Internet Engineering Task Force (IETF)                     M. Westerlund
Request for Comments: 8872                                     B. Burman
Category: Informational                                         Ericsson
ISSN: 2070-1721                                               C. Perkins
                                                  University of Glasgow
                                                          H. Alvestrand
                                                                 Google
                                                                R. Even
                                                           January 2021


   Guidelines for Using the Multiplexing Features of RTP to Support
                        Multiple Media Streams

Abstract

  The Real-time Transport Protocol (RTP) is a flexible protocol that
  can be used in a wide range of applications, networks, and system
  topologies.  That flexibility makes for wide applicability but can
  complicate the application design process.  One particular design
  question that has received much attention is how to support multiple
  media streams in RTP.  This memo discusses the available options and
  design trade-offs, and provides guidelines on how to use the
  multiplexing features of RTP to support multiple media streams.

Status of This Memo

  This document is not an Internet Standards Track specification; it is
  published for informational purposes.

  This document is a product of the Internet Engineering Task Force
  (IETF).  It represents the consensus of the IETF community.  It has
  received public review and has been approved for publication by the
  Internet Engineering Steering Group (IESG).  Not all documents
  approved by the IESG are candidates for any level of Internet
  Standard; see Section 2 of RFC 7841.

  Information about the current status of this document, any errata,
  and how to provide feedback on it may be obtained at
  https://www.rfc-editor.org/info/rfc8872.

Copyright Notice

  Copyright (c) 2021 IETF Trust and the persons identified as the
  document authors.  All rights reserved.

  This document is subject to BCP 78 and the IETF Trust's Legal
  Provisions Relating to IETF Documents
  (https://trustee.ietf.org/license-info) in effect on the date of
  publication of this document.  Please review these documents
  carefully, as they describe your rights and restrictions with respect
  to this document.  Code Components extracted from this document must
  include Simplified BSD License text as described in Section 4.e of
  the Trust Legal Provisions and are provided without warranty as
  described in the Simplified BSD License.

Table of Contents

  1.  Introduction
  2.  Definitions
    2.1.  Terminology
    2.2.  Focus of This Document
  3.  RTP Multiplexing Overview
    3.1.  Reasons for Multiplexing and Grouping RTP Streams
    3.2.  RTP Multiplexing Points
      3.2.1.  RTP Session
      3.2.2.  Synchronization Source (SSRC)
      3.2.3.  Contributing Source (CSRC)
      3.2.4.  RTP Payload Type
    3.3.  Issues Related to RTP Topologies
    3.4.  Issues Related to RTP and RTCP
      3.4.1.  The RTP Specification
      3.4.2.  Multiple SSRCs in a Session
      3.4.3.  Binding Related Sources
      3.4.4.  Forward Error Correction
  4.  Considerations for RTP Multiplexing
    4.1.  Interworking Considerations
      4.1.1.  Application Interworking
      4.1.2.  RTP Translator Interworking
      4.1.3.  Gateway Interworking
      4.1.4.  Legacy Considerations for Multiple SSRCs
    4.2.  Network Considerations
      4.2.1.  Quality of Service
      4.2.2.  NAT and Firewall Traversal
      4.2.3.  Multicast
    4.3.  Security and Key-Management Considerations
      4.3.1.  Security Context Scope
      4.3.2.  Key Management for Multi-party Sessions
      4.3.3.  Complexity Implications
  5.  RTP Multiplexing Design Choices
    5.1.  Multiple Media Types in One Session
    5.2.  Multiple SSRCs of the Same Media Type
    5.3.  Multiple Sessions for One Media Type
    5.4.  Single SSRC per Endpoint
    5.5.  Summary
  6.  Guidelines
  7.  IANA Considerations
  8.  Security Considerations
  9.  References
    9.1.  Normative References
    9.2.  Informative References
  Appendix A.  Dismissing Payload Type Multiplexing
  Appendix B.  Signaling Considerations
    B.1.  Session-Oriented Properties
    B.2.  SDP Prevents Multiple Media Types
    B.3.  Signaling RTP Stream Usage
  Acknowledgments
  Contributors
  Authors' Addresses

1.  Introduction

  The Real-time Transport Protocol (RTP) [RFC3550] is a commonly used
  protocol for real-time media transport.  It is a protocol that
  provides great flexibility and can support a large set of different
  applications.  From the beginning, RTP was designed for multiple
  participants in a communication session.  It supports many topology
  paradigms and usages, as defined in [RFC7667].  RTP has several
  multiplexing points designed for different purposes; these points
  enable support of multiple RTP streams and switching between
  different encoding or packetization techniques for the media.  By
  using multiple RTP sessions, sets of RTP streams can be structured
  for efficient processing or identification.  Thus, to meet an
  application's needs, an RTP application designer needs to understand
  how best to use the RTP session, the RTP stream identifier
  (synchronization source (SSRC)), and the RTP payload type.

  There has been increased interest in more-advanced usage of RTP.  For
  example, multiple RTP streams can be used when a single endpoint has
  multiple media sources (like multiple cameras or microphones) from
  which streams of media need to be sent simultaneously.  Consequently,
  questions are raised regarding the most appropriate RTP usage.  The
  limitations in some implementations, RTP/RTCP extensions, and
  signaling have also been exposed.  This document aims to clarify the
  usefulness of some functionalities in RTP that, hopefully, will
  result in future implementations that are more complete.

  The purpose of this document is to provide clear information about
  the possibilities of RTP when it comes to multiplexing.  The RTP
  application designer needs to understand the implications arising
  from a particular usage of the RTP multiplexing points.  This
  document provides some guidelines and recommends against some usages
  as being unsuitable, in general or for particular purposes.

  This document starts with some definitions and then goes into
  existing RTP functionalities around multiplexing.  Both the desired
  behavior and the implications of a particular behavior depend on
  which topologies are used; therefore, this topic requires some
  consideration.  We then discuss some choices regarding multiplexing
  behavior and the impacts of those choices.  Some designs of RTP usage
  are also discussed.  Finally, some guidelines and examples are
  provided.

2.  Definitions

2.1.  Terminology

  The definitions in Section 3 of [RFC3550] are referenced normatively.

  The taxonomy defined in [RFC7656] is referenced normatively.

  The following terms and abbreviations are used in this document:

  Multi-party:
     Communication that includes multiple endpoints.  In this document,
     "multi-party" will be used to refer to scenarios where more than
     two endpoints communicate.

  Multiplexing:
     An operation that takes multiple entities as input, aggregating
     them onto some common resource while keeping the individual
     entities addressable such that they can later be fully and
     unambiguously separated (demultiplexed) again.

  RTP Receiver:
     An endpoint or middlebox receiving RTP streams and RTCP messages.
     It uses at least one SSRC to send RTCP messages.  An RTP receiver
     may also be an RTP sender.

  RTP Sender:
     An endpoint sending one or more RTP streams but also sending RTCP
     messages.

  RTP Session Group:
     One or more RTP sessions that are used together to perform some
     function.  Examples include multiple RTP sessions used to carry
     different layers of a layered encoding.  In an RTP Session Group,
     CNAMEs are assumed to be valid across all RTP sessions and
     designate synchronization contexts that can cross RTP sessions;
     i.e., SSRCs that map to a common CNAME can be assumed to have RTCP
     Sender Report (SR) timing information derived from a common clock
     such that they can be synchronized for playout.

  Signaling:
     The process of configuring endpoints to participate in one or more
     RTP sessions.

     |  Note: The above definitions of "RTP receiver" and "RTP sender"
     |  are consistent with the usage in [RFC3550].

2.2.  Focus of This Document

  This document is focused on issues that affect RTP.  Thus, issues
  that involve signaling protocols -- such as whether SIP [RFC3261],
  Jingle [JINGLE], or some other protocol is in use for session
  configuration; the particular syntaxes used to define RTP session
  properties; or the constraints imposed by particular choices in the
  signaling protocols -- are mentioned only as examples in order to
  describe the RTP issues more precisely.

  This document assumes that the applications will use RTCP.  While
  there are applications that don't send RTCP, they do not conform to
  the RTP specification and thus can be regarded as reusing the RTP
  packet format but not implementing RTP.

3.  RTP Multiplexing Overview

3.1.  Reasons for Multiplexing and Grouping RTP Streams

  There are several reasons why an endpoint might choose to send
  multiple media streams.  In the discussion below, please keep in mind
  that the reasons for having multiple RTP streams vary and include,
  but are not limited to, the following:

  *  There might be multiple media sources.

  *  Multiple RTP streams might be needed to represent one media
     source, for example:

     -  To carry different layers of a scalable encoding of a media
        source

     -  Alternative encodings during simulcast, using different codecs
        for the same audio stream

     -  Alternative formats during simulcast, multiple resolutions of
        the same video stream

  *  A retransmission stream might repeat some parts of the content of
     another RTP stream.

  *  A Forward Error Correction (FEC) stream might provide material
     that can be used to repair another RTP stream.

  For each of these reasons, it is necessary to decide whether each
  additional RTP stream is sent within the same RTP session as the
  other RTP streams or it is necessary to use additional RTP sessions
  to group the RTP streams.  For a combination of reasons, the suitable
  choice for one situation might not be the suitable choice for another
  situation.  The choice is easiest when multiplexing multiple media
  sources of the same media type.  However, all reasons warrant
  discussion and clarification regarding how to deal with them.  As the
  discussion below will show, a single solution does not suit all
  purposes.  To utilize RTP well and as efficiently as possible, both
  are needed.  The real issue is knowing when to create multiple RTP
  sessions versus when to send multiple RTP streams in a single RTP
  session.

3.2.  RTP Multiplexing Points

  This section describes the multiplexing points present in RTP that
  can be used to distinguish RTP streams and groups of RTP streams.
  Figure 1 outlines the process of demultiplexing incoming RTP streams,
  starting with one or more sockets representing the reception of one
  or more transport flows, e.g., based on the UDP destination port.  It
  also demultiplexes RTP/RTCP from any other protocols, such as Session
  Traversal Utilities for NAT (STUN) [RFC5389] and DTLS-SRTP [RFC5764]
  on the same transport as described in [RFC7983].  The Processing and
  Buffering (PB) step in Figure 1 terminates RTP/RTCP and prepares the
  RTP payload for input to the decoder.

                     |   |   |
                     |   |   | packets
          +--        v   v   v
          |        +------------+
          |        |  Socket(s) |   Transport Protocol Demultiplexing
          |        +------------+
          |            ||  ||
     RTP  |       RTP/ ||  |+-----> DTLS (SRTP keying, SCTP, etc.)
  Session |       RTCP ||  +------> STUN (multiplexed using same port)
          +--          ||
          +--          ||
          |      ++(split by SSRC)-++---> Identify SSRC collision
          |      ||    ||    ||    ||
          | (associate with signaling by MID/RID)
          |      vv    vv    vv    vv
    RTP   |     +--+  +--+  +--+  +--+ Jitter buffer,
  Streams |     |PB|  |PB|  |PB|  |PB| process RTCP, etc.
          |     +--+  +--+  +--+  +--+
          +--     |    |      |    |
            (select decoder based on payload type (PT))
          +--     |   /       |  /
          |       +-----+     | /
          |         /   |     |/
  Payload |        v    v     v
  Formats |     +---+ +---+ +---+
          |     |Dec| |Dec| |Dec| Decoders
          |     +---+ +---+ +---+
          +--

                   Figure 1: RTP Demultiplexing Process

3.2.1.  RTP Session

  An RTP session is the highest semantic layer in RTP and represents an
  association between a group of communicating endpoints.  RTP does not
  contain a session identifier, yet different RTP sessions must be
  possible to identify both across a set of different endpoints and
  from the perspective of a single endpoint.

