit
          stream must be interpreted.

   Length (16 bits): The length of the VBCM octet string in octets
          exclusive of any padding octets.

   VBCM Octet String (variable length): This is the octet string
          generated by the decoder carrying a specific feedback sub-
          message.

   Padding (variable length): Bits set to 0 to make up a 32-bit
          boundary.



Wenger, et al.              Standards Track                    [Page 50]

RFC 5104             Codec Control Messages in AVPF        February 2008


4.3.4.2.  Semantics



   The "payload" of the VBCM indication carries different types of
   codec-specific, feedback information.  The type of feedback
   information can be classified as a 'status report' (such as an
   indication that a bit stream was received without errors, or that a
   partial or complete picture or block was lost) or 'update requests'
   (such as complete refresh of the bit stream).

          Note: There are possible overlaps between the VBCM sub-
          messages and CCM/AVPF feedback messages, such as FIR.  Please
          see section 3.5.3 for further discussion.

   The different types of feedback sub-messages carried in the VBCM are
   indicated by the "payloadType" as defined in [H.271].  These sub-
   message types are reproduced below for convenience.  "payloadType",
   in ITU-T Rec. H.271 terminology, refers to the sub-type of the H.271
   message and should not be confused with an RTP payload type.

   Payload          Message Content
   Type
   ---------------------------------------------------------------------
   0      One or more pictures without detected bit stream error
          mismatch
   1      One or more pictures that are entirely or partially lost
   2      A set of blocks of one picture that is entirely or partially
          lost
   3      CRC for one parameter set
   4      CRC for all parameter sets of a certain type
   5      A "reset" request indicating that the sender should completely
          refresh the video bit stream as if no prior bit stream data
          had been received
   > 5    Reserved for future use by ITU-T

   Table 2: H.271 message types ("payloadTypes")

   The bit string or the "payload" of a VBCM is of variable length and
   is self-contained and coded in a variable-length, binary format.  The
   media sender necessarily has to be able to parse this optimized
   binary format to make use of VBCMs.

   Each of the different types of sub-messages (indicated by
   payloadType) may have different semantics depending on the codec
   used.

   Within the common packet header for feedback messages (as defined in
   section 6.1 of [RFC4585]), the "SSRC of packet sender" field
   indicates the source of the request, and the "SSRC of media source"



Wenger, et al.              Standards Track                    [Page 51]

RFC 5104             Codec Control Messages in AVPF        February 2008


   is not used and SHALL be set to 0.  The SSRCs of the media senders to
   which the VBCM applies are in the corresponding FCI entries.  The
   sender of the VBCM MAY send H.271 messages to multiple media senders
   and MAY send more than one H.271 message to the same media sender
   within the same VBCM.

4.3.4.3.  Timing Rules



   The timing follows the rules outlined in section 3 of [RFC4585].  The
   different sub-message types may have different properties in regards
   to the timing of messages that should be used.  If several different
   types are included in the same feedback packet, then the requirements
   for the sub-message type with the most stringent requirements should
   be followed.

4.3.4.4.  Handling of Message in Mixers or Translators



   The handling of a VBCM in a mixer or translator is sub-message type
   dependent.

4.3.4.5.  Remarks



   Please see section 3.5.3 for a discussion of the usage of H.271
   messages and messages defined in AVPF [RFC4585] and this memo with
   similar functionality.

     Note: There has been some discussion whether the RTP payload type
     field in this message is needed.  It will be needed if there is
     potentially more than one VBCM-capable RTP payload type in the same
     session, and the semantics of a given VBCM changes between payload
     types.  For example, the picture identification mechanism in
     messages of H.271 type 0 is fundamentally different between H.263
     and H.264 (although both use the same syntax).  Therefore, the
     payload field is justified here.  There was a further comment that
     for TSTR and FIR such a need does not exist, because the semantics
     of TSTR and FIR are either loosely enough defined, or generic
     enough, to apply to all video payloads currently in
     existence/envisioned.

5.  Congestion Control

   The correct application of the AVPF [RFC4585] timing rules prevents
   the network from being flooded by feedback messages.  Hence, assuming
   a correct implementation and configuration, the RTCP channel cannot
   break its bit rate commitment and introduce congestion.

   The reception of some of the feedback messages modifies the behaviour
   of the media senders or, more specifically, the media encoders.



