idnits 2.17.1 draft-westerlund-avtcore-multiplex-architecture-02.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The document seems to lack a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 1283: '... SHOULD be carried in a separate RTP...' RFC 2119 keyword, line 1286: '...nd video streams SHOULD NOT be carried...' RFC 2119 keyword, line 1368: '...ely, each medium SHOULD be carried in ...' Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (July 16, 2012) is 4300 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Outdated reference: A later version (-10) exists of draft-ietf-avtcore-rtp-security-options-00 == Outdated reference: A later version (-11) exists of draft-ietf-avtext-multiple-clock-rates-05 == Outdated reference: A later version (-54) exists of draft-ietf-mmusic-sdp-bundle-negotiation-00 == Outdated reference: A later version (-14) exists of draft-ietf-payload-rtp-howto-02 == Outdated reference: A later version (-02) exists of draft-lennox-avtcore-rtp-multi-stream-00 == Outdated reference: A later version (-05) exists of draft-lennox-mmusic-sdp-source-selection-04 == Outdated reference: A later version (-02) exists of draft-westerlund-avtcore-max-ssrc-01 == Outdated reference: A later version (-07) exists of draft-westerlund-avtcore-transport-multiplexing-02 == Outdated reference: A later version (-03) exists of draft-westerlund-avtext-rtcp-sdes-srcname-00 -- Obsolete informational reference (is this intentional?): RFC 2326 (Obsoleted by RFC 7826) -- Obsolete informational reference (is this intentional?): RFC 4566 (Obsoleted by RFC 8866) -- Obsolete informational reference (is this intentional?): RFC 5117 (Obsoleted by RFC 7667) -- Obsolete informational reference (is this intentional?): RFC 6222 (Obsoleted by RFC 7022) Summary: 1 error (**), 0 flaws (~~), 10 warnings (==), 5 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group M. Westerlund 3 Internet-Draft B. Burman 4 Intended status: Informational Ericsson 5 Expires: January 17, 2013 C. Perkins 6 University of Glasgow 7 H. Alvestrand 8 Google 9 July 16, 2012 11 Guidelines for using the Multiplexing Features of RTP 12 draft-westerlund-avtcore-multiplex-architecture-02 14 Abstract 16 Real-time Transport Protocol (RTP) is a flexible protocol possible to 17 use in a wide range of applications and network and system 18 topologies. This flexibility and the implications of different 19 choices should be understood by any application developer using RTP. 20 To facilitate that understanding, this document contains an in-depth 21 discussion of the usage of RTP's multiplexing points; the RTP session 22 and the Synchronisation Source Identifier (SSRC). The document tries 23 to give guidance and source material for an analysis on the most 24 suitable choices for the application being designed. 26 Status of this Memo 28 This Internet-Draft is submitted in full conformance with the 29 provisions of BCP 78 and BCP 79. 31 Internet-Drafts are working documents of the Internet Engineering 32 Task Force (IETF). Note that other groups may also distribute 33 working documents as Internet-Drafts. The list of current Internet- 34 Drafts is at http://datatracker.ietf.org/drafts/current/. 36 Internet-Drafts are draft documents valid for a maximum of six months 37 and may be updated, replaced, or obsoleted by other documents at any 38 time. It is inappropriate to use Internet-Drafts as reference 39 material or to cite them other than as "work in progress." 41 This Internet-Draft will expire on January 17, 2013. 43 Copyright Notice 45 Copyright (c) 2012 IETF Trust and the persons identified as the 46 document authors. All rights reserved. 48 This document is subject to BCP 78 and the IETF Trust's Legal 49 Provisions Relating to IETF Documents 50 (http://trustee.ietf.org/license-info) in effect on the date of 51 publication of this document. Please review these documents 52 carefully, as they describe your rights and restrictions with respect 53 to this document. Code Components extracted from this document must 54 include Simplified BSD License text as described in Section 4.e of 55 the Trust Legal Provisions and are provided without warranty as 56 described in the Simplified BSD License. 58 Table of Contents 60 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 61 2. Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 5 62 2.1. Terminology . . . . . . . . . . . . . . . . . . . . . . . 5 63 2.2. Subjects Out of Scope . . . . . . . . . . . . . . . . . . 7 64 3. RTP Concepts . . . . . . . . . . . . . . . . . . . . . . . . . 7 65 3.1. Session . . . . . . . . . . . . . . . . . . . . . . . . . 7 66 3.2. SSRC . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 67 3.3. CSRC . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 68 3.4. Payload Type . . . . . . . . . . . . . . . . . . . . . . . 10 69 4. Multiple Streams Alternatives . . . . . . . . . . . . . . . . 12 70 5. RTP Topologies and Issues . . . . . . . . . . . . . . . . . . 13 71 5.1. Point to Point . . . . . . . . . . . . . . . . . . . . . . 13 72 5.1.1. Translators & Gateways . . . . . . . . . . . . . . . . 14 73 5.2. Point to Multipoint Using Multicast . . . . . . . . . . . 15 74 5.3. Point to Multipoint Using an RTP Transport Translator . . 17 75 5.4. Point to Multipoint Using an RTP Mixer . . . . . . . . . . 18 76 5.4.1. Media Mixing . . . . . . . . . . . . . . . . . . . . . 19 77 5.4.2. Media Switching . . . . . . . . . . . . . . . . . . . 22 78 5.4.3. RTP Source Projecting . . . . . . . . . . . . . . . . 24 79 5.5. Point to Multipoint using Multiple Unicast flows . . . . . 26 80 5.6. De-composite Endpoint . . . . . . . . . . . . . . . . . . 27 81 6. Multiple Streams Discussion . . . . . . . . . . . . . . . . . 28 82 6.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 28 83 6.2. RTP/RTCP Aspects . . . . . . . . . . . . . . . . . . . . . 28 84 6.2.1. The RTP Specification . . . . . . . . . . . . . . . . 29 85 6.2.2. Multiple SSRCs in a Session . . . . . . . . . . . . . 31 86 6.2.3. Handling Varying sets of Senders . . . . . . . . . . . 32 87 6.2.4. Cross Session RTCP Requests . . . . . . . . . . . . . 32 88 6.2.5. Binding Related Sources . . . . . . . . . . . . . . . 33 89 6.2.6. Forward Error Correction . . . . . . . . . . . . . . . 35 90 6.2.7. Transport Translator Sessions . . . . . . . . . . . . 36 91 6.3. Interworking . . . . . . . . . . . . . . . . . . . . . . . 36 92 6.3.1. Types of Interworking . . . . . . . . . . . . . . . . 36 93 6.3.2. RTP Translator Interworking . . . . . . . . . . . . . 36 94 6.3.3. Gateway Interworking . . . . . . . . . . . . . . . . . 37 95 6.3.4. Multiple SSRC Legacy Considerations . . . . . . . . . 38 97 6.4. Network Aspects . . . . . . . . . . . . . . . . . . . . . 38 98 6.4.1. Quality of Service . . . . . . . . . . . . . . . . . . 39 99 6.4.2. NAT and Firewall Traversal . . . . . . . . . . . . . . 39 100 6.4.3. Multicast . . . . . . . . . . . . . . . . . . . . . . 41 101 6.4.4. Multiplexing multiple RTP Session on a Single 102 Transport . . . . . . . . . . . . . . . . . . . . . . 41 103 6.5. Security Aspects . . . . . . . . . . . . . . . . . . . . . 42 104 6.5.1. Security Context Scope . . . . . . . . . . . . . . . . 42 105 6.5.2. Key Management for Multi-party session . . . . . . . . 42 106 6.5.3. Complexity Implications . . . . . . . . . . . . . . . 43 107 7. Arch-Types . . . . . . . . . . . . . . . . . . . . . . . . . . 43 108 7.1. Single SSRC per Session . . . . . . . . . . . . . . . . . 43 109 7.2. Multiple SSRCs of the Same Media Type . . . . . . . . . . 45 110 7.3. Multiple Sessions for one Media type . . . . . . . . . . . 46 111 7.4. Multiple Media Types in one Session . . . . . . . . . . . 48 112 7.5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . 49 113 8. Summary considerations and guidelines . . . . . . . . . . . . 50 114 8.1. Guidelines . . . . . . . . . . . . . . . . . . . . . . . . 50 115 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 51 116 10. Security Considerations . . . . . . . . . . . . . . . . . . . 51 117 11. References . . . . . . . . . . . . . . . . . . . . . . . . . . 51 118 11.1. Normative References . . . . . . . . . . . . . . . . . . . 51 119 11.2. Informative References . . . . . . . . . . . . . . . . . . 51 120 Appendix A. Dismissing Payload Type Multiplexing . . . . . . . . 55 121 Appendix B. Proposals for Future Work . . . . . . . . . . . . . . 57 122 Appendix C. RTP Specification Clarifications . . . . . . . . . . 57 123 C.1. RTCP Reporting from all SSRCs . . . . . . . . . . . . . . 58 124 C.2. RTCP Self-reporting . . . . . . . . . . . . . . . . . . . 58 125 C.3. Combined RTCP Packets . . . . . . . . . . . . . . . . . . 58 126 Appendix D. Signalling considerations . . . . . . . . . . . . . . 58 127 D.1. Signalling Aspects . . . . . . . . . . . . . . . . . . . . 59 128 D.1.1. Session Oriented Properties . . . . . . . . . . . . . 59 129 D.1.2. SDP Prevents Multiple Media Types . . . . . . . . . . 60 130 D.1.3. Signalling Media Stream Usage . . . . . . . . . . . . 60 131 Appendix E. Changes from -01 to -02 . . . . . . . . . . . . . . . 61 132 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 61 134 1. Introduction 136 Real-time Transport Protocol (RTP) [RFC3550] is a commonly used 137 protocol for real-time media transport. It is a protocol that 138 provides great flexibility and can support a large set of different 139 applications. RTP has several multiplexing points designed for 140 different purposes. These enable support of multiple media streams 141 and switching between different encoding or packetization of the 142 media. By using multiple RTP sessions, sets of media streams can be 143 structured for efficient processing or identification. Thus the 144 question for any RTP application designer is how to best use the RTP 145 session, the SSRC and the payload type to meet the application's 146 needs. 148 The purpose of this document is to provide clear information about 149 the possibilities of RTP when it comes to multiplexing. The RTP 150 application designer should understand the implications that come 151 from a particular usage of the RTP multiplexing points. The document 152 will recommend against some usages as being unsuitable, in general or 153 for particular purposes. 155 RTP was from the beginning designed for multiple participants in a 156 communication session. This is not restricted to multicast, as some 157 may believe, but also provides functionality over unicast, using 158 either multiple transport flows below RTP or a network node that re- 159 distributes the RTP packets. The re-distributing node can for 160 example be a transport translator (relay) that forwards the packets 161 unchanged, a translator performing media or protocol translation in 162 addition to forwarding, or an RTP mixer that creates new conceptual 163 sources from the received streams. In addition, multiple streams may 164 occur when a single endpoint have multiple media sources, like 165 multiple cameras or microphones that need to be sent simultaneously. 