  For RTP session separation across endpoints, the set of participants
  that form an RTP session is defined as those that share a single SSRC
  space [RFC3550].  That is, if a group of participants are each aware
  of the SSRC identifiers belonging to the other participants, then
  those participants are in a single RTP session.  A participant can
  become aware of an SSRC identifier by receiving an RTP packet
  containing the identifier in the SSRC field or contributing source
  (CSRC) list, by receiving an RTCP packet listing it in an SSRC field,
  or through signaling (e.g., the Session Description Protocol (SDP)
  [RFC4566] "a=ssrc:" attribute [RFC5576]).  Thus, the scope of an RTP
  session is determined by the participants' network interconnection
  topology, in combination with RTP and RTCP forwarding strategies
  deployed by the endpoints and any middleboxes, and by the signaling.

  For RTP session separation within a single endpoint, RTP relies on
  the underlying transport layer and the signaling to identify RTP
  sessions in a manner that is meaningful to the application.  A single
  endpoint can have one or more transport flows for the same RTP
  session, and a single RTP session can span multiple transport-layer
  flows even if all endpoints use a single transport-layer flow per
  endpoint for that RTP session.  The signaling layer might give RTP
  sessions an explicit identifier, or the identification might be
  implicit based on the addresses and ports used.  Accordingly, a
  single RTP session can have multiple associated identifiers, explicit
  and implicit, belonging to different contexts.  For example, when
  running RTP on top of UDP/IP, an endpoint can identify and delimit an
  RTP session from other RTP sessions by their UDP source and
  destination IP addresses and their UDP port numbers.  A single RTP
  session can be using multiple IP/UDP flows for receiving and/or
  sending RTP packets to other endpoints or middleboxes, even if the
  endpoint does not have multiple IP addresses.  Using multiple IP
  addresses only makes it more likely that multiple IP/UDP flows will
  be required.  Another example is SDP media descriptions (the "m="
  line and the subsequent associated lines) that signal the transport
  flow and RTP session configuration for the endpoint's part of the RTP
  session.  The SDP grouping framework [RFC5888] allows labeling of the
  media descriptions to be used so that RTP Session Groups can be
  created.  Through the use of "Negotiating Media Multiplexing Using
  the Session Description Protocol (SDP)" [RFC8843], multiple media
  descriptions become part of a common RTP session where each media
  description represents the RTP streams sent or received for a media
  source.

  RTP makes no normative statements about the relationship between
  different RTP sessions; however, applications that use more than one
  RTP session need to understand how the different RTP sessions that
  they create relate to one another.

3.2.2.  Synchronization Source (SSRC)

  An SSRC identifies a source of an RTP stream, or an RTP receiver when
  sending RTCP.  Every endpoint has at least one SSRC identifier, even
  if it does not send RTP packets.  RTP endpoints that are only RTP
  receivers still send RTCP and use their SSRC identifiers in the RTCP
  packets they send.  An endpoint can have multiple SSRC identifiers if
  it sends multiple RTP streams.  Endpoints that function as both RTP
  sender and RTP receiver use the same SSRC(s) in both roles.

  The SSRC is a 32-bit identifier.  It is present in every RTP and RTCP
  packet header and in the payload of some RTCP packet types.  It can
  also be present in SDP signaling.  Unless presignaled, e.g., using
  the SDP "a=ssrc:" attribute [RFC5576], the SSRC is chosen at random.
  It is not dependent on the network address of the endpoint and is
  intended to be unique within an RTP session.  SSRC collisions can
  occur and are handled as specified in [RFC3550] and [RFC5576],
  resulting in the SSRC of the colliding RTP streams or receivers
  changing.  An endpoint that changes its network transport address
  during a session has to choose a new SSRC identifier to avoid being
  interpreted as a looped source, unless a mechanism providing a
  virtual transport (such as Interactive Connectivity Establishment
  (ICE) [RFC8445]) abstracts the changes.

  SSRC identifiers that belong to the same synchronization context
  (i.e., that represent RTP streams that can be synchronized using
  information in RTCP SR packets) use identical CNAME chunks in
  corresponding RTCP source description (SDES) packets.  SDP signaling
  can also be used to provide explicit SSRC grouping [RFC5576].

  In some cases, the same SSRC identifier value is used to relate
  streams in two different RTP sessions, such as in RTP retransmission
  [RFC4588].  This is to be avoided, since there is no guarantee that
  SSRC values are unique across RTP sessions.  In the case of RTP
  retransmission [RFC4588], it is recommended to use explicit binding
  of the source RTP stream and the redundancy stream, e.g., using the
  RepairedRtpStreamId RTCP SDES item [RFC8852].  The
  RepairedRtpStreamId is a rather recent mechanism, so one cannot
  expect older applications to follow this recommendation.

  Note that the RTP sequence number and RTP timestamp are scoped by the
  SSRC and are thus specific per RTP stream.

  Different types of entities use an SSRC to identify themselves, as
  follows:

  *  A real media source uses the SSRC to identify a "physical" media
     source.

  *  A conceptual media source uses the SSRC to identify the result of
     applying some filtering function in a network node -- for example,
     a filtering function in an RTP mixer that provides the most active
     speaker based on some criteria, or a mix representing a set of
     other sources.

  *  An RTP receiver uses the SSRC to identify itself as the source of
     its RTCP reports.

  An endpoint that generates more than one media type, e.g., a
  conference participant sending both audio and video, need not (and,
  indeed, should not) use the same SSRC value across RTP sessions.
  Using RTCP compound packets containing the CNAME SDES item is the
  designated method for binding an SSRC to a CNAME, effectively cross-
  correlating SSRCs within and between RTP sessions as coming from the
  same endpoint.  The main property attributed to SSRCs associated with
  the same CNAME is that they are from a particular synchronization
  context and can be synchronized at playback.

  An RTP receiver receiving a previously unseen SSRC value will
  interpret it as a new source.  It might in fact be a previously
  existing source that had to change its SSRC number due to an SSRC
  conflict.  Using the media identification (MID) extension [RFC8843]
  helps to identify which media source the new SSRC represents, and
  using the restriction identifier (RID) extension [RFC8851] helps to
  identify what encoding or redundancy stream it represents, even
  though the SSRC changed.  However, the originator of the previous
  SSRC ought to have ended the conflicting source by sending an RTCP
  BYE for it prior to starting to send with the new SSRC, making the
  new SSRC a new source.

3.2.3.  Contributing Source (CSRC)

  The CSRC is not a separate identifier.  Rather, an SSRC identifier is
  listed as a CSRC in the RTP header of a packet generated by an RTP
  mixer or video Multipoint Control Unit (MCU) / switch, if the
  corresponding SSRC was in the header of one of the packets that
  contributed to the output.

  It is not possible, in general, to extract media represented by an
  individual CSRC, since it is typically the result of a media merge
  (e.g., mix) operation on the individual media streams corresponding
  to the CSRC identifiers.  The exception is the case where only a
  single CSRC is indicated, as this represents the forwarding of an RTP
  stream that might have been modified.  The RTP header extension ("A
  Real-time Transport Protocol (RTP) Header Extension for
  Mixer-to-Client Audio Level Indication" [RFC6465]) expands on the
  receiver's information about a packet with a CSRC list.  Due to these
  restrictions, a CSRC will not be considered a fully qualified
  multiplexing point and will be disregarded in the rest of this
  document.

3.2.4.  RTP Payload Type

  Each RTP stream utilizes one or more RTP payload formats.  An RTP
  payload format describes how the output of a particular media codec
  is framed and encoded into RTP packets.  The payload format is
  identified by the payload type (PT) field in the RTP packet header.
  The combination of SSRC and PT therefore identifies a specific RTP
  stream in a specific encoding format.  The format definition can be
  taken from [RFC3551] for statically allocated payload types but ought
  to be explicitly defined in signaling, such as SDP, for both static
  and dynamic payload types.  The term "format" here includes those
  aspects described by out-of-band signaling means; in SDP, the term
  "format" includes media type, RTP timestamp sampling rate, codec,
  codec configuration, payload format configurations, and various
  robustness mechanisms such as redundant encodings [RFC2198].

  The RTP payload type is scoped by the sending endpoint within an RTP
  session.  PT has the same meaning across all RTP streams in an RTP
  session.  All SSRCs sent from a single endpoint share the same
  payload type definitions.  The RTP payload type is designed such that
  only a single payload type is valid at any instant in time in the RTP
  stream's timestamp timeline, effectively time-multiplexing different
  payload types if any change occurs.  The payload type can change on a
  per-packet basis for an SSRC -- for example, a speech codec making
  use of generic comfort noise [RFC3389].  If there is a true need to
  send multiple payload types for the same SSRC that are valid for the
  same instant, then redundant encodings [RFC2198] can be used.
  Several additional constraints, other than those mentioned above,
  need to be met to enable this usage, one of which is that the
  combined payload sizes of the different payload types ought not
  exceed the transport MTU.

  Other aspects of using the RTP payload format are described in "How
  to Write an RTP Payload Format" [RFC8088].

  The payload type is not a multiplexing point at the RTP layer (see
  Appendix A for a detailed discussion of why using the payload type as
  an RTP multiplexing point does not work).  The RTP payload type is,
  however, used to determine how to consume and decode an RTP stream.
  The RTP payload type number is sometimes used to associate an RTP
  stream with the signaling, which in general requires that unique RTP
  payload type numbers be used in each context.  Using MID, e.g., when
  bundling "m=" sections [RFC8843], can replace the payload type as a
  signaling association, and unique RTP payload types are then no
  longer required for that purpose.