Wenger, et al.              Standards Track                    [Page 52]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Thus, modified behaviour MUST respect the bandwidth limits that the
   application of congestion control provides.  For example, when a
   media sender is reacting to a FIR, the unusually high number of
   packets that form the decoder refresh point have to be paced in
   compliance with the congestion control algorithm, even if the user
   experience suffers from a slowly transmitted decoder refresh point.

   A change of the Temporary Maximum Media Stream Bit Rate value can
   only mitigate congestion, but not cause congestion as long as
   congestion control is also employed.  An increase of the value by a
   request REQUIRES the media sender to use congestion control when
   increasing its transmission rate to that value.  A reduction of the
   value results in a reduced transmission bit rate, thus reducing the
   risk for congestion.

6.  Security Considerations



   The defined messages have certain properties that have security
   implications.  These must be addressed and taken into account by
   users of this protocol.

   The defined setup signaling mechanism is sensitive to modification
   attacks that can result in session creation with sub-optimal
   configuration, and, in the worst case, session rejection.  To prevent
   this type of attack, authentication and integrity protection of the
   setup signaling is required.

   Spoofed or maliciously created feedback messages of the type defined
   in this specification can have the following implications:

        a. severely reduced media bit rate due to false TMMBR messages
           that sets the maximum to a very low value;

        b. assignment of the ownership of a bounding tuple to the wrong
           participant within a TMMBN message, potentially causing
           unnecessary oscillation in the bounding set as the mistakenly
           identified owner reports a change in its tuple and the true
           owner possibly holds back on changes until a correct TMMBN
           message reaches the participants;

        c. sending TSTRs that result in a video quality different from
           the user's desire, rendering the session less useful;

        d. sending multiple FIR commands to reduce the frame rate, and
           make the video jerky, due to the frequent usage of decoder
           refresh points.





Wenger, et al.              Standards Track                    [Page 53]

RFC 5104             Codec Control Messages in AVPF        February 2008


   To prevent these attacks, there is a need to apply authentication and
   integrity protection of the feedback messages.  This can be
   accomplished against threats external to the current RTP session
   using the RTP profile that combines Secure RTP [SRTP] and AVPF into
   SAVPF [SAVPF].  In the mixer cases, separate security contexts and
   filtering can be applied between the mixer and the participants, thus
   protecting other users on the mixer from a misbehaving participant.

7.  SDP Definitions



   Section 4 of [RFC4585] defines a new SDP [RFC4566] attribute, rtcp-
   fb, that may be used to negotiate the capability to handle specific
   AVPF commands and indications, such as Reference Picture Selection,
   Picture Loss Indication, etc.  The ABNF for rtcp-fb is described in
   section 4.2 of [RFC4585].  In this section, we extend the rtcp-fb
   attribute to include the commands and indications that are described
   for codec control in the present document.  We also discuss the
   Offer/Answer implications for the codec control commands and
   indications.

7.1.  Extension of the rtcp-fb Attribute



   As described in AVPF [RFC4585], the rtcp-fb attribute indicates the
   capability of using RTCP feedback.  AVPF specifies that the rtcp-fb
   attribute must only be used as a media level attribute and must not
   be provided at session level.  All the rules described in [RFC4585]
   for rtcp-fb attribute relating to payload type and to multiple rtcp-
   fb attributes in a session description also apply to the new feedback
   messages defined in this memo.

   The ABNF [RFC4234] for rtcp-fb as defined in [RFC4585] is

     "a=rtcp-fb: " rtcp-fb-pt SP rtcp-fb-val CRLF

   where rtcp-fb-pt is the payload type and rtcp-fb-val defines the type
   of the feedback message such as ack, nack, trr-int, and rtcp-fb-id.
   For example, to indicate the support of feedback of Picture Loss
   Indication, the sender declares the following in SDP

         v=0
         o=alice 3203093520 3203093520 IN IP4 host.example.com
         s=Media with feedback
         t=0 0
         c=IN IP4 host.example.com
         m=audio 49170 RTP/AVPF 98
         a=rtpmap:98 H263-1998/90000
         a=rtcp-fb:98 nack pli




Wenger, et al.              Standards Track                    [Page 54]

RFC 5104             Codec Control Messages in AVPF        February 2008


   In this document, we define a new feedback value "ccm", which
   indicates the support of codec control using RTCP feedback messages.
   The "ccm" feedback value SHOULD be used with parameters that indicate
   the specific codec control commands supported.  In this document, we
   define four such parameters, namely:

      o  "fir" indicates support of the Full Intra Request (FIR).
      o  "tmmbr" indicates support of the Temporary Maximum Media Stream
         Bit Rate Request/Notification (TMMBR/TMMBN).  It has an
         optional sub-parameter to indicate the session maximum packet
         rate (measured in packets per second) to be used.  If not
         included, this defaults to infinity.
      o  "tstr" indicates support of the Temporal-Spatial Trade-off
         Request/Notification (TSTR/TSTN).
      o  "vbcm" indicates support of H.271 Video Back Channel Messages
         (VBCMs).  It has zero or more subparameters identifying the
         supported H.271 "payloadType" values.