167 This document has been written due to increased interest in more 168 advanced usage of RTP, resulting in questions regarding the most 169 appropriate RTP usage. The limitations in some implementations, RTP/ 170 RTCP extensions, and signalling has also been exposed. It is 171 expected that some limitations will be addressed by updates or new 172 extensions resolving the shortcomings. The authors also hope that 173 clarification on the usefulness of some functionalities in RTP will 174 result in more complete implementations in the future. 176 The document starts with some definitions and then goes into the 177 existing RTP functionalities around multiplexing. Both the desired 178 behaviour and the implications of a particular behaviour depend on 179 which topologies are used, which requires some consideration. This 180 is followed by a discussion of some choices in multiplexing behaviour 181 and their impacts. Some arch-types of RTP usage are discussed. 183 Finally, some recommendations and examples are provided. 185 This document is currently an individual contribution, but it is the 186 intention of the authors that this should become a WG document that 187 objectively describes and provides suitable recommendations for which 188 there is WG consensus. Currently this document only represents the 189 views of the authors. The authors gladly accept any feedback on the 190 document and will be happy to discuss suitable recommendations. 192 2. Definitions 194 2.1. Terminology 196 The following terms and abbreviations are used in this document: 198 Endpoint: A single entity sending or receiving RTP packets. It may 199 be decomposed into several functional blocks, but as long as it 200 behaves a single RTP stack entity it is classified as a single 201 endpoint. 203 Multiparty: A communication situation including multiple end-points. 204 In this document it will be used to refer to situations where more 205 than two end-points communicate. 207 Media Source: The source of a stream of data of one Media Type, It 208 can either be a single media capturing device such as a video 209 camera, a microphone, or a specific output of a media production 210 function, such as an audio mixer, or some video editing function. 211 Sending data from a Media Source may cause multiple RTP sources to 212 send multiple Media Streams. 214 Media Stream: A sequence of RTP packets using a single SSRC that 215 together carries part or all of the content of a specific Media 216 Type from a specific sender source within a given RTP session. 218 RTP Source: The originator or source of a particular Media Stream. 219 Identified using an SSRC in a particular RTP session. An RTP 220 source is the source of a single media stream, and is associated 221 with a single endpoint and a single Media Source. An RTP Source 222 is just called a Source in RFC 3550. 224 Media Sink: A recipient of a Media Stream. The endpoint sinking 225 media are Identified using one or more SSRCs. There may be more 226 than one Media Sink for one RTP source. 228 CNAME: "Canonical name" - identifier associated with one or more RTP 229 sources from a single endpoint. Defined in the RTP specification 230 [RFC3550]. A CNAME identifies a synchronisation context. A CNAME 231 is associated with a single endpoint, although some RTP nodes will 232 use an end-points CNAME on that end-points behalf. An endpoint 233 may use multiple CNAMEs. A CNAME is intended to be globally 234 unique and stable for the full duration of a communication 235 session. [RFC6222] gives updated guidelines for choosing CNAMEs. 237 Media Type: Audio, video, text or data whose form and meaning are 238 defined by a specific real-time application. 240 Multiplex: The operation of taking multiple entities as input, 241 aggregating them onto some common resource while keeping the 242 individual entities addressable such that they can later be fully 243 and unambiguously separated (de-multiplexed) again. 245 RTP Session: As defined by [RFC3550], the endpoints belonging to the 246 same RTP Session are those that share a single SSRC space. That 247 is, those endpoints can see an SSRC identifier transmitted by any 248 one of the other endpoints. An endpoint can receive an SSRC 249 either as SSRC or as CSRC in RTP and RTCP packets. Thus, the RTP 250 Session scope is decided by the endpoints' network interconnection 251 topology, in combination with RTP and RTCP forwarding strategies 252 deployed by endpoints and any interconnecting middle nodes. 254 RTP Session Group: One or more RTP sessions that are used together 255 to perform some function. Examples are multiple RTP sessions used 256 to carry different layers of a layered encoding. In an RTP 257 Session Group, CNAMEs are assumed to be valid across all RTP 258 sessions, and designate synchronisation contexts that can cross 259 RTP sessions. 261 Source: Term that should not be used alone. An RTP Source, as 262 identified by its SSRC, is the source of a single Media Stream; a 263 Media Source can be the source of mutiple Media Streams. 265 SSRC: An RTP 32-bit unsigned integer used as identifier for a RTP 266 Source. 268 CSRC: Contributing Source, A SSRC identifier used in a context, like 269 the RTP headers CSRC list, where it is clear that the Media Source 270 is not the source of the media stream, instead only a contributor 271 to the Media Stream. 273 Signalling: The process of configuring endpoints to participate in 274 one or more RTP sessions. 276 (tbd: The terms "SSRC multiplexing" and "session multiplexing" are 277 confusing, with unclear historical meanings; they need to be removed 278 from this document in the interests of clarity) 280 2.2. Subjects Out of Scope 282 This document is focused on issues that affect RTP. Thus, issues 283 that involve signalling protocols, such as whether SIP, Jingle or 284 some other protocol is in use for session configuration, the 285 particular syntaxes used to define RTP session properties, or the 286 constraints imposed by particular choices in the signalling 287 protocols, are mentioned only as examples in order to describe the 288 RTP issues more precisely. 290 This document assumes the applications will use RTCP. While there 291 are such applications that don't send RTCP, they do not conform to 292 the RTP specification, and thus should be regarded as reusing the RTP 293 packet format, not as implementing the RTP protocol. 295 3. RTP Concepts 297 This section describes the existing RTP tools that are particularly 298 important when discussing multiplexing of different media streams. 300 3.1. Session 302 The RTP Session is the highest semantic level in RTP and contains all 303 of the RTP functionality. RTP itself has no normative statements 304 about the relationship between different RTP sessions. 306 Identifier: RTP in itself does not contain any Session identifier, 307 but relies either on the underlying transport or on the used 308 signalling protocol, depending on in which context the identifier 309 is used (e.g. transport or signalling). Due to this, a single RTP 310 Session may have multiple associated identifiers belonging to 311 different contexts. 313 Position: Depending on underlying transport and signalling 314 protocol. For example, when running RTP on top of UDP, an RTP 315 endpoint can identify and delimit an RTP Session from other RTP 316 Sessions through the UDP source and destination transport 317 address, consisting of network address and port number(s). 319 Commonly, RTP and RTCP use separate ports and the destination 320 transport address is in fact an address pair, but in the case 321 of RTP/RTCP multiplex [RFC5761] there is only a single port. 322 Another example is SDP signalling [RFC4566], where the grouping 323 framework [RFC5888] uses an identifier per "m="-line. If there 324 is a one-to-one mapping between "m="-line and RTP Session, that 325 grouping framework identifier can identify a single RTP 326 Session. 328 Usage: Identify separate RTP Sessions. 330 Uniqueness: Globally unique, but identity can only be detected by 331 the general communication context for the specific endpoint. 333 Inter-relation: Depending on the underlying transport and 334 signalling protocol. 336 Special Restrictions: None. 338 A RTP source in an RTP session that changes its source transport 339 address during a session must also choose a new SSRC identifier to 340 avoid being interpreted as a looped source. 342 The set of participants considered part of the same RTP Session is 343 defined by the RTP specification [RFC3550] as those that share a 344 single SSRC space. That is, those participants that can see an SSRC 345 identifier transmitted by any one of the other participants. A 346 participant can receive an SSRC either as SSRC or CSRC in RTP and 347 RTCP packets. Thus, the RTP Session scope is decided by the 348 participants' network interconnection topology, in combination with 349 RTP and RTCP forwarding strategies deployed by endpoints and any 350 interconnecting middle nodes. 352 3.2. SSRC 354 An SSRC identifies a RTP source or a media sink. For end-points that 355 both source and sink media streams its SSRCs are used in both roles. 356 At any given time, a RTP source has one and only one SSRC - although 357 that may change over the lifetime of the RTP source or sink. An RTP 358 Session serves one or more RTP sources, each sending a Media Stream. 360 Identifier: Synchronisation Source (SSRC), 32-bit unsigned number. 362 Position: In every RTP and RTCP packet header. May be present in 363 RTCP payload. May be present in SDP signalling. 365 Usage: Identify individual RTP sources and media sinks within an 366 RTP Session. Refer to individual RTP sources and media sinks 367 in RTCP messages and SDP signalling. 369 Uniqueness: Randomly chosen, intended to be globally unique 370 within an RTP Session and not dependent on network address. 371 SSRC value collisions may occur and must be handled as 372 specified in RTP [RFC3550]. 374 Inter-relation: SSRC belonging to the same synchronisation 375 context (originating from the same endpoint), within or between 376 RTP Sessions, are indicated through use of identical SDES CNAME 377 items in RTCP compound packets with those SSRC as originating 378 source. SDP signalling can provide explicit SSRC grouping 379 [RFC5576]. When CNAME is inappropriate or insufficient, there 380 exist a few other methods to relate different SSRC. One such 381 case is session-based RTP retransmission [RFC4588]. In some 382 cases, the same SSRC Identifier value is used to relate streams 383 in two different RTP Sessions, such as in Multi-Session 384 Transmission of scalable video [RFC6190]. 386 Special Restrictions: All RTP implementations must be prepared to 387 use procedures for SSRC collision handling, which results in an 388 SSRC number change. A RTP source that changes its RTP Session 389 identifier (e.g. source transport address) during a session must 390 also choose a new SSRC identifier to avoid being interpreted as 391 looped source. 393 Note that RTP sequence number and RTP timestamp are scoped by SSRC 394 and thus independent between different SSRCs. 396 A RTP source having an SSRC identifier can be of different types: 398 Real: Connected to a "physical" media source, for example a camera 399 or microphone. 401 Conceptual: A source with some attributed property generated by some 402 network node, for example a filtering function in an RTP mixer 403 that provides the most active speaker based on some criteria, or a 404 mix representing a set of other sources. 