3.3.  Issues Related to RTP Topologies

  The impact of how RTP multiplexing is performed will in general vary
  with how the RTP session participants are interconnected, as
  described in "RTP Topologies" [RFC7667].

  Even the most basic use case -- "Topo-Point-to-Point" as described in
  [RFC7667] -- raises a number of considerations, which are discussed
  in detail in the following sections.  They range over such aspects as
  the following:

  *  Does my communication peer support RTP as defined with multiple
     SSRCs per RTP session?

  *  Do I need network differentiation in the form of QoS
     (Section 4.2.1)?

  *  Can the application more easily process and handle the media
     streams if they are in different RTP sessions?

  *  Do I need to use additional RTP streams for RTP retransmission or
     FEC?

  For some point-to-multipoint topologies (e.g., Topo-ASM and Topo-SSM
  [RFC7667]), multicast is used to interconnect the session
  participants.  Special considerations (documented in Section 4.2.3)
  are then needed, as multicast is a one-to-many distribution system.

  Sometimes, an RTP communication session can end up in a situation
  where the communicating peers are not compatible, for various
  reasons:

  *  No common media codec for a media type, thus requiring
     transcoding.

  *  Different support for multiple RTP streams and RTP sessions.

  *  Usage of different media transport protocols (i.e., one peer uses
     RTP, but the other peer uses a different transport protocol).

  *  Usage of different transport protocols, e.g., UDP, the Datagram
     Congestion Control Protocol (DCCP), or TCP.

  *  Different security solutions (e.g., IPsec, TLS, DTLS, or the
     Secure Real-time Transport Protocol (SRTP)) with different keying
     mechanisms.

  These compatibility issues can often be resolved by the inclusion of
  a translator between the two peers -- the Topo-PtP-Translator, as
  described in [RFC7667].  The translator's main purpose is to make the
  peers look compatible to each other.  There can also be reasons other
  than compatibility for inserting a translator in the form of a
  middlebox or gateway -- for example, a need to monitor the RTP
  streams.  Beware that changing the stream transport characteristics
  in the translator can require a thorough understanding of aspects
  ranging from congestion control and media-level adaptations to
  application-layer semantics.

  Within the uses enabled by the RTP standard, the point-to-point
  topology can contain one or more RTP sessions with one or more media
  sources per session, each having one or more RTP streams per media
  source.

3.4.  Issues Related to RTP and RTCP

  Using multiple RTP streams is a well-supported feature of RTP.
  However, for most implementers or people writing RTP/RTCP
  applications or extensions attempting to apply multiple streams, it
  can be unclear when it is most appropriate to add an additional RTP
  stream in an existing RTP session and when it is better to use
  multiple RTP sessions.  This section discusses the various
  considerations that need to be taken into account.

3.4.1.  The RTP Specification

  RFC 3550 contains some recommendations and a numbered list
  (Section 5.2 of [RFC3550]) of five arguments regarding different
  aspects of RTP multiplexing.  Please review Section 5.2 of [RFC3550].
  Five important aspects are quoted below.

  1.  |  If, say, two audio streams shared the same RTP session and the
      |  same SSRC value, and one were to change encodings and thus
      |  acquire a different RTP payload type, there would be no
      |  general way of identifying which stream had changed encodings.

      This argument advocates the use of different SSRCs for each
      individual RTP stream, as this is fundamental to RTP operation.

  2.  |  An SSRC is defined to identify a single timing and sequence
      |  number space.  Interleaving multiple payload types would
      |  require different timing spaces if the media clock rates
      |  differ and would require different sequence number spaces to
      |  tell which payload type suffered packet loss.

      This argument advocates against demultiplexing RTP streams within
      a session based only on their RTP payload type numbers; it still
      stands, as can be seen by the extensive list of issues discussed
      in Appendix A.

  3.  |  The RTCP sender and receiver reports (see Section 6.4) can
      |  only describe one timing and sequence number space per SSRC
      |  and do not carry a payload type field.

      This argument is yet another argument against payload type
      multiplexing.

  4.  |  An RTP mixer would not be able to combine interleaved streams
      |  of incompatible media into one stream.

      This argument advocates against multiplexing RTP packets that
      require different handling into the same session.  In most cases,
      the RTP mixer must embed application logic to handle streams; the
      separation of streams according to stream type is just another
      piece of application logic, which might or might not be
      appropriate for a particular application.  One type of
      application that can mix different media sources blindly is the
      audio-only telephone bridge, although the ability to do that
      comes from the well-defined scenario that is aided by the use of
      a single media type, even though individual streams may use
      incompatible codec types; most other types of applications need
      application-specific logic to perform the mix correctly.

  5.  |  Carrying multiple media in one RTP session precludes: the use
      |  of different network paths or network resource allocations if
      |  appropriate; reception of a subset of the media if desired,
      |  for example just audio if video would exceed the available
      |  bandwidth; and receiver implementations that use separate
      |  processes for the different media, whereas using separate RTP
      |  sessions permits either single- or multiple-process
      |  implementations.

      This argument discusses network aspects that are described in
      Section 4.2.  It also goes into aspects of implementation, like
      split component terminals (see Section 3.10 of [RFC7667]) --
      endpoints where different processes or interconnected devices
      handle different aspects of the whole multimedia session.

  To summarize, RFC 3550's view on multiplexing is to use unique SSRCs
  for anything that is its own media/packet stream and use different
  RTP sessions for media streams that don't share a media type.  This
  document supports the first point; it is very valid.  The latter
  needs further discussion, as imposing a single solution on all usages
  of RTP is inappropriate.  "Sending Multiple Types of Media in a
  Single RTP Session" [RFC8860] updates RFC 3550 to allow multiple
  media types in an RTP session and provides a detailed analysis of the
  potential benefits and issues related to having multiple media types
  in the same RTP session.  Thus, [RFC8860] provides a wider scope for
  an RTP session and considers multiple media types in one RTP session
  as a possible choice for the RTP application designer.

3.4.2.  Multiple SSRCs in a Session

  Using multiple SSRCs at one endpoint in an RTP session requires that
  some unclear aspects of the RTP specification be resolved.  These
  items could potentially lead to some interoperability issues as well
  as some potential significant inefficiencies, as further discussed in
  "Sending Multiple RTP Streams in a Single RTP Session" [RFC8108].  An
  RTP application designer should consider these issues and the
  application's possible impact caused by a lack of appropriate RTP
  handling or optimization in the peer endpoints.

  Using multiple RTP sessions can potentially mitigate application
  issues caused by multiple SSRCs in an RTP session.

3.4.3.  Binding Related Sources

  A common problem in a number of various RTP extensions has been how
  to bind related RTP streams together.  This issue is common to both
  using additional SSRCs and multiple RTP sessions.

  The solutions can be divided into a few groups:

  *  RTP/RTCP based

  *  Signaling based, e.g., SDP

  *  Grouping related RTP sessions

  *  Grouping SSRCs within an RTP session

  Most solutions are explicit, but some implicit methods have also been
  applied to the problem.

  The SDP-based signaling solutions are:

  SDP media description grouping:
     The SDP grouping framework [RFC5888] uses various semantics to
     group any number of media descriptions.  SDP media description
     grouping has primarily been used to group RTP sessions, but in
     combination with [RFC8843], it can also group multiple media
     descriptions within a single RTP session.

  SDP media multiplexing:
     "Negotiating Media Multiplexing Using the Session Description
     Protocol (SDP)" [RFC8843] uses information taken from both SDP and
     RTCP to associate RTP streams to SDP media descriptions.  This
     allows both SDP and RTCP to group RTP streams belonging to an SDP
     media description and group multiple SDP media descriptions into a
     single RTP session.

  SDP SSRC grouping:
     "Source-Specific Media Attributes in the Session Description
     Protocol (SDP)" [RFC5576] includes a solution for grouping SSRCs
     in the same way that the grouping framework groups media
     descriptions.

  The above grouping constructs support many use cases.  Those
  solutions have shortcomings in cases where the session's dynamic
  properties are such that it is difficult or a drain on resources to
  keep the list of related SSRCs up to date.

  One RTP/RTCP-based grouping solution is to use the RTCP SDES CNAME to
  bind related RTP streams to an endpoint or a synchronization context.
  For applications with a single RTP stream per type (media, source, or
  redundancy stream), the CNAME is sufficient for that purpose,
  independent of whether one or more RTP sessions are used.  However,
  some applications choose not to use a CNAME because of perceived
  complexity or a desire not to implement RTCP and instead use the same
  SSRC value to bind related RTP streams across multiple RTP sessions.
  RTP retransmission [RFC4588], when configured to use multiple RTP
  sessions, and generic FEC [RFC5109] both use the CNAME method to
  relate the RTP streams, which may work but might have some downsides
  in RTP sessions with many participating SSRCs.  It is not recommended
  to use identical SSRC values across RTP sessions to relate RTP
  streams; when an SSRC collision occurs, this will force a change of
  that SSRC in all RTP sessions and will thus resynchronize all of the
  streams instead of only the single media stream experiencing the
  collision.

  Another method for implicitly binding SSRCs is used by RTP
  retransmission [RFC4588] when using the same RTP session as the
  source RTP stream for retransmissions.  A receiver that is missing a
  packet issues an RTP retransmission request and then awaits a new
  SSRC carrying the RTP retransmission payload, where that SSRC is from
  the same CNAME.  This limits a requester to having only one
  outstanding retransmission request on any new SSRCs per endpoint.

  "RTP Payload Format Restrictions" [RFC8851] provides an RTP/RTCP-
  based mechanism to unambiguously identify the RTP streams within an
  RTP session and restrict the streams' payload format parameters in a
  codec-agnostic way beyond what is provided with the regular payload
  types.  The mapping is done by specifying an "a=rid" value in the SDP
  offer/answer signaling and having the corresponding RtpStreamId value
  as an SDES item and an RTP header extension [RFC8852].  The RID
  solution also includes a solution for binding redundancy RTP streams
  to their original source RTP streams, given that those streams use
  RID identifiers.  The redundancy stream uses the RepairedRtpStreamId
  SDES item and RTP header extension to declare the RtpStreamId value
  of the source stream to create the binding.

  Experience has shown that an explicit binding between the RTP
  streams, agnostic of SSRC values, behaves well.  That way, solutions
  using multiple RTP streams in a single RTP session and in multiple
  RTP sessions will use the same type of binding.