   In the ABNF for rtcp-fb-val defined in [RFC4585], there is a
   placeholder called rtcp-fb-id to define new feedback types.  "ccm" is
   defined as a new feedback type in this document, and the ABNF for the
   parameters for ccm is defined here (please refer to section 4.2 of
   [RFC4585] for complete ABNF syntax).

   rtcp-fb-val        =/ "ccm" rtcp-fb-ccm-param

   rtcp-fb-ccm-param  = SP "fir"   ; Full Intra Request
                      / SP "tmmbr" [SP "smaxpr=" MaxPacketRateValue]
                                   ; Temporary max media bit rate
                      / SP "tstr"  ; Temporal-Spatial Trade-Off
                      / SP "vbcm" *(SP subMessageType) ; H.271 VBCMs
                      / SP token [SP byte-string]
                              ; for future commands/indications
   subMessageType = 1*8DIGIT
   byte-string = <as defined in section 4.2 of [RFC4585] >
   MaxPacketRateValue = 1*15DIGIT

7.2.  Offer-Answer



   The Offer/Answer [RFC3264] implications for codec control protocol
   feedback messages are similar to those described in [RFC4585].  The
   offerer MAY indicate the capability to support selected codec
   commands and indications.  The answerer MUST remove all CCM
   parameters corresponding to the CCMs that it does not wish to support
   in this particular media session (for example, because it does not
   implement the message in question, or because its application logic
   suggests that support of the message adds no value).  The answerer
   MUST NOT add new ccm parameters in addition to what has been offered.



Wenger, et al.              Standards Track                    [Page 55]

RFC 5104             Codec Control Messages in AVPF        February 2008


   The answer is binding for the media session and both offerer and
   answerer MUST NOT use any feedback messages other than what both
   sides have explicitly indicated as being supported.  In other words,
   only the joint subset of CCM parameters from the offer and answer may
   be used.

   Note that including a CCM parameter in an offer or answer indicates
   that the party (offerer or answerer) is at least capable of receiving
   the corresponding CCM(s) and act upon them.  In cases when the
   reception of a negotiated CCM mandates the party to respond with
   another CCM, it must also have that capability.  Although it is not
   mandated to initiate CCMs of any negotiated type, it is generally
   expected that a party will initiate CCMs when appropriate.

   The session maximum packet rate parameter part of the TMMBR
   indication is declarative, and the highest value from offer and
   answer SHALL be used.  If the session maximum packet rate parameter
   is not present in an offer, it SHALL NOT be included by the answerer.

7.3.  Examples



   Example 1: The following SDP describes a point-to-point video call
   with H.263, with the originator of the call declaring its capability
   to support the FIR and TSTR/TSTN codec control messages.  The SDP is
   carried in a high-level signaling protocol like SIP.

         v=0
         o=alice 3203093520 3203093520 IN IP4 host.example.com
         s=Point-to-Point call
         c=IN IP4 192.0.2.124
         m=audio 49170 RTP/AVP 0
         a=rtpmap:0 PCMU/8000
         m=video 51372 RTP/AVPF 98
         a=rtpmap:98 H263-1998/90000
         a=rtcp-fb:98 ccm tstr
         a=rtcp-fb:98 ccm fir

   In the above example, when the sender receives a TSTR message from
   the remote party it is capable of adjusting the trade-off as
   indicated in the RTCP TSTN feedback message.

   Example 2: The following SDP describes a SIP end point joining a
   video mixer that is hosting a multiparty video conferencing session.
   The participant supports only the FIR (Full Intra Request) codec
   control command and it declares it in its session description.