406 Media Sink: A source that does not generate any RTP media stream in 407 itself (e.g. an endpoint only receiving in an RTP session), but 408 anyway need a sender SSRC for use as source in RTCP reports. 410 Note that a endpoint that generates more than one media type, e.g. a 411 conference participant sending both audio and video, need not (and 412 commonly should not) use the same SSRC value across RTP sessions. 414 RTCP Compound packets containing the CNAME SDES item is the 415 designated method to bind an SSRC to a CNAME, effectively cross- 416 correlating SSRCs within and between RTP Sessions as coming from the 417 same endpoint. The main property attributed to SSRCs associated with 418 the same CNAME is that they are from a particular synchronisation 419 context and may be synchronised at playback. 421 Note also that RTP sequence number and RTP timestamp are scoped by 422 SSRC and thus independent between different SSRCs. 424 An RTP receiver receiving a previously unseen SSRC value must 425 interpret it as a new source. It may in fact be a previously 426 existing source that had to change SSRC number due to an SSRC 427 conflict. However, the originator of the previous SSRC should have 428 ended the conflicting source by sending an RTCP BYE for it prior to 429 starting to send with the new SSRC, so the new SSRC is anyway 430 effectively a new source. 432 3.3. CSRC 434 The Contributing Source (CSRC) is not a separate identifier, but an 435 usage of the SSRC identifier. It is optionally included in the RTP 436 header as list of up to 15 contributing RTP sources. CSRC shares the 437 SSRC number space and specifies which set of SSRCs that has 438 contributed to the RTP payload. However, even though each RTP packet 439 and SSRC can be tagged with the contained CSRCs, the media 440 representation of an individual CSRC is in general not possible to 441 extract from the RTP payload since it is typically the result of a 442 media mixing (merge) operation (by an RTP mixer) on the individual 443 media streams corresponding to the CSRC identifiers. The exception 444 is the case when only a single CSRC is indicated as this represent 445 forwarding of a media stream, possibly modified. The RTP header 446 extension for Mixer-to-Client Audio Level Indication [RFC6465] 447 expands on the receivers information about a packet with CSRC list. 448 Due to these restrictions, CSRC will not be considered a fully 449 qualified multiplex point and will be disregarded in the rest of this 450 document. 452 3.4. Payload Type 454 Each Media Stream utilises one or more encoding formats, identified 455 by the Payload Type. 457 The Payload Type is not a multiplexing point. Appendix A gives some 458 of the many reasons why attempting to use it as a multiplexing point 459 will have bad results. 461 Identifier: Payload Type number. 463 Position: In every RTP header and in SDP signalling. 465 Usage: Identify a specific Media Stream encoding format. The 466 format definition may be taken from [RFC3551] for statically 467 allocated Payload Types, but should be explicitly defined in 468 signalling, such as SDP, both for static and dynamic Payload 469 Types. The term "format" here includes whatever can be 470 described by out-of-band signalling means. In SDP, the term 471 "format" includes media type, RTP timestamp sampling rate, 472 codec, codec configuration, payload format configurations, and 473 various robustness mechanisms such as redundant encodings 474 [RFC2198]. 476 Uniqueness: Scoped by sending endpoint within an RTP Session. To 477 avoid any potential for ambiguity, it is desirable that payload 478 types are unique across all sending endpoints within an RTP 479 session, but this is often not true in practice. All SSRC in 480 an RTP session sent from an single endpoint share the same 481 Payload Types definitions. The RTP Payload Type is designed 482 such that only a single Payload Type is valid at any time 483 instant in the SSRC's RTP timestamp time line, effectively 484 time-multiplexing different Payload Types if any change occurs. 485 Used Payload Type may change on a per-packet basis for an SSRC, 486 for example a speech codec making use of generic Comfort Noise 487 [RFC3389]. 489 Inter-relation: There are some uses where Payload Type numbers 490 need to be unique across RTP Sessions. This is for example the 491 case in Media Decoding Dependency [RFC5583] where Payload Types 492 are used to describe media dependency across RTP Sessions. 493 Another example is session-based RTP retransmission [RFC4588]. 495 Special Restrictions: Using different RTP timestamp clock rates for 496 the RTP Payload Types in use in the same RTP Session have issues 497 such as loss of synchronisation. Payload Type clock rate 498 switching requires some special consideration that is described in 499 the multiple clock rates specification 500 [I-D.ietf-avtext-multiple-clock-rates]. 502 If there is a true need to send multiple Payload Types for the same 503 SSRC that are valid for the same RTP Timestamps, then redundant 504 encodings [RFC2198] can be used. Several additional constraints than 505 the ones mentioned above need to be met to enable this use, one of 506 which is that the combined payload sizes of the different Payload 507 Types must not exceed the transport MTU. 509 Other aspects of RTP payload format use are described in RTP Payload 510 HowTo [I-D.ietf-payload-rtp-howto]. 512 4. Multiple Streams Alternatives 514 The reasons why an endpoint may choose to send multiple media streams 515 are widespread. In the below discussion, please keep in mind that 516 the reasons for having multiple media streams vary and include but 517 are not limited to the following: 519 o Multiple Media Sources 521 o Multiple Media Streams may be needed to represent one Media Source 522 (for instance when using layered encodings) 524 o A Retransmission stream may repeat the content of another Media 525 Stream 527 o An FEC stream may provide material that can be used to repair 528 another Media Stream 530 o Alternative Encodings, for instance different codecs for the same 531 audio stream 533 o Alternative formats, for instance multiple resolutions of the same 534 video stream 536 Thus the choice made due to one reason may not be the choice suitable 537 for another reason. In the above list, the different items have 538 different levels of maturity in the discussion on how to solve them. 539 The clearest understanding is associated with multiple media sources 540 of the same media type. However, all warrant discussion and 541 clarification on how to deal with them. 543 This section reviews the alternatives to enable multi-stream 544 handling. Let's start with describing mechanisms that could enable 545 multiple media streams, independent of the purpose for having 546 multiple streams. 548 SSRC Multiplexing: Each additional Media Stream gets its own SSRC 549 within a RTP Session. 551 Session Multiplexing: Using additional RTP Sessions to handle 552 additional Media Streams 554 As the below discussion will show, in reality we cannot choose a 555 single one of the two solutions. To utilise RTP well and as 556 efficiently as possible, both are needed. The real issue is finding 557 the right guidance on when to create RTP sessions and when additional 558 SSRCs in an RTP session is the right choice. 560 5. RTP Topologies and Issues 562 The impact of how RTP Multiplex is performed will in general vary 563 with how the RTP Session participants are interconnected; the RTP 564 Topology [RFC5117]. This section describes the topologies and 565 attempts to highlight the important behaviours concerning RTP 566 multiplexing and multi-stream handling. It lists any identified 567 issues regarding RTP and RTCP handling, and introduces additional 568 topologies that are supported by RTP beyond those included in RTP 569 Topologies [RFC5117]. The RTP Topologies that do not follow the RTP 570 specification or do not attempt to utilise the facilities of RTP are 571 ignored in this document. 573 5.1. Point to Point 575 This is the most basic use case with two endpoints directly 576 interconnected and no additional entities are involved in the 577 communication. 579 +---+ +---+ 580 | A |<------->| B | 581 +---+ +---+ 583 Figure 1: Point to Point 585 A number of applications are using a single RTP session with one RTP 586 source per endpoint. This is likely the simplest case, as you 587 basically doesn't have to make any choices regarding multiplexing. 588 When you add an additional source to either endpoint you immediately 589 create the question do one send the media stream in the existing RTP 590 session or should I use an additional RTP session. 592 This raises a number of considerations that are discussed in detail 593 below (Section 6). But the range over such aspects as: 595 o Does my communication peer support RTP as defined with multiple 596 SSRCs? 598 o Do I need network differentiation in form of QoS? 600 o Can the application more easily process and handle the media 601 streams if they are in different RTP sessions? 603 o etc. 605 The application designer will have to make choices here. The point 606 to point topology can contain one to many RTP sessions with one to 607 many RTP sources (SSRC) per session. 609 5.1.1. Translators & Gateways 611 A point to point communication can end up in a situation when the 612 peer it is communicating with is not compatible with the other peer 613 for various reasons. This is in many situation resolved by the 614 inclusion of a translator in-between the two peers. 616 +---+ +---+ +---+ 617 | A |<------>| T |<------->| B | 618 +---+ +---+ +---+ 620 Point to Point with Translator 622 The translator main purpose is to make the peer look to the other 623 peer like something it is compatible with. An RTP translator will 624 commonly not be distinguishable from the actual end-point. It is 625 intentional not identifiable on RTP level. Reasons a translator can 626 be required are: 628 o No common media codec for a media type thus requiring transcoding. 630 o Different usages of the RTP multiplexing points 632 o Usage of different media transport protocols 634 o Usage of different transport protocols 636 o Different security solutions 638 The RTP translator will rewrite RTP and RTCP as required to provide a 639 consistent view to each peer of the traffic the translator forwards 640 and the feedback being provided back to the RTP source. 642 In some case security policies or the need for monitoring the media 643 streams the direct communication are directed to a pass through a 644 specific middlebox, commonly called a gateway. This is often placed 645 on the border of administrative domain where the security policies 646 are in effect. Many gateways simple relay the RTP and RTCP traffic 647 between the domains, but some may do more by including above 648 mentioned translator functions or even go as far as terminating the 649 RTP session and do application level forwarding of the media traffic. 