3.4.4.  Forward Error Correction

  There exist a number of FEC-based schemes designed to mitigate packet
  loss in the original streams.  Most of the FEC schemes protect a
  single source flow.  This protection is achieved by transmitting a
  certain amount of redundant information that is encoded such that it
  can repair one or more instances of packet loss over the set of
  packets the redundant information protects.  This sequence of
  redundant information needs to be transmitted as its own media stream
  or, in some cases, instead of the original media stream.  Thus, many
  of these schemes create a need for binding related flows, as
  discussed above.  Looking at the history of these schemes, there are
  schemes using multiple SSRCs and schemes using multiple RTP sessions,
  and some schemes that support both modes of operation.

  Using multiple RTP sessions supports the case where some set of
  receivers might not be able to utilize the FEC information.  By
  placing it in a separate RTP session and if separating RTP sessions
  at the transport level, FEC can easily be ignored at the transport
  level, without considering any RTP-layer information.

  In usages involving multicast, sending FEC information in a separate
  multicast group allows for similar flexibility.  This is especially
  useful when receivers see heterogeneous packet loss rates.  A
  receiver can decide, based on measurement of experienced packet loss
  rates, whether to join a multicast group with suitable FEC data
  repair capabilities.

4.  Considerations for RTP Multiplexing

4.1.  Interworking Considerations

  There are several different kinds of interworking, and this section
  discusses two: interworking directly between different applications
  and the interworking of applications through an RTP translator.  The
  discussion includes the implications of potentially different RTP
  multiplexing point choices and limitations that have to be considered
  when working with some legacy applications.

4.1.1.  Application Interworking

  It is not uncommon that applications or services of similar but not
  identical usage, especially those intended for interactive
  communication, encounter a situation where one wants to interconnect
  two or more of these applications.

  In these cases, one ends up in a situation where one might use a
  gateway to interconnect applications.  This gateway must then either
  change the multiplexing structure or adhere to the respective
  limitations in each application.

  There are two fundamental approaches to building a gateway: using RTP
  translator interworking (RTP bridging), where the gateway acts as an
  RTP translator with the two interconnected applications being members
  of the same RTP session; or using gateway interworking
  (Section 4.1.3) with RTP termination, where there are independent RTP
  sessions between each interconnected application and the gateway.

  For interworking to be feasible, any security solution in use needs
  to be compatible and capable of exchanging keys with either the peer
  or the gateway under the trust model being used.  Secondly, the
  applications need to use media streams in a way that makes sense in
  both applications.

4.1.2.  RTP Translator Interworking

  From an RTP perspective, the RTP translator approach could work if
  all the applications are using the same codecs with the same payload
  types, have made the same multiplexing choices, and have the same
  capabilities regarding the number of simultaneous RTP streams
  combined with the same set of RTP/RTCP extensions being supported.
  Unfortunately, this might not always be true.

  When a gateway is implemented via an RTP translator, an important
  consideration is if the two applications being interconnected need to
  use the same approach to multiplexing.  If one side is using RTP
  session multiplexing and the other is using SSRC multiplexing with
  BUNDLE [RFC8843], it may be possible for the RTP translator to map
  the RTP streams between both sides using some method, e.g., based on
  the number and order of SDP "m=" lines from each side.  There are
  also challenges related to SSRC collision handling, since, unless
  SSRC translation is applied on the RTP translator, there may be a
  collision on the SSRC multiplexing side that the RTP session
  multiplexing side will not be aware of.  Furthermore, if one of the
  applications is capable of working in several modes (such as being
  able to use additional RTP streams in one RTP session or multiple RTP
  sessions at will) and the other one is not, successful
  interconnection depends on locking the more flexible application into
  the operating mode where interconnection can be successful, even if
  none of the participants are using the less flexible application when
  the RTP sessions are being created.

4.1.3.  Gateway Interworking

  When one terminates RTP sessions at the gateway, there are certain
  tasks that the gateway has to carry out:

  *  Generating appropriate RTCP reports for all RTP streams (possibly
     based on incoming RTCP reports) originating from SSRCs controlled
     by the gateway.

  *  Handling SSRC collision resolution in each application's RTP
     sessions.

  *  Signaling, choosing, and policing appropriate bitrates for each
     session.

  For applications that use any security mechanism, e.g., in the form
  of SRTP, the gateway needs to be able to decrypt and verify source
  integrity of the incoming packets and then re-encrypt, integrity
  protect, and sign the packets as the peer in the other application's
  security context.  This is necessary even if all that's needed is a
  simple remapping of SSRC numbers.  If this is done, the gateway also
  needs to be a member of the security contexts of both sides and thus
  a trusted entity.

  The gateway might also need to apply transcoding (for incompatible
  codec types), media-level adaptations that cannot be solved through
  media negotiation (such as rescaling for incompatible video size
  requirements), suppression of content that is known not to be handled
  in the destination application, or the addition or removal of
  redundancy coding or scalability layers to fit the needs of the
  destination domain.

  From the above, we can see that the gateway needs to have an intimate
  knowledge of the application requirements; a gateway is by its nature
  application specific and not a commodity product.

  These gateways might therefore potentially block application
  evolution by blocking RTP and RTCP extensions that the applications
  have been extended with but that are unknown to the gateway.

  If one uses a security mechanism like SRTP, the gateway and the
  necessary trust in it by the peers pose an additional risk to
  communication security.  The gateway also incurs additional
  complexities in the form of the decrypt-encrypt cycles needed for
  each forwarded packet.  SRTP, due to its keying structure, also
  requires that each RTP session need different master keys, as the use
  of the same key in two RTP sessions can, for some ciphers, result in
  a reuse of a one-time pad that completely breaks the confidentiality
  of the packets.

4.1.4.  Legacy Considerations for Multiple SSRCs

  Historically, the most common RTP use cases have been point-to-point
  Voice over IP (VoIP) or streaming applications, commonly with no more
  than one media source per endpoint and media type (typically audio or
  video).  Even in conferencing applications, especially voice-only,
  the conference focus or bridge provides to each participant a single
  stream containing a mix of the other participants.  It is also common
  to have individual RTP sessions between each endpoint and the RTP
  mixer, meaning that the mixer functions as an RTP-terminating
  gateway.

  Applications and systems that aren't updated to handle multiple
  streams following these recommendations can have issues with
  participating in RTP sessions containing multiple SSRCs within a
  single session, such as:

  1.  The need to handle more than one stream simultaneously rather
      than replacing an already-existing stream with a new one.

  2.  Being capable of decoding multiple streams simultaneously.

  3.  Being capable of rendering multiple streams simultaneously.

  This indicates that gateways attempting to interconnect to this class
  of devices have to make sure that only one RTP stream of each media
  type gets delivered to the endpoint if it's expecting only one and
  that the multiplexing format is what the device expects.  It is
  highly unlikely that RTP translator-based interworking can be made to
  function successfully in such a context.

4.2.  Network Considerations

  The RTP implementer needs to consider that the RTP multiplexing
  choice also impacts network-level mechanisms.

4.2.1.  Quality of Service

  QoS mechanisms are either flow based or packet marking based.  RSVP
  [RFC2205] is an example of a flow-based mechanism, while Diffserv
  [RFC2474] is an example of a packet-marking-based mechanism.

  For a flow-based scheme, additional SSRCs will receive the same QoS
  as all other RTP streams being part of the same 5-tuple (protocol,
  source address, destination address, source port, destination port),
  which is the most common selector for flow-based QoS.

  For a packet-marking-based scheme, the method of multiplexing will
  not affect the possibility of using QoS.  Different Differentiated
  Services Code Points (DSCPs) can be assigned to different packets
  within a transport flow (5-tuple) as well as within an RTP stream,
  assuming the usage of UDP or other transport protocols that do not
  have issues with packet reordering within the transport flow
  (5-tuple).  To avoid packet-reordering issues, packets belonging to
  the same RTP flow should limit their use of DSCPs to packets whose
  corresponding Per-Hop Behavior (PHB) do not enable reordering.  If
  the transport protocol being used assumes in-order delivery of
  packets (e.g., TCP and the Stream Control Transmission Protocol
  (SCTP)), then a single DSCP should be used.  For more discussion on
  this topic, see [RFC7657].

  The method for assigning marking to packets can impact what number of
  RTP sessions to choose.  If this marking is done using a network
  ingress function, it can have issues discriminating the different RTP
  streams.  The network API on the endpoint also needs to be capable of
  setting the marking on a per-packet basis to reach full
  functionality.

4.2.2.  NAT and Firewall Traversal

  In today's networks, there exist a large number of middleboxes.
  Those that normally have the most impact on RTP are Network Address
  Translators (NATs) and Firewalls (FWs).

  Below, we analyze and comment on the impact of requiring more
  underlying transport flows in the presence of NATs and FWs:

  Endpoint Port Consumption:
     A given IP address only has 65536 available local ports per
     transport protocol for all consumers of ports that exist on the
     machine.  This is normally never an issue for an end-user machine.
     It can become an issue for servers that handle a large number of
     simultaneous streams.  However, if the application uses ICE to
     authenticate STUN requests, a server can serve multiple endpoints
     from the same local port and use the whole 5-tuple (source and
     destination address, source and destination port, protocol) as the
     identifier of flows after having securely bound them to the remote
     endpoint address using the STUN request.  In theory, the minimum
     number of media server ports needed is the maximum number of
     simultaneous RTP sessions a single endpoint can use.  In practice,
     implementations will probably benefit from using more server ports
     to simplify implementation or avoid performance bottlenecks.

  NAT State:
     If an endpoint sits behind a NAT, each flow it generates to an
     external address will result in a state that has to be kept in the
     NAT.  That state is a limited resource.  In home or Small
     Office/Home Office (SOHO) NATs, the most limited resource is
     memory or processing.  For large-scale NATs serving many internal
     endpoints, available external ports are likely the scarce
     resource.  Port limitations are primarily a problem for larger
     centralized NATs where endpoint-independent mapping requires each
     flow to use one port for the external IP address.  This affects
     the maximum number of internal users per external IP address.
     However, as a comparison, a real-time video conference session
     with audio and video likely uses less than 10 UDP flows, compared
     to certain web applications that can use 100+ TCP flows to various
     servers from a single browser instance.

  Extra Delay Added by NAT Traversal:
     Performing the NAT/FW traversal takes a certain amount of time for
     each flow.  The best-case scenario for additional NAT/FW traversal
     time after finding the first valid candidate pair following the
     specified ICE procedures is 1.5*RTT + Ta*(Additional_Flows-1),
     where Ta is the pacing timer.  That assumes a message in one
     direction, immediately followed by a return message in the
     opposite direction to confirm reachability.  It isn't more,
     because ICE first finds one candidate pair that works, prior to
     attempting to establish multiple flows.  Thus, there is no extra
     time until one has found a working candidate pair.  Based on that
     working pair, the extra time is needed to establish the additional
     flows (two or three, in most cases) in parallel.  However, packet
     loss causes extra delays of at least 500 ms (the minimal
     retransmission timer for ICE).