Wenger, et al.              Standards Track                    [Page 56]

RFC 5104             Codec Control Messages in AVPF        February 2008


         v=0
         o=alice 3203093520 3203093520 IN IP4 host.example.com
         s=Multiparty Video Call
         c=IN IP4 192.0.2.124
         m=audio 49170 RTP/AVP 0
         a=rtpmap:0 PCMU/8000
         m=video 51372 RTP/AVPF 98
         a=rtpmap:98 H263-1998/90000
         a=rtcp-fb:98 ccm fir

   When the video MCU decides to route the video of this participant, it
   sends an RTCP FIR feedback message.  Upon receiving this feedback
   message, the end point is required to generate a full intra request.

   Example 3: The following example describes the Offer/Answer
   implications for the codec control messages.  The offerer wishes to
   support "tstr", "fir" and "tmmbr".  The offered SDP is

   -------------> Offer
         v=0
         o=alice 3203093520 3203093520 IN IP4 host.example.com
         s=Offer/Answer
         c=IN IP4 192.0.2.124
         m=audio 49170 RTP/AVP 0
         a=rtpmap:0 PCMU/8000
         m=video 51372 RTP/AVPF 98
         a=rtpmap:98 H263-1998/90000
         a=rtcp-fb:98 ccm tstr
         a=rtcp-fb:98 ccm fir
         a=rtcp-fb:* ccm tmmbr smaxpr=120

   The answerer wishes to support only the FIR and TSTR/TSTN messages
   and the answerer SDP is

   <---------------- Answer

         v=0
         o=alice 3203093520 3203093524 IN IP4 otherhost.example.com
         s=Offer/Answer
         c=IN IP4 192.0.2.37
         m=audio 47190 RTP/AVP 0
         a=rtpmap:0 PCMU/8000
         m=video 53273 RTP/AVPF 98
         a=rtpmap:98 H263-1998/90000
         a=rtcp-fb:98 ccm tstr
         a=rtcp-fb:98 ccm fir





Wenger, et al.              Standards Track                    [Page 57]

RFC 5104             Codec Control Messages in AVPF        February 2008


   Example 4: The following example describes the Offer/Answer
   implications for H.271 Video Back Channel Messages (VBCMs).  The
   offerer wishes to support VBCM and the sub-messages of payloadType 1
   (one or more pictures that are entirely or partially lost) and 2 (a
   set of blocks of one picture that are entirely or partially lost).

   -------------> Offer
         v=0
         o=alice 3203093520 3203093520 IN IP4 host.example.com
         s=Offer/Answer
         c=IN IP4 192.0.2.124
         m=audio 49170 RTP/AVP 0
         a=rtpmap:0 PCMU/8000
         m=video 51372 RTP/AVPF 98
         a=rtpmap:98 H263-1998/90000
         a=rtcp-fb:98 ccm vbcm 1 2

   The answerer only wishes to support sub-messages of type 1 only

   <---------------- Answer

         v=0
         o=alice 3203093520 3203093524 IN IP4 otherhost.example.com
         s=Offer/Answer
         c=IN IP4 192.0.2.37
         m=audio 47190 RTP/AVP 0
         a=rtpmap:0 PCMU/8000
         m=video 53273 RTP/AVPF 98
         a=rtpmap:98 H263-1998/90000
         a=rtcp-fb:98 ccm vbcm 1

   So, in the above example, only VBCM indications comprised of
   "payloadType" 1 will be supported.

8.  IANA Considerations



   The new value "ccm" has been registered with IANA in the "rtcp-fb"
   Attribute Values registry located at the time of publication at:
   http://www.iana.org/assignments/sdp-parameters

      Value name:       ccm
      Long Name:        Codec Control Commands and Indications
      Reference:        RFC 5104

   A new registry "Codec Control Messages" has been created to hold
   "ccm" parameters located at time of publication at:
   http://www.iana.org/assignments/sdp-parameters




Wenger, et al.              Standards Track                    [Page 58]

RFC 5104             Codec Control Messages in AVPF        February 2008


   New registration in this registry follows the "Specification
   required" policy as defined by [RFC2434].  In addition, they are
   required to indicate any additional RTCP feedback types, such as
   "nack" and "ack".