650 The later places requirements on the gateway to have full 651 understanding of the application logic and especially be able to cope 652 with any congestion control or media adaptation. 654 A variant of translator behaviour worth pointing out is when an 655 endpoint A sends a media flow to B. On the path there is a device T 656 that on A's behalf does something with the media streams, for example 657 adds an RTP session with FEC information for A's media streams. T 658 will in this case need to bind the new FEC streams to A's media 659 stream by using the same CNAME as A. 661 +------+ +------+ +------+ 662 | | | | | | 663 | A |------->| T |-------->| B | 664 | | | |---FEC-->| | 665 +------+ +------+ +------+ 667 Figure 2: When De-composition is a Translator 669 This type of functionality where T does something with the media 670 stream on behalf of A is clearly covered under the media translator 671 definition (Section 5.3). 673 5.2. Point to Multipoint Using Multicast 675 This section discusses the Point to Multi-point using Multicast to 676 interconnect the session participants. This needs to consider both 677 Any Source Multicast (ASM) and Source-Specific Multicast (SSM). 678 There are large commercial deployments of multicast for applications 679 like IPTV. 681 +-----+ 682 +---+ / \ +---+ 683 | A |----/ \---| B | 684 +---+ / Multi- \ +---+ 685 + Cast + 686 +---+ \ Network / +---+ 687 | C |----\ /---| D | 688 +---+ \ / +---+ 689 +-----+ 691 Figure 3: Point to Multipoint Using Any Source Multicast 693 In Any Source Multicast, any of the participants can send to all the 694 other participants, simply by sending a packet to the multicast 695 group. That is not possible in Source Specific Multicast [RFC4607] 696 where only a single source (Distribution Source) can send to the 697 multicast group, creating a topology that looks like the one below: 699 +--------+ +-----+ 700 |Media | | | Source-specific 701 |Sender 1|<----->| D S | Multicast 702 +--------+ | I O | +--+----------------> R(1) 703 | S U | | | | 704 +--------+ | T R | | +-----------> R(2) | 705 |Media |<----->| R C |->+ | : | | 706 |Sender 2| | I E | | +------> R(n-1) | | 707 +--------+ | B | | | | | | 708 : | U | +--+--> R(n) | | | 709 : | T +-| | | | | 710 : | I | |<---------+ | | | 711 +--------+ | O |F|<---------------+ | | 712 |Media | | N |T|<--------------------+ | 713 |Sender M|<----->| | |<-------------------------+ 714 +--------+ +-----+ RTCP Unicast 716 FT = Feedback Target 717 Transport from the Feedback Target to the Distribution 718 Source is via unicast or multicast RTCP if they are not 719 co-located. 721 Figure 4: Point to Multipoint using Source Specific Multicast 723 In the SSM topology (Figure 4) a number of RTP sources (1 to M) are 724 allowed to send media to the SSM group. These send media to the 725 distribution source which then forwards the media streams to the 726 multicast group. The media streams reach the Receivers (R(1) to 727 R(n)). The Receivers' RTCP cannot be sent to the multicast group. 728 To support RTCP, an RTP extension for SSM [RFC5760] was defined to 729 use unicast transmission to send RTCP from the receivers to one or 730 more Feedback Targets (FT). 732 As multicast is a one to many distribution system, this must be taken 733 into consideration. For example, the only practical method for 734 adapting the bit-rate sent towards a given receiver for large groups 735 is to use a set of multicast groups, where each multicast group 736 represents a particular bit-rate. Otherwise the whole group gets 737 media adapted to the participant with the worst conditions. The 738 media encoding is either scalable, where multiple layers can be 739 combined, or simulcast where a single version is selected. By either 740 selecting or combing multicast groups, the receiver can control the 741 bit-rate sent on the path to itself. It is also common that streams 742 that improve transport robustness are sent in their own multicast 743 group to allow for interworking with legacy or to support different 744 levels of protection. 746 The result of this is some common behaviours for RTP multicast: 748 1. Multicast applications use a group of RTP sessions, not one. 749 Each endpoint will need to be a member of a number of RTP 750 sessions in order to perform well. 752 2. Within each RTP session, the number of media sinks is likely to 753 be much larger than the number of RTP sources. 755 3. Multicast applications need signalling functions to identify the 756 relationships between RTP sessions. 758 4. Multicast applications need signalling functions to identify the 759 relationships between SSRCs in different RTP sessions. 761 All multicast configurations share a signalling requirement; all of 762 the participants will need to have the same RTP and payload type 763 configuration. Otherwise, A could for example be using payload type 764 97 as the video codec H.264 while B thinks it is MPEG-2. It should 765 be noted that SDP offer/answer [RFC3264] has issues with ensuring 766 this property. The signalling aspects of multicast are not explored 767 further in this memo. 769 Security solutions for this type of group communications are also 770 challenging. First of all the key-management and the security 771 protocol must support group communication. Source authentication 772 becomes more difficult and requires special solutions. For more 773 discussion on this please review Options for Securing RTP Sessions 774 [I-D.ietf-avtcore-rtp-security-options]. 776 5.3. Point to Multipoint Using an RTP Transport Translator 778 This mode is described in section 3.3 of RFC 5117. 780 Transport Translators (Relays) result in an RTP session situation 781 that is very similar to how an ASM group RTP session would behave. 783 +---+ +------------+ +---+ 784 | A |<---->| |<---->| B | 785 +---+ | | +---+ 786 | Translator | 787 +---+ | | +---+ 788 | C |<---->| |<---->| D | 789 +---+ +------------+ +---+ 791 Figure 5: Transport Translator (Relay) 793 An RTP translator forwards both RTP and RTCP packets from all 794 participants to all other participants. 796 One of the most important aspects with the simple relay is that it is 797 only rewriting transport headers, no RTP modifications nor media 798 transcoding occur. The most obvious downside of this basic relaying 799 is that the translator has no control over how many streams need to 800 be delivered to a receiver. Nor can it simply select to deliver only 801 certain streams, as this creates session inconsistencies: If the 802 translator temporarily stops a stream, this prevents some receivers 803 from reporting on it. From the sender's perspective it will look 804 like a transport failure. Applications having needs to stop or 805 switch streams in the central node should consider using an RTP mixer 806 to avoid this issue. 808 The Transport Translator does not need to have an SSRC of itself, nor 809 does it need to send any RTCP reports on the flows that pass it. 810 This as the RTP source will receive feedback for the full path in the 811 RTCP being sent back. However the transport translator may choose to 812 send RTCP reports using its own SSRC, as if it itself contained a 813 media sink, in order to make information about the source-to- 814 translator link available to monitors. 816 Use of a transport translator results in that all the endpoints will 817 receive multiple SSRCs over a single unicast transport flow from the 818 translator. That is independent of whether the other endpoints have 819 only a single or several SSRCs. 821 The Transport Translator has the same signalling requirement as 822 multicast: All participants must have the same payload type 823 configuration. Also most of the ASM security issues also arise here. 824 Some alternative when it comes to solution do exist as there after 825 all exist a central node to communicate with. One that also can 826 enforce some security policies depending on the level of trust placed 827 in the node. 829 5.4. Point to Multipoint Using an RTP Mixer 831 An mixer (Figure 6) is a centralised node that selects or mixes 832 content in a conference to optimise the RTP session so that each 833 endpoint only needs connect to one entity, the mixer. The mixer can 834 also reduce the bit-rate needed from the mixer down to a conference 835 participants as the media sent from the mixer to the end-point can be 836 optimised in different ways. These optimisations include methods 837 like only choosing media from the currently most active speaker or 838 mixing together audio so that only one audio stream is required 839 instead of three in the depicted scenario (Figure 6). 841 +---+ +------------+ +---+ 842 | A |<---->| |<---->| B | 843 +---+ | | +---+ 844 | Mixer | 845 +---+ | | +---+ 846 | C |<---->| |<---->| D | 847 +---+ +------------+ +---+ 849 Figure 6: RTP Mixer 851 Mixers has some downsides, the first is that the mixer must be a 852 trusted node as they either performs media operations or at least 853 repacketize the media. Both type of operations requires when using 854 SRTP that the mixer verifies integrity, decrypts the content, perform 855 its operation and form new RTP packets, encrypts and integrity 856 protect them. This applies to all types of mixers described below. 857 The second downside is that all these operations and optimisation of 858 the session requires processing. How much depends on the 859 implementation as will become evident below. 861 A mixer, unlike a pure transport translator, is always application 862 specific: the application logic for stream mixing or stream selection 863 has to be embedded within the mixer, and controlled using application 864 specific signalling. The implementation of an mixer can take several 865 different forms and we will discuss the main themes available that 866 doesn't break RTP. 868 Please note that a Mixer could also contain translator 869 functionalities, like a media transcoder to adjust the media bit-rate 870 or codec used on a particular RTP media stream. 872 5.4.1. Media Mixing 874 This type of mixer is one which clearly can be called RTP mixer is 875 likely the one that most thinks of when they hear the term mixer. 876 Its basic pattern of operation is that it will receive the different 877 participants RTP media stream. Select which that are to be included 878 in a media domain mix of the incoming RTP media streams. Then create 879 a single outgoing stream from this mix. 881 The most commonly deployed media mixer is probably the audio mixer, 882 used in voice conferencing, where the output consists of some mixture 883 of all the input streams; this needs minimal signalling to be 884 successful. Audio mixing is straight forward and commonly possible 885 to do for a number of participants. Lets assume that you want to mix 886 N number of streams from different participants. Then the mixer need 887 to perform N decodings. Then it needs to produce N or N+1 mixes, the 888 reasons that different mixes are needed are so that each contributing 889 source get a mix which don't contain themselves, as this would result 890 in an echo. When N is lower than the number of all participants one 891 may produce a Mix of all N streams for the group that are currently 892 not included in the mix, thus N+1 mixes. These audio streams are 893 then encoded again, RTP packetised and sent out. 895 Video can't really be "mixed" and produce something particular useful 896 for the users, however creating an composition out of the contributed 897 video streams can be done. In fact it can be done in a number of 898 ways, tiling the different streams creating a chessboard, selecting 899 someone as more important and showing them large and a number of 900 other sources as smaller overlays is another. Also here one commonly 901 need to produce a number of different compositions so that the 902 contributing part doesn't need to see themselves. Then the mixer re- 903 encodes the created video stream, RTP packetise it and send it out 905 The problem with media mixing is that it both consume large amount of 906 media processing and encoding resources. The second is the quality 907 degradation created by decoding and re-encoding the RTP media stream. 