  NAT Traversal Failure Rate:
     Due to the need to establish more than a single flow through the
     NAT, there is some risk that establishing the first flow will
     succeed but one or more of the additional flows will fail.  The
     risk of this happening is hard to quantify but should be fairly
     low, as one flow from the same interfaces has just been
     successfully established.  Thus, only such rare events as NAT
     resource overload, selecting particular port numbers that are
     filtered, etc., ought to be reasons for failure.

  Deep Packet Inspection and Multiple Streams:
     FWs differ in how deeply they inspect packets.  Previous
     experience using FWs and Session Border Gateways (SBGs) with RTP
     shows that there is a significant risk that the FWs and SBGs will
     reject RTP sessions that use multiple SSRCs.

  Using additional RTP streams in the same RTP session and transport
  flow does not introduce any additional NAT traversal complexities per
  RTP stream.  This can be compared with (normally) one or two
  additional transport flows per RTP session when using multiple RTP
  sessions.  Additional lower-layer transport flows will be needed,
  unless an explicit demultiplexing layer is added between RTP and the
  transport protocol.  At the time of this writing, no such mechanism
  was defined.

4.2.3.  Multicast

  Multicast groups provide a powerful tool for a number of real-time
  applications, especially those that desire broadcast-like behaviors
  with one endpoint transmitting to a large number of receivers, like
  in IPTV.  An RTP/RTCP extension to better support Source-Specific
  Multicast (SSM) [RFC5760] is also available.  Many-to-many
  communication, which RTP [RFC3550] was originally built to support,
  has several limitations in common with multicast.

  One limitation is that, for any group, sender-side adaptations with
  the intent to suit all receivers would have to adapt to the most
  limited receiver experiencing the worst conditions among the group
  participants, which imposes degradation for all participants.  For
  broadcast-type applications with a large number of receivers, this is
  not acceptable.  Instead, various receiver-based solutions are
  employed to ensure that the receivers achieve the best possible
  performance.  By using scalable encoding and placing each scalability
  layer in a different multicast group, the receiver can control the
  amount of traffic it receives.  To have each scalability layer in a
  different multicast group, one RTP session per multicast group is
  used.

  In addition, the transport flow considerations in multicast are a bit
  different from unicast; NATs with port translation are not useful in
  the multicast environment, meaning that the entire port range of each
  multicast address is available for distinguishing between RTP
  sessions.

  Thus, when using broadcast applications it appears easiest and most
  straightforward to use multiple RTP sessions for sending different
  media flows used for adapting to network conditions.  It is also
  common that streams improving transport robustness are sent in their
  own multicast group to allow for interworking with legacy
  applications or to support different levels of protection.

  Many-to-many applications have different needs, and the most
  appropriate multiplexing choice will depend on how the actual
  application is realized.  Multicast applications that are capable of
  using sender-side congestion control can avoid the use of multiple
  multicast sessions and RTP sessions that result from the use of
  receiver-side congestion control.

  The properties of a broadcast application using RTP multicast are as
  follows:

  1.  The application uses a group of RTP sessions -- not just one.
      Each endpoint will need to be a member of a number of RTP
      sessions in order to perform well.

  2.  Within each RTP session, the number of RTP receivers is likely to
      be much larger than the number of RTP senders.

  3.  The application needs signaling functions to identify the
      relationships between RTP sessions.

  4.  The application needs signaling or RTP/RTCP functions to identify
      the relationships between SSRCs in different RTP sessions when
      more complex relations than those that can be expressed by the
      CNAME exist.

  Both broadcast and many-to-many multicast applications share a
  signaling requirement; all of the participants need the same RTP and
  payload type configuration.  Otherwise, A could, for example, be
  using payload type 97 as the video codec H.264 while B thinks it is
  MPEG-2.  SDP offer/answer [RFC3264] is not appropriate for ensuring
  this property in a broadcast/multicast context.  The signaling
  aspects of broadcast/multicast are not explored further in this memo.

  Security solutions for this type of group communication are also
  challenging.  First, the key-management mechanism and the security
  protocol need to support group communication.  Second, source
  authentication requires special solutions.  For more discussion on
  this topic, please review "Options for Securing RTP Sessions"
  [RFC7201].

4.3.  Security and Key-Management Considerations

  When dealing with point-to-point two-member RTP sessions only, there
  are few security issues that are relevant to the choice of having one
  RTP session or multiple RTP sessions.  However, there are a few
  aspects of multi-party sessions that might warrant consideration.
  For general information regarding possible methods of securing RTP,
  please review [RFC7201].

4.3.1.  Security Context Scope

  When using SRTP [RFC3711], the security context scope is important
  and can be a necessary differentiation in some applications.  As
  SRTP's crypto suites are (so far) built around symmetric keys, the
  receiver will need to have the same key as the sender.  As a result,
  no one in a multi-party session can be certain that a received packet
  was really sent by the claimed sender and not by another party having
  access to the key.  The single SRTP algorithm not having this
  property is Timed Efficient Stream Loss-Tolerant Authentication
  (TESLA) source authentication [RFC4383].  However, TESLA adds delay
  to achieve source authentication.  In most cases, symmetric ciphers
  provide sufficient security properties, but in a few cases they can
  create issues.

  The first case is when someone leaves a multi-party session and one
  wants to ensure that the party that left can no longer access the RTP
  streams.  This requires that everyone rekey without disclosing the
  new keys to the excluded party.

  A second case is when security is used as an enforcing mechanism for
  stream access differentiation between different receivers.  Take, for
  example, a scalable layer or a high-quality simulcast version that
  only users paying a premium are allowed to access.  The mechanism
  preventing a receiver from getting the high-quality stream can be
  based on the stream being encrypted with a key that users can't
  access without paying a premium, using the key-management mechanism
  to limit access to the key.

  As specified in [RFC3711], SRTP uses unique keys per SSRC; however,
  the original assumption was a single-session master key from which
  SSRC-specific RTP and RTCP keys were derived.  However, that
  assumption was proven incorrect, as the application usage and the
  developed key-management mechanisms have chosen many different
  methods for ensuring unique keys per SSRC.  The key-management
  functions have different abilities to establish different sets of
  keys, normally on a per-endpoint basis.  For example, DTLS-SRTP
  [RFC5764] and Security Descriptions [RFC4568] establish different
  keys for outgoing and incoming traffic from an endpoint.  This key
  usage has to be written into the cryptographic context, possibly
  associated with different SSRCs.  Thus, limitations do exist,
  depending on the chosen key-management method and due to the
  integration of particular implementations of the key-management
  method and SRTP.

4.3.2.  Key Management for Multi-party Sessions

  The capabilities of the key-management method combined with the RTP
  multiplexing choices affect the resulting security properties,
  control over the secured media, and who has access to it.

  Multi-party sessions contain at least one RTP stream from each active
  participant.  Depending on the multi-party topology [RFC7667], each
  participant can both send and receive multiple RTP streams.
  Transport translator-based sessions (Topo-Trn-Translator) and
  multicast sessions (Topo-ASM) can use neither Security Descriptions
  [RFC4568] nor DTLS-SRTP [RFC5764] without an extension, because each
  endpoint provides its own set of keys.  In centralized conferences,
  the signaling counterpart is a conference server, and the transport
  translator is the media-plane unicast counterpart (to which DTLS
  messages would be sent).  Thus, an extension like Encrypted Key
  Transport [RFC8870] or a solution based on Multimedia Internet KEYing
  (MIKEY) [RFC3830] that allows for keying all session participants
  with the same master key is needed.

  Privacy-Enhanced RTP Conferencing (PERC) also enables a different
  trust model with semi-trusted media-switching RTP middleboxes
  [RFC8871].

4.3.3.  Complexity Implications

  There can be complex interactions between the choice of multiplexing
  and topology and the security functions.  This becomes especially
  evident in RTP topologies having any type of middlebox that processes
  or modifies RTP/RTCP packets.  While the overhead of an RTP
  translator or mixer rewriting an SSRC value in the RTP packet of an
  unencrypted session is low, the cost is higher when using
  cryptographic security functions.  For example, if using SRTP
  [RFC3711], the actual security context and exact crypto key are
  determined by the SSRC field value.  If one changes the SSRC value,
  the encryption and authentication must use another key.  Thus,
  changing the SSRC value implies a decryption using the old SSRC and
  its security context, followed by an encryption using the new one.

5.  RTP Multiplexing Design Choices

  This section discusses how some RTP multiplexing design choices can
  be used in applications to achieve certain goals and summarizes the
  implications of such choices.  The benefits and downsides of each
  design are also discussed.

5.1.  Multiple Media Types in One Session

  This design uses a single RTP session for multiple different media
  types, like audio and video, and possibly also transport robustness
  mechanisms like FEC or retransmission.  An endpoint can send zero,
  one, or multiple media sources per media type, resulting in a number
  of RTP streams of various media types for both source and redundancy
  streams.

  Advantages:

  1.  Only a single RTP session is used, which implies:

      *  Minimal need to keep NAT/FW state.

      *  Minimal NAT/FW traversal cost.

      *  Fate-sharing for all media flows.

      *  Minimal overhead for security association establishment.

  2.  Dynamic allocation of RTP streams can be handled almost entirely
      at the RTP level.  The extent to which this allocation can be
      kept at the RTP level depends on the application's needs for an
      explicit indication of stream usage and in how timely a fashion
      that information can be signaled.

  Disadvantages:

  1.  It is less suitable for interworking with other applications that
      use individual RTP sessions per media type or multiple sessions
      for a single media type, due to the risk of SSRC collisions and
      thus a potential need for SSRC translation.

  2.  Negotiation of individual bandwidths for the different media
      types is currently only possible in SDP when using RID [RFC8851].

  3.  It is not suitable for split component terminals (see
      Section 3.10 of [RFC7667]).

  4.  Flow-based QoS cannot be used to provide separate treatment of
      RTP streams compared to others in the single RTP session.

  5.  If there is significant asymmetry between the RTP streams' RTCP
      reporting needs, there are some challenges related to
      configuration and usage to avoid wasting RTCP reporting on the
      RTP stream that does not need such frequent reporting.

  6.  It is not suitable for applications where some receivers like to
      receive only a subset of the RTP streams, especially if multicast
      or a transport translator is being used.

  7.  There are some additional concerns regarding legacy
      implementations that do not support the RTP specification fully
      when it comes to handling multiple SSRCs per endpoint, as
      multiple simultaneous media types are sent as separate SSRCs in
      the same RTP session.

  8.  If the applications need finer control over which session
      participants are included in different sets of security
      associations, most key-management mechanisms will have
      difficulties establishing such a session.