   The initial content of the registry is the following values:

      Value name:       fir
      Long name:        Full Intra Request Command
      Usable with:      ccm
      Reference:        RFC 5104

      Value name:       tmmbr
      Long name:        Temporary Maximum Media Stream Bit Rate
      Usable with:      ccm
      Reference:        RFC 5104

      Value name:       tstr
      Long name:        Temporal Spatial Trade Off
      Usable with:      ccm
      Reference:        RFC 5104

      Value name:       vbcm
      Long name:        H.271 video back channel messages
      Usable with:      ccm
      Reference:        RFC 5104

   The following values have been registered as FMT values in the "FMT
   Values for RTPFB Payload Types" registry located at the time of
   publication at: http://www.iana.org/assignments/rtp-parameters

   RTPFB range
   Name           Long Name                         Value  Reference
   -------------- --------------------------------- -----  ---------
                  Reserved                             2   [RFC5104]
   TMMBR          Temporary Maximum Media Stream Bit   3   [RFC5104]
                  Rate Request
   TMMBN          Temporary Maximum Media Stream Bit   4   [RFC5104]
                  Rate Notification

   The following values have been registered as FMT values in the "FMT
   Values for PSFB Payload Types" registry located at the time of
   publication at: http://www.iana.org/assignments/rtp-parameters








Wenger, et al.              Standards Track                    [Page 59]

RFC 5104             Codec Control Messages in AVPF        February 2008


   PSFB range
   Name           Long Name                             Value Reference
   -------------- ---------------------------------     ----- ---------
   FIR            Full Intra Request Command              4   [RFC5104]
   TSTR           Temporal-Spatial Trade-off Request      5   [RFC5104]
   TSTN           Temporal-Spatial Trade-off Notification 6   [RFC5104]
   VBCM           Video Back Channel Message              7   [RFC5104]

9.  Contributors

   Tom Taylor has made a very significant contribution to this
   specification, for which the authors are very grateful, by helping
   rewrite the specification.  Especially the parts regarding the
   algorithm for determining bounding sets for TMMBR have benefited.

10.  Acknowledgements



   The authors would like to thank Andrea Basso, Orit Levin, and Nermeen
   Ismail for their work on the requirement and discussion document
   [Basso].

   Versions of this memo were reviewed and extensively commented on by
   Roni Even, Colin Perkins, Randell Jesup, Keith Lantz, Harikishan
   Desineni, Guido Franceschini, and others.  The authors appreciate
   these reviews.

11.  References



11.1.  Normative References



   [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.

   [RFC2119]   Bradner, S., "Key words for use in RFCs to Indicate
               Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC3550]   Schulzrinne, H.,  Casner, S., Frederick, R., and V.
               Jacobson, "RTP: A Transport Protocol for Real-Time
               Applications", STD 64, RFC 3550, July 2003.

   [RFC4566]   Handley, M., Jacobson, V., and C. Perkins, "SDP: Session
               Description Protocol", RFC 4566, July 2006.

   [RFC3264]   Rosenberg, J. and H. Schulzrinne, "An Offer/Answer Model
               with Session Description Protocol (SDP)", RFC 3264, June
               2002.



Wenger, et al.              Standards Track                    [Page 60]

RFC 5104             Codec Control Messages in AVPF        February 2008


   [RFC2434]   Narten, T. and H. Alvestrand, "Guidelines for Writing an
               IANA Considerations Section in RFCs", BCP 26, RFC 2434,
               October 1998.

   [RFC4234]   Crocker, D. and P. Overell, "Augmented BNF for Syntax
               Specifications: ABNF", RFC 4234, October 2005.

11.2.  Informative References



   [Basso]     Basso, A., Levin, O., and N. Ismail, "Requirements for
               transport of video control commands", Work in Progress,
               October 2004.

   [AVC]       Joint Video Team of ITU-T and ISO/IEC JTC 1, Draft ITU-T
               Recommendation and Final Draft International Standard of
               Joint Video Specification (ITU-T Rec. H.264 | ISO/IEC
               14496-10 AVC), Joint Video Team (JVT) of ISO/IEC MPEG and
               ITU-T VCEG, JVT-G050, March 2003.

   [H245]      ITU-T Rec. H.245, "Control protocol for multimedia
               communication", May 2006.

   [NEWPRED]   S. Fukunaga, T. Nakai, and H. Inoue, "Error Resilient
               Video Coding by Dynamic Replacing of Reference Pictures",
               in Proc. Globcom'96, vol. 3, pp. 1503 - 1508, 1996.

   [SRTP]      Baugher, M., McGrew, D., Naslund, M., Carrara, E., and K.
               Norrman, "The Secure Real-time Transport Protocol
               (SRTP)", RFC 3711, March 2004.