908 Its advantage is that it is quite simplistic for the clients to 909 handle as they don't need to handle local mixing and composition. 911 +-A---------+ +-MIXER----------------------+ 912 | +-RTP1----| |-RTP1------+ +-----+ | 913 | | +-Audio-| |-Audio---+ | +---+ | | | 914 | | | AA1|--------->|---------+-+-|DEC|->| | | 915 | | | |<---------|MA1 <----+ | +---+ | | | 916 | | | | |(BA1+CA1)|\| +---+ | | | 917 | | +-------| |---------+ +-|ENC|<-| B+C | | 918 | +---------| |-----------+ +---+ | | | 919 +-----------+ | | | | 920 | | M | | 921 +-B---------+ | | E | | 922 | +-RTP2----| |-RTP2------+ | D | | 923 | | +-Audio-| |-Audio---+ | +---+ | I | | 924 | | | BA1|--------->|---------+-+-|DEC|->| A | | 925 | | | |<---------|MA2 <----+ | +---+ | | | 926 | | +-------| |(BA1+CA1)|\| +---+ | | | 927 | +---------| |---------+ +-|ENC|<-| A+C | | 928 +-----------+ |-----------+ +---+ | | | 929 | | M | | 930 +-C---------+ | | I | | 931 | +-RTP3----| |-RTP3------+ | X | | 932 | | +-Audio-| |-Audio---+ | +---+ | E | | 933 | | | CA1|--------->|---------+-+-|DEC|->| R | | 934 | | | |<---------|MA3 <----+ | +---+ | | | 935 | | +-------| |(BA1+CA1)|\| +---+ | | | 936 | +---------| |---------+ +-|ENC|<-| A+B | | 937 +-----------+ |-----------+ +---+ +-----+ | 938 +----------------------------+ 940 Figure 7: Session and SSRC details for Media Mixer 942 From an RTP perspective media mixing can be very straight forward as 943 can be seen in Figure 7. The mixer present one SSRC towards the peer 944 client, e.g. MA1 to Peer A, which is the media mix of the other 945 participants. As each peer receives a different version produced by 946 the mixer there are no actual relation between the different RTP 947 sessions in the actual media or the transport level information. 948 There is however one connection between RTP1-RTP3 in this figure. It 949 has to do with the SSRC space and the identity information. When A 950 receives the MA1 stream which is a combination of BA1 and CA1 streams 951 in the other PeerConnections RTP could enable the mixer to include 952 CSRC information in the MA1 stream to identify the contributing 953 source BA1 and CA1. 955 The CSRC has in its turn utility in RTP extensions, like the in Mixer 956 to Client audio levels RTP header extension [RFC6465]. If the SSRC 957 from endpoint to mixer leg are used as CSRC in another PeerConnection 958 then RTP1, RTP2 and RTP3 becomes one joint session as they have a 959 common SSRC space. At this stage the mixer also need to consider 960 which RTCP information it need to expose in the different legs. For 961 the above situation commonly nothing more than the Source Description 962 (SDES) information and RTCP BYE for CSRC need to be exposed. The 963 main goal would be to enable the correct binding against the 964 application logic and other information sources. This also enables 965 loop detection in the RTP session. 967 5.4.1.1. RTP Session Termination 969 There exist an possible implementation choice to have the RTP 970 sessions being separated between the different legs in the multi- 971 party communication session and only generate RTP media streams in 972 each without carrying on RTP/RTCP level any identity information 973 about the contributing sources. This removes both the functionality 974 that CSRC can provide and the possibility to use any extensions that 975 build on CSRC and the loop detection. It may appear a simplification 976 if SSRC collision would occur between two different end-points as 977 they can be avoided to be resolved and instead remapped between the 978 independent sessions if at all exposed. However, SSRC/CSRC remapping 979 requires that SSRC/CSRC are never used in the application level as 980 reference. This as they only have local importance, if they are used 981 on a multi-party session scope the result would be miss-referencing. 983 Session termination may appear to resolve some issues, it however 984 creates other issues that needs resolving, like loop detection, 985 identification of contributing sources and the need to handle mapped 986 identities and ensure that the right one is used towards the right 987 identities and never used directly between multiple end-points. 989 5.4.2. Media Switching 991 An RTP Mixer based on media switching avoids the media decoding and 992 encoding cycle in the mixer, but not the decryption and re-encryption 993 cycle as one rewrites RTP headers. This both reduces the amount of 994 computational resources needed in the mixer and increases the media 995 quality per transmitted bit. This is achieve by letting the mixer 996 have a number of SSRCs that represents conceptual or functional 997 streams the mixer produces. These streams are created by selecting 998 media from one of the by the mixer received RTP media streams and 999 forward the media using the mixers own SSRCs. The mixer can then 1000 switch between available sources if that is required by the concept 1001 for the source, like currently active speaker. 1003 To achieve a coherent RTP media stream from the mixer's SSRC the 1004 mixer is forced to rewrite the incoming RTP packet's header. First 1005 the SSRC field must be set to the value of the Mixer's SSRC. 1006 Secondly, the sequence number must be the next in the sequence of 1007 outgoing packets it sent. Thirdly the RTP timestamp value needs to 1008 be adjusted using an offset that changes each time one switch media 1009 source. Finally depending on the negotiation the RTP payload type 1010 value representing this particular RTP payload configuration may have 1011 to be changed if the different endpoint mixer legs have not arrived 1012 on the same numbering for a given configuration. This also requires 1013 that the different end-points do support a common set of codecs, 1014 otherwise media transcoding for codec compatibility is still 1015 required. 1017 Lets consider the operation of media switching mixer that supports a 1018 video conference with six participants (A-F) where the two latest 1019 speakers in the conference are shown to each participants. Thus the 1020 mixer has two SSRCs sending video to each peer. 1021 +-A---------+ +-MIXER----------------------+ 1022 | +-RTP1----| |-RTP1------+ +-----+ | 1023 | | +-Video-| |-Video---+ | | | | 1024 | | | AV1|------------>|---------+-+------->| S | | 1025 | | | |<------------|MV1 <----+-+-BV1----| W | | 1026 | | | |<------------|MV2 <----+-+-EV1----| I | | 1027 | | +-------| |---------+ | | T | | 1028 | +---------| |-----------+ | C | | 1029 +-----------+ | | H | | 1030 | | | | 1031 +-B---------+ | | M | | 1032 | +-RTP2----| |-RTP2------+ | A | | 1033 | | +-Video-| |-Video---+ | | T | | 1034 | | | BV1|------------>|---------+-+------->| R | | 1035 | | | |<------------|MV3 <----+-+-AV1----| I | | 1036 | | | |<------------|MV4 <----+-+-EV1----| X | | 1037 | | +-------| |---------+ | | | | 1038 | +---------| |-----------+ | | | 1039 +-----------+ | | | | 1040 : : : : 1041 : : : : 1042 +-F---------+ | | | | 1043 | +-RTP6----| |-RTP6------+ | | | 1044 | | +-Video-| |-Video---+ | | | | 1045 | | | CV1|------------>|---------+-+------->| | | 1046 | | | |<------------|MV11 <---+-+-AV1----| | | 1047 | | | |<------------|MV12 <---+-+-EV1----| | | 1048 | | +-------| |---------+ | | | | 1049 | +---------| |-----------+ +-----+ | 1050 +-----------+ +----------------------------+ 1052 Figure 8: Media Switching RTP Mixer 1054 The Media Switching RTP mixer can similar to the Media Mixing one 1055 reduce the bit-rate needed towards the different peers by selecting 1056 and switching in a sub-set of RTP media streams out of the ones it 1057 receives from the conference participations. 1059 To ensure that a media receiver can correctly decode the RTP media 1060 stream after a switch, it becomes necessary to ensure for state 1061 saving codecs that they start from default state at the point of 1062 switching. Thus one common tool for video is to request that the 1063 encoding creates an intra picture, something that isn't dependent on 1064 earlier state. This can be done using Full Intra Request [RFC5104] 1065 RTCP codec control message. 1067 Also in this type of mixer one could consider to terminate the RTP 1068 sessions fully between the different end-point and mixer legs. The 1069 same arguments and considerations as discussed in Section 5.4.1.1 1070 applies here. 1072 5.4.3. RTP Source Projecting 1074 Another method for handling media in the RTP mixer is to project all 1075 potential RTP sources (SSRCs) into a per end-point independent RTP 1076 session. The mixer can then select which of the potential sources 1077 that are currently actively transmitting media, despite that the 1078 mixer in another RTP session receives media from that end-point. 1079 This is similar to the media switching Mixer but have some important 1080 differences in RTP details. 1082 +-A---------+ +-MIXER---------------------+ 1083 | +-RTP1----| |-RTP1------+ +-----+ | 1084 | | +-Video-| |-Video---+ | | | | 1085 | | | AV1|------------>|---------+-+------>| | | 1086 | | | |<------------|BV1 <----+-+-------| S | | 1087 | | | |<------------|CV1 <----+-+-------| W | | 1088 | | | |<------------|DV1 <----+-+-------| I | | 1089 | | | |<------------|EV1 <----+-+-------| T | | 1090 | | | |<------------|FV1 <----+-+-------| C | | 1091 | | +-------| |---------+ | | H | | 1092 | +---------| |-----------+ | | | 1093 +-----------+ | | M | | 1094 | | A | | 1095 +-B---------+ | | T | | 1096 | +-RTP2----| |-RTP2------+ | R | | 1097 | | +-Video-| |-Video---+ | | I | | 1098 | | | BV1|------------>|---------+-+------>| X | | 1099 | | | |<------------|AV1 <----+-+-------| | | 1100 | | | |<------------|CV1 <----+-+-------| | | 1101 | | | | : : : |: : : : : : : : :| | | 1102 | | | |<------------|FV1 <----+-+-------| | | 1103 | | +-------| |---------+ | | | | 1104 | +---------| |-----------+ | | | 1105 +-----------+ | | | | 1106 : : : : 1107 : : : : 1108 +-F---------+ | | | | 1109 | +-RTP6----| |-RTP6------+ | | | 1110 | | +-Video-| |-Video---+ | | | | 1111 | | | CV1|------------>|---------+-+------>| | | 1112 | | | |<------------|AV1 <----+-+-------| | | 1113 | | | | : : : |: : : : : : : : :| | | 1114 | | | |<------------|EV1 <----+-+-------| | | 1115 | | +-------| |---------+ | | | | 1116 | +---------| |-----------+ +-----+ | 1117 +-----------+ +---------------------------+ 1119 Figure 9: Media Projecting Mixer 1121 So in this six participant conference depicted above in (Figure 9) 1122 one can see that end-point A will in this case be aware of 5 incoming 1123 SSRCs, BV1-FV1. If this mixer intend to have the same behaviour as 1124 in Section 5.4.2 where the mixer provides the end-points with the two 1125 latest speaking end-points, then only two out of these five SSRCs 1126 will concurrently transmit media to A. As the mixer selects which 1127 source in the different RTP sessions that transmit media to the end- 1128 points each RTP media stream will require some rewriting when being 1129 projected from one session into another. The main thing is that the 1130 sequence number will need to be consecutively incremented based on 1131 the packet actually being transmitted in each RTP session. Thus the 1132 RTP sequence number offset will change each time a source is turned 1133 on in a RTP session. 