5.2.  Multiple SSRCs of the Same Media Type

  In this design, each RTP session serves only a single media type.
  The RTP session can contain multiple RTP streams, from either a
  single endpoint or multiple endpoints.  This commonly creates a low
  number of RTP sessions, typically only one for audio and one for
  video, with a corresponding need for two listening ports when using
  RTP/RTCP multiplexing [RFC5761].

  Advantages:

  1.  It works well with split component terminals (see Section 3.10 of
      [RFC7667]) where the split is per media type.

  2.  It enables flow-based QoS with different prioritization levels
      between media types.

  3.  For applications with dynamic usage of RTP streams (i.e., streams
      are frequently added and removed), having much of the state
      associated with the RTP session rather than per individual SSRC
      can avoid the need for in-session signaling of meta-information
      about each SSRC.  In simple cases, this allows for unsignaled RTP
      streams where session-level information and an RTCP SDES item
      (e.g., CNAME) are sufficient.  In the more complex cases where
      more source-specific metadata needs to be signaled, the SSRC can
      be associated with an intermediate identifier, e.g., the MID
      conveyed as an SDES item as defined in Section 15 of [RFC8843].

  4.  The overhead of security association establishment is low.

  Disadvantages:

  1.  A slightly higher number of RTP sessions are needed, compared to
      multiple media types in one session (Section 5.1).  This implies
      the following:

      *  More NAT/FW state is needed.

      *  The cost of NAT/FW traversal is increased in terms of both
         processing and delay.

  2.  There is some potential for concern regarding legacy
      implementations that don't support the RTP specification fully
      when it comes to handling multiple SSRCs per endpoint.

  3.  It is not possible to control security associations for sets of
      RTP streams within the same media type with today's key-
      management mechanisms, unless these are split into different RTP
      sessions (Section 5.3).

  For RTP applications where all RTP streams of the same media type
  share the same usage, this structure provides efficiency gains in the
  amount of network state used and provides more fate-sharing with
  other media flows of the same type.  At the same time, it still
  maintains almost all functionalities for the negotiation signaling of
  properties per individual media type and also enables flow-based QoS
  prioritization between media types.  It handles multi-party sessions
  well, independently of multicast or centralized transport
  distribution, as additional sources can dynamically enter and leave
  the session.

5.3.  Multiple Sessions for One Media Type

  This design goes one step further than the design discussed in
  Section 5.2 by also using multiple RTP sessions for a single media
  type.  The main reason for going in this direction is that the RTP
  application needs separation of the RTP streams according to their
  usage, such as, for example, scalability over multicast, simulcast,
  the need for extended QoS prioritization, or the need for fine-
  grained signaling using RTP session-focused signaling tools.

  Advantages:

  1.  This design is more suitable for multicast usage where receivers
      can individually select which RTP sessions they want to
      participate in, assuming that each RTP session has its own
      multicast group.

  2.  When multiple different usages exist, the application can
      indicate its usage of the RTP streams at the RTP session level.

  3.  There is less need for SSRC-specific explicit signaling for each
      media stream and thus a reduced need for explicit and timely
      signaling when RTP streams are added or removed.

  4.  It enables detailed QoS prioritization for flow-based mechanisms.

  5.  It works well with split component terminals (see Section 3.10 of
      [RFC7667]).

  6.  The scope for who is included in a security association can be
      structured around the different RTP sessions, thus enabling such
      functionality with existing key-management mechanisms.

  Disadvantages:

  1.  There is an increased amount of session configuration state
      compared to multiple SSRCs of the same media type (Section 5.2),
      due to the increased amount of RTP sessions.

  2.  For RTP streams that are part of scalability, simulcast, or
      transport robustness, a method for binding sources across
      multiple RTP sessions is needed.

  3.  There is some potential for concern regarding legacy
      implementations that don't support the RTP specification fully
      when it comes to handling multiple SSRCs per endpoint.

  4.  The overhead of security association establishment is higher, due
      to the increased number of RTP sessions.

  5.  If the applications need finer control over which participants in
      a given RTP session are included in different sets of security
      associations, most of today's key-management mechanisms will have
      difficulties establishing such a session.

  For more-complex RTP applications that have several different usages
  for RTP streams of the same media type or that use scalability or
  simulcast, this solution can enable those functions, at the cost of
  increased overhead associated with the additional sessions.  This
  type of structure is suitable for more-advanced applications as well
  as multicast-based applications requiring differentiation to
  different participants.

5.4.  Single SSRC per Endpoint

  In this design, each endpoint in a point-to-point session has only a
  single SSRC; thus, the RTP session contains only two SSRCs -- one
  local and one remote.  This session can be used either
  unidirectionally (i.e., one SSRC sends an RTP stream that is received
  by the other SSRC) or bidirectionally (i.e., the two SSRCs both send
  an RTP stream and receive the RTP stream sent by the other endpoint).
  If the application needs additional media flows between the
  endpoints, it will have to establish additional RTP sessions.

  Advantages:

  1.  This design has great potential for interoperability with legacy
      applications, as it will not tax any RTP stack implementations.

  2.  The signaling system makes it possible to negotiate and describe
      the exact formats and bitrates for each RTP stream, especially
      using today's tools in SDP.

  3.  It is possible to control security associations per RTP stream
      with current key-management functions, since each RTP stream is
      directly related to an RTP session and the most commonly used
      keying mechanisms operate on a per-session basis.

  Disadvantages:

  1.  The amount of NAT/FW state grows linearly with the number of RTP
      streams.

  2.  NAT/FW traversal increases delay and resource consumption.

  3.  There are likely more signaling message and signaling processing
      requirements due to the increased amount of session-related
      information.

  4.  There is higher potential for a single RTP stream to fail during
      transport between the endpoints, due to the need for a separate
      NAT/FW traversal for every RTP stream, since there is only one
      stream per session.

  5.  The amount of explicit state for relating RTP streams grows,
      depending on how the application relates RTP streams.

  6.  Port consumption might become a problem for centralized services,
      where the central node's port or 5-tuple filter consumption grows
      rapidly with the number of sessions.

  7.  For applications where RTP stream usage is highly dynamic, i.e.,
      entities frequently enter and leave sessions, the amount of
      signaling can become high.  Issues can also arise from the need
      for timely establishment of additional RTP sessions.

  8.  If, against the recommendation in [RFC3550], the same SSRC value
      is reused in multiple RTP sessions rather than being randomly
      chosen, interworking with applications that use a different
      multiplexing structure will require SSRC translation.

  RTP applications with a strong need to interwork with legacy RTP
  applications can potentially benefit from this structure.  However, a
  large number of media descriptions in SDP can also run into issues
  with existing implementations.  For any application needing a larger
  number of media flows, the overhead can become very significant.
  This structure is also not suitable for non-mixed multi-party
  sessions, as any given RTP stream from each participant, although
  having the same usage in the application, needs its own RTP session.
  In addition, the dynamic behavior that can arise in multi-party
  applications can tax the signaling system and make timely media
  establishment more difficult.

5.5.  Summary

  Both the "single SSRC per endpoint" (Section 5.4) and "multiple media
  types in one session" (Section 5.1) cases require full explicit
  signaling of the media stream relationships.  However, they operate
  on two different levels, where the first primarily enables session-
  level binding and the second needs SSRC-level binding.  From another
  perspective, the two solutions are the two extremes when it comes to
  the number of RTP sessions needed.

  The two other designs -- multiple SSRCs of the same media type
  (Section 5.2) and multiple sessions for one media type (Section 5.3)
  -- are two examples that primarily allow for some implicit mapping of
  the role or usage of the RTP streams based on which RTP session they
  appear in.  Thus, they potentially allow for less signaling and, in
  particular, reduce the need for real-time signaling in sessions with
  a dynamically changing number of RTP streams.  They also represent
  points between the first two designs when it comes to the amount of
  RTP sessions established, i.e., they represent an attempt to balance
  the amount of RTP sessions with the functionality the communication
  session provides at both the network level and the signaling level.

6.  Guidelines

  This section contains a number of multi-stream guidelines for
  implementers, system designers, and specification writers.

  Do not require the use of the same SSRC value across RTP sessions:
     As discussed in Section 3.4.3, there are downsides to using the
     same SSRC in multiple RTP sessions as a mechanism to bind related
     RTP streams together.  It is instead recommended to use a
     mechanism to explicitly signal the relationship, in either
     RTP/RTCP or the signaling mechanism used to establish the RTP
     session(s).

  Use additional RTP streams for additional media sources:
     In the cases where an RTP endpoint needs to transmit additional
     RTP streams of the same media type in the application, with the
     same processing requirements at the network and RTP layers, it is
     suggested to send them in the same RTP session.  For example, in
     the case of a telepresence room where there are three cameras and
     each camera captures two persons sitting at the table, we suggest
     that each camera send its own RTP stream within a single RTP
     session.

  Use additional RTP sessions for streams with different
  requirements:
     When RTP streams have different processing requirements from the
     network or the RTP layer at the endpoints, it is suggested that
     the different types of streams be put in different RTP sessions.
     This includes the case where different participants want different
     subsets of the set of RTP streams.

  Use grouping when using multiple RTP sessions:
     When using multiple RTP session solutions, it is suggested to
     explicitly group the involved RTP sessions when needed using a
     signaling mechanism -- for example, see "The Session Description
     Protocol (SDP) Grouping Framework" [RFC5888] -- using some
     appropriate grouping semantics.

  Ensure that RTP/RTCP extensions support multiple RTP streams as
  well as multiple RTP sessions:
     When defining an RTP or RTCP extension, the creator needs to
     consider if this extension is applicable for use with additional
     SSRCs and multiple RTP sessions.  Any extension intended to be
     generic must support both.  Extensions that are not as generally
     applicable will have to consider whether interoperability is
     better served by defining a single solution or providing both
     options.

  Provide adequate extensions for transport support:
     When defining new RTP/RTCP extensions intended for transport
     support, like the retransmission or FEC mechanisms, they must
     include support for both multiple RTP streams in the same RTP
     session and multiple RTP sessions, such that application
     developers can choose freely from the set of mechanisms without
     concerning themselves with which of the multiplexing choices a
     particular solution supports.

7.  IANA Considerations

  This document has no IANA actions.

8.  Security Considerations

  The security considerations discussed in the RTP specification
  [RFC3550]; any applicable RTP profile [RFC3551] [RFC4585] [RFC3711];
  and the extensions for sending multiple media types in a single RTP
  session [RFC8860], RID [RFC8851], BUNDLE [RFC8843], [RFC5760], and
  [RFC5761] apply if selected and thus need to be considered in the
  evaluation.