   [RFC2032]   Turletti, T. and C. Huitema, "RTP Payload Format for
               H.261 Video Streams", RFC 2032, October 1996.

   [SAVPF]     Ott, J. and E. Carrara, "Extended Secure RTP Profile for
               RTCP-based Feedback (RTP/SAVPF)", Work in Progress,
               November 2007.

   [RFC3525]   Groves, C., Pantaleo, M., Anderson, T., and T. Taylor,
               "Gateway Control Protocol Version 1", RFC 3525, June
               2003.

   [RFC3448]   Handley, M., Floyd, S., Padhye, J., and J. Widmer, "TCP
               Friendly Rate Control (TFRC): Protocol Specification",
               RFC 3448, January 2003.

   [H.271]     ITU-T Rec. H.271, "Video Back Channel Messages", June
               2006.




Wenger, et al.              Standards Track                    [Page 61]

RFC 5104             Codec Control Messages in AVPF        February 2008


   [RFC3890]   Westerlund, M., "A Transport Independent Bandwidth
               Modifier for the Session Description Protocol (SDP)", RFC
               3890, September 2004.

   [RFC4340]   Kohler, E., Handley, M., and S. Floyd, "Datagram
               Congestion Control Protocol (DCCP)", RFC 4340, March
               2006.

   [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.

   [RFC2198]   Perkins, C., Kouvelas, I., Hodson, O., Hardman, V.,
               Handley, M., Bolot, J., Vega-Garcia, A., and S. Fosse-
               Parisis, "RTP Payload for Redundant Audio Data", RFC
               2198, September 1997.

   [RFC4587]   Even, R., "RTP Payload Format for H.261 Video Streams",
               RFC 4587, August 2006.

   [RFC5117]   Westerlund, M. and S. Wenger, "RTP Topologies", RFC 5117,
               January 2008.

   [XML-MC]    Levin, O., Even, R., and P. Hagendorf, "XML Schema for
               Media Control", Work in Progress, November 2007.

























Wenger, et al.              Standards Track                    [Page 62]

RFC 5104             Codec Control Messages in AVPF        February 2008


Authors' Addresses



   Stephan Wenger
   Nokia Corporation
   975, Page Mill Road,
   Palo Alto,CA 94304
   USA

   Phone: +1-650-862-7368
   EMail: stewe@stewe.org


   Umesh Chandra
   Nokia Research Center
   975, Page Mill Road,
   Palo Alto,CA 94304
   USA

   Phone: +1-650-796-7502
   Email: Umesh.1.Chandra@nokia.com


   Magnus Westerlund
   Ericsson Research
   Ericsson AB
   SE-164 80 Stockholm, SWEDEN

   Phone: +46 8 7190000
   EMail: magnus.westerlund@ericsson.com


   Bo Burman
   Ericsson Research
   Ericsson AB
   SE-164 80 Stockholm, SWEDEN

   Phone: +46 8 7190000
   EMail: bo.burman@ericsson.com













Wenger, et al.              Standards Track                    [Page 63]

RFC 5104             Codec Control Messages in AVPF        February 2008


Full Copyright Statement



   Copyright (C) The IETF Trust (2008).

   This document is subject to the rights, licenses and restrictions
   contained in BCP 78, and except as set forth therein, the authors
   retain all their rights.

   This document and the information contained herein are provided on an
   "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
   OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND
   THE INTERNET ENGINEERING TASK FORCE DISCLAIM 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.

Intellectual Property



   The IETF takes no position regarding the validity or scope of any
   Intellectual Property Rights or other rights that might be claimed to
   pertain to the implementation or use of the technology described in
   this document or the extent to which any license under such rights
   might or might not be available; nor does it represent that it has
   made any independent effort to identify any such rights.  Information
   on the procedures with respect to rights in RFC documents can be
   found in BCP 78 and BCP 79.

   Copies of IPR disclosures made to the IETF Secretariat and any
   assurances of licenses to be made available, or the result of an
   attempt made to obtain a general license or permission for the use of
   such proprietary rights by implementers or users of this
   specification can be obtained from the IETF on-line IPR repository at
   http://www.ietf.org/ipr.

   The IETF invites any interested party to bring to its attention any
   copyrights, patents or patent applications, or other proprietary
   rights that may cover technology that may be required to implement
   this standard.  Please address the information to the IETF at
   ietf-ipr@ietf.org.












Wenger, et al.              Standards Track                    [Page 64]