1135 As the RTP sessions are independent the SSRC numbers used can be 1136 handled independently also thus working around any SSRC collisions by 1137 having remapping tables between the RTP sessions. This will result 1138 that each endpoint may have a different view of the application usage 1139 of a particular SSRC. Thus the application must not use SSRC as 1140 references to RTP media streams when communicating with other peers 1141 directly. 1143 The mixer will also be responsible to act on any RTCP codec control 1144 requests coming from an end-point and decide if it can act on it 1145 locally or needs to translate the request into the RTP session that 1146 contains the media source. Both end-points and the mixer will need 1147 to implement conference related codec control functionalities to 1148 provide a good experience. Full Intra Request to request from the 1149 media source to provide switching points between the sources, 1150 Temporary Maximum Media Bit-rate Request (TMMBR) to enable the mixer 1151 to aggregate congestion control response towards the media source and 1152 have it adjust its bit-rate in case the limitation is not in the 1153 source to mixer link. 1155 This version of the mixer also puts different requirements on the 1156 end-point when it comes to decoder instances and handling of the RTP 1157 media streams providing media. As each projected SSRC can at any 1158 time provide media the end-point either needs to handle having thus 1159 many allocated decoder instances or have efficient switching of 1160 decoder contexts in a more limited set of actual decoder instances to 1161 cope with the switches. The WebRTC application also gets more 1162 responsibility to update how the media provides is to be presented to 1163 the user. 1165 5.5. Point to Multipoint using Multiple Unicast flows 1167 Based on the RTP session definition, it is clearly possible to have a 1168 joint RTP session over multiple transport flows like the below three 1169 endpoint joint session. In this case, A needs to send its' media 1170 streams and RTCP packets to both B and C over their respective 1171 transport flows. As long as all participants do the same, everyone 1172 will have a joint view of the RTP session. 1174 +---+ +---+ 1175 | A |<---->| B | 1176 +---+ +---+ 1177 ^ ^ 1178 \ / 1179 \ / 1180 v v 1181 +---+ 1182 | C | 1183 +---+ 1185 Figure 10: Point to Multi-Point using Multiple Unicast Transports 1187 This doesn't create any additional requirements beyond the need to 1188 have multiple transport flows associated with a single RTP session. 1189 Note that an endpoint may use a single local port to receive all 1190 these transport flows, or it might have separate local reception 1191 ports for each of the endpoints. 1193 There exists an alternative structure for establishing the above 1194 communication scenario (Figure 10) which uses independent RTP 1195 sessions between each pair of peers, i.e. three different RTP 1196 sessions. Unless independently adapted the same RTP media stream 1197 could be sent in both of the RTP sessions an endpoint has. The 1198 difference exists in the behaviours around RTCP, for example common 1199 RTCP bandwidth for one joint session, rather than three independent 1200 pools, and the awareness based on RTCP reports between the peers of 1201 how that third leg is doing. 1203 5.6. De-composite Endpoint 1205 The implementation of an application may desire to send a subset of 1206 the application's data to each of multiple devices, each with their 1207 own network address. A very basic use case for this would be to 1208 separate audio and video processing for a particular endpoint, like a 1209 conference room, into one device handling the audio and another 1210 handling the video, being interconnected by some control functions 1211 allowing them to behave as a single endpoint in all aspects except 1212 for transport Figure 11. 1214 Which decomposition that is possible is highly dependent on the RTP 1215 session usage. It is not really feasible to decomposed one logical 1216 end-point into two different transport node in one RTP session. From 1217 a third party monitor of such an attempt the two entities would look 1218 like two different end-points with a CNAME collision. This put a 1219 requirement on that the only type of de-composited endpoint that RTP 1220 really supports is one where the different parts have separate RTP 1221 sessions to send and/or receive media streams intended for them. 1223 +---------------------+ 1224 | Endpoint A | 1225 | Local Area Network | 1226 | +------------+ | 1227 | +->| Audio |<+----\ 1228 | | +------------+ | \ +------+ 1229 | | +------------+ | +-->| | 1230 | +->| Video |<+--------->| B | 1231 | | +------------+ | +-->| | 1232 | | +------------+ | / +------+ 1233 | +->| Control |<+----/ 1234 | +------------+ | 1235 +---------------------+ 1237 Figure 11: De-composite End-Point 1239 In the above usage, let us assume that the RTP sessions are different 1240 for audio and video. The audio and video parts will use a common 1241 CNAME and also have a common clock to ensure that synchronisation and 1242 clock drift handling works despite the decomposition. Also the RTCP 1243 handling works correctly as long as only one part of the de-composite 1244 is part of each RTP session. That way any differences in the path 1245 between A's audio entity and B and A's video and B are related to 1246 different SSRCs in different RTP sessions. 1248 The requirements that can derived from the above usage is that the 1249 transport flows for each RTP session might be under common control 1250 but still go to what looks like different endpoints based on 1251 addresses and ports. This geometry cannot be accomplished using one 1252 RTP session, so in this case, multiple RTP sessions are needed. 1254 6. Multiple Streams Discussion 1256 6.1. Introduction 1258 Using multiple media streams is a well supported feature of RTP. 1259 However, it can be unclear for most implementers or people writing 1260 RTP/RTCP applications or extensions attempting to apply multiple 1261 streams when it is most appropriate to add an additional SSRC in an 1262 existing RTP session and when it is better to use multiple RTP 1263 sessions. This section tries to discuss the various considerations 1264 needed. The next section then concludes with some guidelines. 1266 6.2. RTP/RTCP Aspects 1268 This section discusses RTP and RTCP aspects worth considering when 1269 selecting between SSRC multiplexing and Session multiplexing. 1271 6.2.1. The RTP Specification 1273 RFC 3550 contains some recommendations and a bullet list with 5 1274 arguments for different aspects of RTP multiplexing. Let's review 1275 Section 5.2 of [RFC3550], reproduced below: 1277 "For efficient protocol processing, the number of multiplexing points 1278 should be minimised, as described in the integrated layer processing 1279 design principle [ALF]. In RTP, multiplexing is provided by the 1280 destination transport address (network address and port number) which 1281 is different for each RTP session. For example, in a teleconference 1282 composed of audio and video media encoded separately, each medium 1283 SHOULD be carried in a separate RTP session with its own destination 1284 transport address. 1286 Separate audio and video streams SHOULD NOT be carried in a single 1287 RTP session and demultiplexed based on the payload type or SSRC 1288 fields. Interleaving packets with different RTP media types but 1289 using the same SSRC would introduce several problems: 1291 1. If, say, two audio streams shared the same RTP session and the 1292 same SSRC value, and one were to change encodings and thus 1293 acquire a different RTP payload type, there would be no general 1294 way of identifying which stream had changed encodings. 1296 2. An SSRC is defined to identify a single timing and sequence 1297 number space. Interleaving multiple payload types would require 1298 different timing spaces if the media clock rates differ and would 1299 require different sequence number spaces to tell which payload 1300 type suffered packet loss. 1302 3. The RTCP sender and receiver reports (see Section 6.4) can only 1303 describe one timing and sequence number space per SSRC and do not 1304 carry a payload type field. 1306 4. An RTP mixer would not be able to combine interleaved streams of 1307 incompatible media into one stream. 1309 5. Carrying multiple media in one RTP session precludes: the use of 1310 different network paths or network resource allocations if 1311 appropriate; reception of a subset of the media if desired, for 1312 example just audio if video would exceed the available bandwidth; 1313 and receiver implementations that use separate processes for the 1314 different media, whereas using separate RTP sessions permits 1315 either single- or multiple-process implementations. 1317 Using a different SSRC for each medium but sending them in the same 1318 RTP session would avoid the first three problems but not the last 1319 two. 1321 On the other hand, multiplexing multiple related sources of the same 1322 medium in one RTP session using different SSRC values is the norm for 1323 multicast sessions. The problems listed above don't apply: an RTP 1324 mixer can combine multiple audio sources, for example, and the same 1325 treatment is applicable for all of them. It may also be appropriate 1326 to multiplex streams of the same medium using different SSRC values 1327 in other scenarios where the last two problems do not apply." 1329 Let's consider one argument at a time. The first is an argument for 1330 using different SSRC for each individual media stream, which still is 1331 very applicable. 1333 The second argument is advocating against using payload type 1334 multiplexing, which still stands as can been seen by the extensive 1335 list of issues found in Appendix A. 1337 The third argument is yet another argument against payload type 1338 multiplexing. 1340 The fourth is an argument against multiplexing media streams that 1341 require different handling into the same session. As we saw in the 1342 discussion of RTP mixers, the RTP mixer has to embed application 1343 logic in order to handle streams anyway; the separation of streams 1344 according to stream type is just another piece of application logic, 1345 which may or may not be appropriate for a particular application. A 1346 type of application that can mix different media sources "blindly" is 1347 the telephone bridge; most other type of application needs 1348 application-specific logic to perform the mix correctly. 