  Section 4.3 discusses the security implications of choosing multiple
  SSRCs vs. multiple RTP sessions.

9.  References

9.1.  Normative References

  [RFC3550]  Schulzrinne, H., Casner, S., Frederick, R., and V.
             Jacobson, "RTP: A Transport Protocol for Real-Time
             Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550,
             July 2003, <https://www.rfc-editor.org/info/rfc3550>.

  [RFC3551]  Schulzrinne, H. and S. Casner, "RTP Profile for Audio and
             Video Conferences with Minimal Control", STD 65, RFC 3551,
             DOI 10.17487/RFC3551, July 2003,
             <https://www.rfc-editor.org/info/rfc3551>.

  [RFC3711]  Baugher, M., McGrew, D., Naslund, M., Carrara, E., and K.
             Norrman, "The Secure Real-time Transport Protocol (SRTP)",
             RFC 3711, DOI 10.17487/RFC3711, March 2004,
             <https://www.rfc-editor.org/info/rfc3711>.

  [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,
             DOI 10.17487/RFC4585, July 2006,
             <https://www.rfc-editor.org/info/rfc4585>.

  [RFC5576]  Lennox, J., Ott, J., and T. Schierl, "Source-Specific
             Media Attributes in the Session Description Protocol
             (SDP)", RFC 5576, DOI 10.17487/RFC5576, June 2009,
             <https://www.rfc-editor.org/info/rfc5576>.

  [RFC5760]  Ott, J., Chesterfield, J., and E. Schooler, "RTP Control
             Protocol (RTCP) Extensions for Single-Source Multicast
             Sessions with Unicast Feedback", RFC 5760,
             DOI 10.17487/RFC5760, February 2010,
             <https://www.rfc-editor.org/info/rfc5760>.

  [RFC5761]  Perkins, C. and M. Westerlund, "Multiplexing RTP Data and
             Control Packets on a Single Port", RFC 5761,
             DOI 10.17487/RFC5761, April 2010,
             <https://www.rfc-editor.org/info/rfc5761>.

  [RFC7656]  Lennox, J., Gross, K., Nandakumar, S., Salgueiro, G., and
             B. Burman, Ed., "A Taxonomy of Semantics and Mechanisms
             for Real-Time Transport Protocol (RTP) Sources", RFC 7656,
             DOI 10.17487/RFC7656, November 2015,
             <https://www.rfc-editor.org/info/rfc7656>.

  [RFC7667]  Westerlund, M. and S. Wenger, "RTP Topologies", RFC 7667,
             DOI 10.17487/RFC7667, November 2015,
             <https://www.rfc-editor.org/info/rfc7667>.

  [RFC8843]  Holmberg, C., Alvestrand, H., and C. Jennings,
             "Negotiating Media Multiplexing Using the Session
             Description Protocol (SDP)", RFC 8843,
             DOI 10.17487/RFC8843, January 2021,
             <https://www.rfc-editor.org/info/rfc8843>.

  [RFC8851]  Roach, A.B., Ed., "RTP Payload Format Restrictions",
             RFC 8851, DOI 10.17487/RFC8851, January 2021,
             <https://www.rfc-editor.org/info/rfc8851>.

  [RFC8852]  Roach, A.B., Nandakumar, S., and P. Thatcher, "RTP Stream
             Identifier Source Description (SDES)", RFC 8852,
             DOI 10.17487/RFC8852, January 2021,
             <https://www.rfc-editor.org/info/rfc8852>.

  [RFC8860]  Westerlund, M., Perkins, C., and J. Lennox, "Sending
             Multiple Types of Media in a Single RTP Session",
             RFC 8860, DOI 10.17487/RFC8860, January 2021,
             <https://www.rfc-editor.org/info/rfc8860>.

  [RFC8870]  Jennings, C., Mattsson, J., McGrew, D., Wing, D., and F.
             Andreasen, "Encrypted Key Transport for DTLS and Secure
             RTP", RFC 8870, DOI 10.17487/RFC8870, January 2021,
             <https://www.rfc-editor.org/info/rfc8870>.

9.2.  Informative References

  [JINGLE]   Ludwig, S., Beda, J., Saint-Andre, P., McQueen, R., Egan,
             S., and J. Hildebrand, "XEP-0166: Jingle", September 2018,
             <https://xmpp.org/extensions/xep-0166.html>.

  [RFC2198]  Perkins, C., Kouvelas, I., Hodson, O., Hardman, V.,
             Handley, M., Bolot, J.C., Vega-Garcia, A., and S. Fosse-
             Parisis, "RTP Payload for Redundant Audio Data", RFC 2198,
             DOI 10.17487/RFC2198, September 1997,
             <https://www.rfc-editor.org/info/rfc2198>.

  [RFC2205]  Braden, R., Ed., Zhang, L., Berson, S., Herzog, S., and S.
             Jamin, "Resource ReSerVation Protocol (RSVP) -- Version 1
             Functional Specification", RFC 2205, DOI 10.17487/RFC2205,
             September 1997, <https://www.rfc-editor.org/info/rfc2205>.

  [RFC2474]  Nichols, K., Blake, S., Baker, F., and D. Black,
             "Definition of the Differentiated Services Field (DS
             Field) in the IPv4 and IPv6 Headers", RFC 2474,
             DOI 10.17487/RFC2474, December 1998,
             <https://www.rfc-editor.org/info/rfc2474>.

  [RFC2974]  Handley, M., Perkins, C., and E. Whelan, "Session
             Announcement Protocol", RFC 2974, DOI 10.17487/RFC2974,
             October 2000, <https://www.rfc-editor.org/info/rfc2974>.

  [RFC3261]  Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston,
             A., Peterson, J., Sparks, R., Handley, M., and E.
             Schooler, "SIP: Session Initiation Protocol", RFC 3261,
             DOI 10.17487/RFC3261, June 2002,
             <https://www.rfc-editor.org/info/rfc3261>.

  [RFC3264]  Rosenberg, J. and H. Schulzrinne, "An Offer/Answer Model
             with Session Description Protocol (SDP)", RFC 3264,
             DOI 10.17487/RFC3264, June 2002,
             <https://www.rfc-editor.org/info/rfc3264>.

  [RFC3389]  Zopf, R., "Real-time Transport Protocol (RTP) Payload for
             Comfort Noise (CN)", RFC 3389, DOI 10.17487/RFC3389,
             September 2002, <https://www.rfc-editor.org/info/rfc3389>.

  [RFC3830]  Arkko, J., Carrara, E., Lindholm, F., Naslund, M., and K.
             Norrman, "MIKEY: Multimedia Internet KEYing", RFC 3830,
             DOI 10.17487/RFC3830, August 2004,
             <https://www.rfc-editor.org/info/rfc3830>.

  [RFC4103]  Hellstrom, G. and P. Jones, "RTP Payload for Text
             Conversation", RFC 4103, DOI 10.17487/RFC4103, June 2005,
             <https://www.rfc-editor.org/info/rfc4103>.

  [RFC4383]  Baugher, M. and E. Carrara, "The Use of Timed Efficient
             Stream Loss-Tolerant Authentication (TESLA) in the Secure
             Real-time Transport Protocol (SRTP)", RFC 4383,
             DOI 10.17487/RFC4383, February 2006,
             <https://www.rfc-editor.org/info/rfc4383>.

  [RFC4566]  Handley, M., Jacobson, V., and C. Perkins, "SDP: Session
             Description Protocol", RFC 4566, DOI 10.17487/RFC4566,
             July 2006, <https://www.rfc-editor.org/info/rfc4566>.

  [RFC4568]  Andreasen, F., Baugher, M., and D. Wing, "Session
             Description Protocol (SDP) Security Descriptions for Media
             Streams", RFC 4568, DOI 10.17487/RFC4568, July 2006,
             <https://www.rfc-editor.org/info/rfc4568>.

  [RFC4588]  Rey, J., Leon, D., Miyazaki, A., Varsa, V., and R.
             Hakenberg, "RTP Retransmission Payload Format", RFC 4588,
             DOI 10.17487/RFC4588, July 2006,
             <https://www.rfc-editor.org/info/rfc4588>.

  [RFC5104]  Wenger, S., Chandra, U., Westerlund, M., and B. Burman,
             "Codec Control Messages in the RTP Audio-Visual Profile
             with Feedback (AVPF)", RFC 5104, DOI 10.17487/RFC5104,
             February 2008, <https://www.rfc-editor.org/info/rfc5104>.

  [RFC5109]  Li, A., Ed., "RTP Payload Format for Generic Forward Error
             Correction", RFC 5109, DOI 10.17487/RFC5109, December
             2007, <https://www.rfc-editor.org/info/rfc5109>.

  [RFC5389]  Rosenberg, J., Mahy, R., Matthews, P., and D. Wing,
             "Session Traversal Utilities for NAT (STUN)", RFC 5389,
             DOI 10.17487/RFC5389, October 2008,
             <https://www.rfc-editor.org/info/rfc5389>.

  [RFC5764]  McGrew, D. and E. Rescorla, "Datagram Transport Layer
             Security (DTLS) Extension to Establish Keys for the Secure
             Real-time Transport Protocol (SRTP)", RFC 5764,
             DOI 10.17487/RFC5764, May 2010,
             <https://www.rfc-editor.org/info/rfc5764>.

  [RFC5888]  Camarillo, G. and H. Schulzrinne, "The Session Description
             Protocol (SDP) Grouping Framework", RFC 5888,
             DOI 10.17487/RFC5888, June 2010,
             <https://www.rfc-editor.org/info/rfc5888>.

  [RFC6465]  Ivov, E., Ed., Marocco, E., Ed., and J. Lennox, "A Real-
             time Transport Protocol (RTP) Header Extension for Mixer-
             to-Client Audio Level Indication", RFC 6465,
             DOI 10.17487/RFC6465, December 2011,
             <https://www.rfc-editor.org/info/rfc6465>.

  [RFC7201]  Westerlund, M. and C. Perkins, "Options for Securing RTP
             Sessions", RFC 7201, DOI 10.17487/RFC7201, April 2014,
             <https://www.rfc-editor.org/info/rfc7201>.

  [RFC7657]  Black, D., Ed. and P. Jones, "Differentiated Services
             (Diffserv) and Real-Time Communication", RFC 7657,
             DOI 10.17487/RFC7657, November 2015,
             <https://www.rfc-editor.org/info/rfc7657>.

  [RFC7826]  Schulzrinne, H., Rao, A., Lanphier, R., Westerlund, M.,
             and M. Stiemerling, Ed., "Real-Time Streaming Protocol
             Version 2.0", RFC 7826, DOI 10.17487/RFC7826, December
             2016, <https://www.rfc-editor.org/info/rfc7826>.

  [RFC7983]  Petit-Huguenin, M. and G. Salgueiro, "Multiplexing Scheme
             Updates for Secure Real-time Transport Protocol (SRTP)
             Extension for Datagram Transport Layer Security (DTLS)",
             RFC 7983, DOI 10.17487/RFC7983, September 2016,
             <https://www.rfc-editor.org/info/rfc7983>.

  [RFC8088]  Westerlund, M., "How to Write an RTP Payload Format",
             RFC 8088, DOI 10.17487/RFC8088, May 2017,
             <https://www.rfc-editor.org/info/rfc8088>.

  [RFC8108]  Lennox, J., Westerlund, M., Wu, Q., and C. Perkins,
             "Sending Multiple RTP Streams in a Single RTP Session",
             RFC 8108, DOI 10.17487/RFC8108, March 2017,
             <https://www.rfc-editor.org/info/rfc8108>.

  [RFC8445]  Keranen, A., Holmberg, C., and J. Rosenberg, "Interactive
             Connectivity Establishment (ICE): A Protocol for Network
             Address Translator (NAT) Traversal", RFC 8445,
             DOI 10.17487/RFC8445, July 2018,
             <https://www.rfc-editor.org/info/rfc8445>.

  [RFC8871]  Jones, P., Benham, D., and C. Groves, "A Solution
             Framework for Private Media in Privacy-Enhanced RTP
             Conferencing (PERC)", RFC 8871, DOI 10.17487/RFC8871,
             January 2021, <https://www.rfc-editor.org/info/rfc8871>.

Appendix A.  Dismissing Payload Type Multiplexing

  This section documents a number of reasons why using the payload type
  as a multiplexing point is unsuitable for most issues related to
  multiple RTP streams.  Attempting to use payload type multiplexing
  beyond its defined usage has well-known negative effects on RTP, as
  discussed below.  To use the payload type as the single discriminator
  for multiple streams implies that all the different RTP streams are
  being sent with the same SSRC, thus using the same timestamp and
  sequence number space.  The many effects of using payload type
  multiplexing are as follows:

  1.   Constraints are placed on the RTP timestamp rate for the
       multiplexed media.  For example, RTP streams that use different
       RTP timestamp rates cannot be combined, as the timestamp values
       need to be consistent across all multiplexed media frames.
       Thus, streams are forced to use the same RTP timestamp rate.
       When this is not possible, payload type multiplexing cannot be
       used.

  2.   Many RTP payload formats can fragment a media object over
       multiple RTP packets, like parts of a video frame.  These
       payload formats need to determine the order of the fragments to
       correctly decode them.  Thus, it is important to ensure that all
       fragments related to a frame or a similar media object are
       transmitted in sequence and without interruptions within the
       object.  This can be done relatively easily on the sender side
       by ensuring that the fragments of each RTP stream are sent in
       sequence.

  3.   Some media formats require uninterrupted sequence number space
       between media parts.  These are media formats where any missing
       RTP sequence number will result in decoding failure or invoking
       a repair mechanism within a single media context.  The text/t140
       payload format [RFC4103] is an example of such a format.  These
       formats will need a sequence numbering abstraction function
       between RTP and the individual RTP stream before being used with
       payload type multiplexing.

  4.   Sending multiple media streams in the same sequence number space
       makes it impossible to determine which media stream lost a
       packet.  Such a scenario causes difficulties, since the receiver
       cannot determine to which stream it should apply packet-loss
       concealment or other stream-specific loss-mitigation mechanisms.

  5.   If RTP retransmission [RFC4588] is used and packet loss occurs,
       it is possible to ask for the missing packet(s) by SSRC and
       sequence number -- not by payload type.  If only some of the
       payload type multiplexed streams are of interest, there is no
       way to tell which missing packet or packets belong to the stream
       or streams of interest, and all lost packets need to be
       requested, wasting bandwidth.

  6.   The current RTCP feedback mechanisms are built around providing
       feedback on RTP streams based on stream ID (SSRC), packet
       (sequence numbers), and time interval (RTP timestamps).  There
       is almost never a field to indicate which payload type is
       reported, so sending feedback for a specific RTP payload type is
       difficult without extending existing RTCP reporting.

  7.   The current RTCP media control messages specification [RFC5104]
       is oriented around controlling particular media flows, i.e.,
       requests are done by addressing a particular SSRC.  Such
       mechanisms would need to be redefined to support payload type
       multiplexing.

  8.   The number of payload types is inherently limited.  Accordingly,
       using payload type multiplexing limits the number of streams
       that can be multiplexed and does not scale.  This limitation is
       exacerbated if one uses solutions like RTP and RTCP multiplexing
       [RFC5761] where a number of payload types are blocked due to the
       overlap between RTP and RTCP.

  9.   At times, there is a need to group multiplexed streams.  This is
       currently possible for RTP sessions and SSRCs, but there is no
       defined way to group payload types.

  10.  It is currently not possible to signal bandwidth requirements
       per RTP stream when using payload type multiplexing.

  11.  Most existing SDP media-level attributes cannot be applied on a
       per-payload-type basis and would require redefinition in that
       context.

  12.  A legacy endpoint that does not understand the indication that
       different RTP payload types are different RTP streams might be
       slightly confused by the large amount of possibly overlapping or
       identically defined RTP payload types.

Appendix B.  Signaling Considerations

  Signaling is not an architectural consideration for RTP itself, so
  this discussion has been moved to an appendix.  However, it is
  extremely important for anyone building complete applications, so it
  is deserving of discussion.

  We document some issues here that need to be addressed when using
  some form of signaling to establish RTP sessions.  These issues
  cannot be addressed by simply tweaking, extending, or profiling RTP;
  rather, they require a dedicated and in-depth look at the signaling
  primitives that set up the RTP sessions.

  There exist various signaling solutions for establishing RTP
  sessions.  Many are based on SDP [RFC4566]; however, SDP
  functionality is also dependent on the signaling protocols carrying
  the SDP.  The Real-Time Streaming Protocol (RTSP) [RFC7826] and the
  Session Announcement Protocol (SAP) [RFC2974] both use SDP in a
  declarative fashion, while SIP [RFC3261] uses SDP with the additional
  definition of offer/answer [RFC3264].  The impact on signaling, and
  especially on SDP, needs to be considered, as it can greatly affect
  how to deploy a certain multiplexing point choice.

B.1.  Session-Oriented Properties

  One aspect of existing signaling protocols is that they are focused
  on RTP sessions or, in the case of SDP, the concept of media
  descriptions.  A number of things are signaled at the media
  description level, but those are not necessarily strictly bound to an
  RTP session and could be of interest for signaling, especially for a
  particular RTP stream (SSRC) within the session.  The following
  properties have been identified as being potentially useful for
  signaling, and not only at the RTP session level:

  *  Bitrate and/or bandwidth can be specified today only as an
     aggregate limit, or as a common "any RTP stream" limit, unless
     either codec-specific bandwidth limiting or RTCP signaling using
     Temporary Maximum Media Stream Bit Rate Request (TMMBR) messages
     [RFC5104] is used.

  *  Which SSRC will use which RTP payload type (this information will
     be visible in the first media packet but is sometimes useful to
     have before the packet arrives).

  Some of these issues are clearly SDP's problem rather than RTP
  limitations.  However, if the aim is to deploy a solution that uses
  several SSRCs and contains several sets of RTP streams with different
  properties (encoding/packetization parameters, bitrate, etc.),
  putting each set in a different RTP session would directly enable
  negotiation of the parameters for each set.  If insisting on
  additional SSRCs only, a number of signaling extensions are needed to
  clarify that there are multiple sets of RTP streams with different
  properties and that they in fact need to be kept different, since a
  single set will not satisfy the application's requirements.

  For some parameters, such as RTP payload type, resolution, and frame
  rate, an SSRC-linked mechanism has been proposed in [RFC8851].

B.2.  SDP Prevents Multiple Media Types

  SDP uses the "m=" line to both delineate an RTP session and specify
  the top-level media type: audio, video, text, image, application.
  This media type is used as the top-level media type for identifying
  the actual payload format and is bound to a particular payload type
  using the "a=rtpmap:" attribute.  This binding has to be loosened in
  order to use SDP to describe RTP sessions containing multiple top-
  level media types.

  [RFC8843] describes how to let multiple SDP media descriptions use a
  single underlying transport in SDP, which allows the definition of
  one RTP session with different top-level media types.

B.3.  Signaling RTP Stream Usage

  RTP streams being transported in RTP have a particular usage in an
  RTP application.  In many applications to date, this usage of the RTP
  stream is implicitly signaled.  For example, an application might
  choose to take all incoming audio RTP streams, mix them, and play
  them out.  However, in more-advanced applications that use multiple
  RTP streams, there will be more than a single usage or purpose among
  the set of RTP streams being sent or received.  RTP applications will
  need to somehow signal this usage.  The signaling that is used will
  have to identify the RTP streams affected by their RTP-level
  identifiers, which means that they have to be identified by either
  their session or their SSRC + session.

  In some applications, the receiver cannot utilize the RTP stream at
  all before it has received the signaling message describing the RTP
  stream and its usage.  In other applications, there exists a default
  handling method that is appropriate.

  If all RTP streams in an RTP session are to be treated in the same
  way, identifying the session is enough.  If SSRCs in a session are to
  be treated differently, signaling needs to identify both the session
  and the SSRC.

  If this signaling affects how any RTP central node, like an RTP mixer
  or translator that selects, mixes, or processes streams, treats the
  streams, the node will also need to receive the same signaling to
  know how to treat RTP streams with different usages in the right
  fashion.

Acknowledgments

  The authors would like to acknowledge and thank Cullen Jennings, Dale
  R. Worley, Huang Yihong (Rachel), Benjamin Kaduk, Mirja Kühlewind,
  and Vijay Gurbani for review and comments.

Contributors

  Hui Zheng (Marvin) contributed to WG draft versions -04 and -05 of
  the document.

Authors' Addresses

  Magnus Westerlund
  Ericsson
  Torshamnsgatan 23
  SE-164 80 Kista
  Sweden

  Email: [email protected]


  Bo Burman
  Ericsson
  Gronlandsgatan 31
  SE-164 60 Kista
  Sweden

  Email: [email protected]


  Colin Perkins
  University of Glasgow
  School of Computing Science
  Glasgow
  G12 8QQ
  United Kingdom

  Email: [email protected]


  Harald Tveit Alvestrand
  Google
  Kungsbron 2
  SE-11122 Stockholm
  Sweden

  Email: [email protected]


  Roni Even

  Email: [email protected]

You are viewing proxied material from gopher.fnord.one. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.