1350 The fifth argument discusses network aspects that we will discuss 1351 more below in Section 6.4. It also goes into aspects of 1352 implementation, like decomposed endpoints where different processes 1353 or inter-connected devices handle different aspects of the whole 1354 multi-media session. 1356 A summary of RFC 3550's view on multiplexing is to use unique SSRCs 1357 for anything that is its own media/packet stream, and to use 1358 different RTP sessions for media streams that don't share media type. 1359 The first this document support as very valid. The later is one 1360 thing which is further discussed in this document as something the 1361 application developer needs to make a continuous choice for. 1363 6.2.1.1. Different Media Types Recommendations 1365 The above quote from RTP [RFC3550] includes a strong recommendation: 1367 "For example, in a teleconference composed of audio and video 1368 media encoded separately, each medium SHOULD be carried in a 1369 separate RTP session with its own destination transport address." 1371 It was identified in "Why RTP Sessions Should Be Content Neutral" 1372 [I-D.alvestrand-rtp-sess-neutral] that the above statement is poorly 1373 supported by any of the motivations provided in the RTP 1374 specification. This has resulted in the creation of a specification 1375 Multiple Media Types in an RTP Session specification 1376 [I-D.westerlund-avtcore-multi-media-rtp-session] which intend to 1377 update this recommendation. That document has a detailed analysis of 1378 the potential issues in having multiple media types in the same RTP 1379 session. This document tries to provide an more over arching 1380 consideration regarding the usage of RTP session and considers 1381 multiple media types in one RTP session as possible choice for the 1382 RTP application designer. 1384 6.2.2. Multiple SSRCs in a Session 1386 In cases when an endpoint uses multiple SSRCs, we have found two 1387 closely related issues. The first is if every SSRC shall report on 1388 all other SSRC, even the ones originating from the same endpoint. 1389 The reason for this would be to ensure that no monitoring function 1390 should suspect a breakage in the RTP session. No monitoring function 1391 that gives an alert on non-reporting of an endpoint's own SSRCs has 1392 been identified. 1394 The second issue around RTCP reporting arise when an endpoint 1395 receives one or more media streams, and when the receiving endpoint 1396 itself sends multiple SSRC in the same RTP session. As transport 1397 statistics are gathered per endpoint and shared between the nodes, 1398 all the endpoint's SSRC will report based on the same received data, 1399 the only difference will be which SSRCs sends the report. This could 1400 be considered unnecessary overhead, but for consistency it might be 1401 simplest to always have all sending SSRCs send RTCP reports on all 1402 media streams the endpoint receives. 1404 The current RTP text is silent about sending RTCP Receiver Reports 1405 for an endpoint's own sources, but does not preclude either sending 1406 or omitting them. The uncertainty in the expected behaviour in those 1407 cases has likely caused variations in the implementation strategy. 1408 This could cause an interoperability issue where it is not possible 1409 to determine if the lack of reports is a true transport issue, or 1410 simply a result of implementation. 1412 Although this issue is valid already for the simple point to point 1413 case, it needs to be considered in all topologies. From the 1414 perspective of an endpoint, any solution needs to take into account 1415 what a particular endpoint can determine without explicit information 1416 of the topology. For example, a Transport Translator (Relay) 1417 topology will look quite similar to point to point on a transport 1418 level but is different on RTP level. Assume a first scenario with 1419 two SSRC being sent from an endpoint to a Transport Translator, and a 1420 second scenario with two single SSRC remote endpoints sending to the 1421 same Transport Translator. The main differences between those two 1422 scenarios are that in the second scenario, the RTT may vary between 1423 the SSRCs (but it is not guaranteed), and the SSRCs may also have 1424 different CNAMEs. 1426 When an endpoint has multiple SSRCs and it needs to send RTCP packets 1427 on behalf of these SSRCs, the question arises how RTCP packets with 1428 different source SSRCs can be sent in the same compound packet. It 1429 appears allowed, however some consideration of the transmission 1430 scheduling is needed. 1432 These issues are currently being discussed and a recommendation for 1433 how to handle them are developed in "Real-Time Transport Protocol 1434 (RTP) Considerations for Endpoints Sending Multiple Media Streams" 1435 [I-D.lennox-avtcore-rtp-multi-stream]. 1437 6.2.3. Handling Varying sets of Senders 1439 In some applications, the set of simultaneously active sources varies 1440 within a larger set of session members. A receiver can then possibly 1441 try to use a set of decoding chains that is smaller than the number 1442 of senders, switching the decoding chains between different senders. 1443 As each media decoding chain may contain state, either the receiver 1444 must either be able to save the state of swapped-out senders, or the 1445 sender must be able to send data that permits the receiver to 1446 reinitialise when it resumes activity. 1448 This behaviour will cause similar issues independent of SSRC or 1449 Session multiplexing. 1451 6.2.4. Cross Session RTCP Requests 1453 There currently exists no functionality to make truly synchronised 1454 and atomic RTCP messages with some type of request semantics across 1455 multiple RTP Sessions. Instead, separate RTCP messages will have to 1456 be sent in each session. This gives SSRC multiplexed streams a 1457 slight advantage as RTCP messages for different streams in the same 1458 session can be sent in a compound RTCP packet. Thus providing an 1459 atomic operation if different modifications of different streams are 1460 requested at the same time. 1462 In Session multiplexed cases, the RTCP timing rules in the sessions 1463 and the transport aspects, such as packet loss and jitter, prevents a 1464 receiver from relying on atomic operations, forcing it to use more 1465 robust and forgiving mechanisms. 1467 6.2.5. Binding Related Sources 1469 A common problem in a number of various RTP extensions has been how 1470 to bind related sources together. This issue is common to SSRC 1471 multiplexing and Session Multiplexing. 1473 Most, if not all, solutions to this problem are implemented in the 1474 signalling plane, providing metadata information using SDP. 1476 There exists one solution for grouping RTP sessions together in SDP 1477 [RFC5888] to know which RTP session contains for example the FEC data 1478 for the source data in another session. However, this mechanism does 1479 not work on individual media flows and is thus not directly 1480 applicable to the problem. The other solution is also SDP based and 1481 can group SSRCs within a single RTP session [RFC5576]. Thus this 1482 mechanism can bind media streams in SSRC multiplexed cases. Both 1483 solutions have the shortcoming of being restricted to SDP based 1484 signalling and also do not work in cases where the session's dynamic 1485 properties are such that it is difficult or resource consuming to 1486 keep the list of related SSRCs up to date. 1488 One possible solution could be to mandate the same SSRC value being 1489 used in all RTP session in case of session multiplexing. We do note 1490 that Section 8.3 of the RTP Specification [RFC3550] recommends using 1491 a single SSRC space across all RTP sessions for layered coding. 1492 However this recommendation has some downsides and is less applicable 1493 beyond the field of layered coding. To use the same sender SSRC in 1494 all RTP sessions from a particular endpoint can cause issues if an 1495 SSRC collision occurs. If the same SSRC is used as the required 1496 binding between the streams, then all streams in the related RTP 1497 sessions must change their SSRC. This is extra likely to cause 1498 problems if the participant populations are different in the 1499 different sessions. For example, in case of large number of 1500 receivers having selected totally random SSRC values in each RTP 1501 session as RFC 3550 specifies, a change due to a SSRC collision in 1502 one session can then cause a new collision in another session. This 1503 cascading effect is not severe but there is an increased risk that 1504 this occurs for well populated sessions (the birthday paradox ensures 1505 that if you populate a single session with 9292 SSRCs at random, the 1506 chances are approximately 1% that at least one collision will occur). 1507 In addition, being forced to change the SSRC affects all the related 1508 media streams; instead of having to resynchronise only the originally 1509 conflicting stream, all streams will suddenly need to be 1510 resynchronised with each other. This will prevent also the media 1511 streams not having an actual collision from being usable during the 1512 resynchronisation and also increases the time until synchronisation 1513 is finalised. In addition, it requires exception handling in the 1514 SSRC generation. 1516 The above collision issue does not occur in case of having only one 1517 SSRC space across all sessions and all participants will be part of 1518 at least one session, like the base layer in layered encoding. In 1519 that case the only downside is the special behaviour that needs to be 1520 well defined by anyone using this. But, having an exception 1521 behaviour where the SSRC space is common across all session is an 1522 issue as this behaviour does not fit all the RTP extensions or 1523 payload formats. It is possible to create a situation where the 1524 different mechanisms cannot be combined due to the non standard SSRC 1525 allocation behaviour. 1527 Existing mechanisms with known issues: 1529 RTP Retransmission: [RFC4588] Has two modes, one for SSRC 1530 multiplexing and one for Session multiplexing. The session 1531 multiplexing requires the same CNAME and mandates that the same 1532 SSRC is used in both sessions. Using the same SSRC does work but 1533 will as previously stated potentially have issues in certain 1534 cases. In SSRC multiplexed mode the CNAME is used to bind media 1535 and retransmission streams together. However, if multiple media 1536 streams are sent from the same endpoint in the same session this 1537 does not provide non-ambiguous binding. Therefore when the first 1538 retransmission request for a media stream is sent, one must not 1539 have another retransmission request outstanding for an SSRC which 1540 don't have a binding between the original SSRC and the 1541 retransmission stream's SSRC. This works but creates some 1542 limitations that can be avoided by a explicit mechanism. The SDP 1543 based ssrc-group mechanism would be sufficient in this case as 1544 long as the application can rely on the signalling based solution. 1546 Scalable Video Coding : As an example of scalable coding, SVC 1547 [RFC6190] has various modes. The Multi Session Transmission (MST) 1548 uses Session multiplexing to separate scalability layers. 1549 However, this specification has failed to be explicit on how these 1550 layers are bound together in cases where CNAME is not sufficient. 1551 CNAME is no longer sufficient when more than one media source 1552 occur within a session that has the same CNAME, for example due to 1553 multiple video cameras capturing the same lecture hall. This 1554 likely implies that a single SSRC space as recommend by Section 1555 8.3 of RTP [RFC3550] is to be used. 1557 Forward Error Correction: If some type of FEC or redundancy stream 1558 is being sent, it needs its own SSRC, with the exception of 1559 constructions like redundancy encoding [RFC2198]. Thus in case of 1560 transmitting the FEC in the same session as the source data, the 1561 inter SSRC relation within a session is needed. In case of 1562 sending the redundant data in a separate session from the source, 1563 the SSRC in each session needs to be related. This occurs for 1564 example in RFC5109 when using session separation of original and 1565 FEC data. SSRC multiplexing is not supported, only using 1566 redundant encoding is supported. 1568 This issue appears to need action to harmonise and avoid future 1569 shortcomings in extension specifications. A proposed solution for 1570 handling this issue is [I-D.westerlund-avtext-rtcp-sdes-srcname]. 1572 6.2.6. Forward Error Correction 1574 There exist a number of Forward Error Correction (FEC) based schemes 1575 for how to reduce the packet loss of the original streams. Most of 1576 the FEC schemes will protect a single source flow. The protection is 1577 achieved by transmitting a certain amount of redundant information 1578 that is encoded such that it can repair one or more packet loss over 1579 the set of packets they protect. This sequence of redundant 1580 information also needs to be transmitted as its own media stream, or 1581 in some cases instead of the original media stream. Thus many of 1582 these schemes create a need for binding the related flows as 1583 discussed above. They also create additional flows that need to be 1584 transported. Looking at the history of these schemes, there is both 1585 SSRC multiplexed and Session multiplexed solutions and some schemes 1586 that support both. 1588 Using a Session multiplexed solution supports the case where some set 1589 of receivers may not be able to utilise the FEC information. By 1590 placing it in a separate RTP session, it can easily be ignored. 1592 In usages involving multicast, having the FEC information on its own 1593 multicast group, and therefore in its own RTP session, allows for 1594 flexibility, for example when using Rapid Acquisition of Multicast 1595 Groups (RAMS) [RFC6285]. During the RAMS burst where data is 1596 received over unicast and where it is possible to combine with 1597 unicast based retransmission [RFC4588], there is no need to burst the 1598 FEC data related to the burst of the source media streams needed to 1599 catch up with the multicast group. This saves bandwidth to the 1600 receiver during the burst, enabling quicker catch up. When the 1601 receiver has caught up and joins the multicast group(s) for the 1602 source, it can at the same time join the multicast group with the FEC 1603 information. Having the source stream and the FEC in separate groups 1604 allows for easy separation in the Burst/Retransmission Source (BRS) 1605 without having to individually classify packets. 1607 6.2.7. Transport Translator Sessions 1609 A basic Transport Translator relays any incoming RTP and RTCP packets 1610 to the other participants. The main difference between SSRC 1611 multiplexing and Session multiplexing resulting from this use case is 1612 that for SSRC multiplexing it is not possible for a particular 1613 session participant to decide to receive a subset of media streams. 1614 When using separate RTP sessions for the different sets of media 1615 streams, a single participant can choose to leave one of the sessions 1616 but not the other. 1618 6.3. Interworking 1620 There are several different kinds of interworking, and this section 1621 discusses two related ones. The interworking between different 1622 applications and the implications of potentially different choices of 1623 usage of RTP's multiplexing points. The second topic relates to what 1624 limitations may have to be considered working with some legacy 1625 applications. 1627 6.3.1. Types of Interworking 1629 It is not uncommon that applications or services of similar usage, 1630 especially the ones intended for interactive communication, ends up 1631 in a situation where one want to interconnect two or more of these 1632 applications. 1634 In these cases one ends up in a situation where one might use a 1635 gateway to interconnect applications. This gateway then needs to 1636 change the multiplexing structure or adhere to limitations in each 1637 application. 1639 There are two fundamental approaches to gatewaying: RTP bridging, 1640 where the gateway acts as an RTP Translator, and the two applications 1641 are members of the same RTP session, and RTP termination, where there 1642 are independent RTP sessions running from each interconnected 1643 application to the gateway. 1645 6.3.2. RTP Translator Interworking 1647 From an RTP perspective the RTP Translator approach could work if all 1648 the applications are using the same codecs with the same payload 1649 types, have made the same multiplexing choices, have the same 1650 capabilities in number of simultaneous media streams combined with 1651 the same set of RTP/RTCP extensions being supported. Unfortunately 1652 this may not always be true. 1654 When one is gatewaying via an RTP Translator, a natural requirement 1655 is that the two applications being interconnected must use the same 1656 approach to multiplexing. Furthermore, if one of the applications is 1657 capable of working in several modes (such as being able to use SSRC 1658 multiplexing or RTP session multiplexing at will), and the other one 1659 is not, successful interconnection depends on locking the more 1660 flexible application into the operating mode where interconnection 1661 can be successful, even if no participants using the less flexible 1662 application are present when the RTP sessions are being created. 1664 6.3.3. Gateway Interworking 1666 When one terminates RTP sessions at the gateway, there are certain 1667 tasks that the gateway must carry out: 1669 o Generating appropriate RTCP reports for all media streams 1670 (possibly based on incoming RTCP reports), originating from SSRCs 1671 controlled by the gateway. 1673 o Handling SSRC collision resolution in each application's RTP 1674 sessions. 1676 o Signalling, choosing and policing appropriate bit-rates for each 1677 session. 1679 If either of the applications has any security applied, e.g. in the 1680 form of SRTP, the gateway must be able to decrypt incoming packets 1681 and re-encrypt them in the other application's security context. 1682 This is necessary even if all that's required is a simple remapping 1683 of SSRC numbers. If this is done, the gateway also needs to be a 1684 member of the security contexts of both sides, of course. 1686 Other tasks a gateway may need to apply include transcoding (for 1687 incompatible codec types), rescaling (for incompatible video size 1688 requirements), suppression of content that is known not to be handled 1689 in the destination application, or the addition or removal of 1690 redundancy coding or scalability layers to fit the need of the 1691 destination domain. 1693 From the above, we can see that the gateway needs to have an intimate 1694 knowledge of the application requirements; a gateway is by its nature 1695 application specific, not a commodity product. 1697 This fact reveals the potential for these gateways to block evolution 1698 of the applications by blocking unknown RTP and RTCP extensions that 1699 the regular application has been extended with. 1701 If one uses security functions, like SRTP, they can as seen above 1702 incur both additional risk due to the gateway needing to be in 1703 security association between the endpoints, unless the gateway is on 1704 the transport level, and additional complexities in form of the 1705 decrypt-encrypt cycles needed for each forwarded packet. SRTP, due 1706 to its keying structure, also requires that each RTP session must 1707 have different master keys, as use of the same key in two RTP 1708 sessions can result in two-time pads that completely breaks the 1709 confidentiality of the packets. 1711 6.3.4. Multiple SSRC Legacy Considerations 1713 Historically, the most common RTP use cases have been point to point 1714 Voice over IP (VoIP) or streaming applications, commonly with no more 1715 than one media source per endpoint and media type (typically audio 1716 and video). Even in conferencing applications, especially voice 1717 only, the conference focus or bridge has provided a single stream 1718 with a mix of the other participants to each participant. It is also 1719 common to have individual RTP sessions between each endpoint and the 1720 RTP mixer, meaning that the mixer functions as an RTP-terminating 1721 gateway. 1723 When establishing RTP sessions that may contain endpoints that aren't 1724 updated to handle multiple streams following these recommendations, a 1725 particular application can have issues with multiple SSRCs within a 1726 single session. These issues include: 1728 1. Need to handle more than one stream simultaneously rather than 1729 replacing an already existing stream with a new one. 1731 2. Be capable of decoding multiple streams simultaneously. 1733 3. Be capable of rendering multiple streams simultaneously. 1735 This indicates that gateways attempting to interconnect to this class 1736 of devices must make sure that only one media stream of each type 1737 gets delivered to the endpoint if it's expecting only one, and that 1738 the multiplexing format is what the device expects. It is highly 1739 unlikely that RTP translator-based interworking can be made to 1740 function successfully in such a context. 1742 6.4. Network Aspects 1744 The multiplexing choice has impact on network level mechanisms that 1745 need to be considered by the implementor. 1747 6.4.1. Quality of Service 1749 When it comes to Quality of Service mechanisms, they are either flow 1750 based or marking based. RSVP [RFC2205] is an example of a flow based 1751 mechanism, while Diff-Serv [RFC2474] is an example of a Marking based 1752 one. For a marking based scheme, the method of multiplexing will not 1753 affect the possibility to use QoS. 1755 However, for a flow based scheme there is a clear difference between 1756 the methods. SSRC multiplexing will result in all media streams 1757 being part of the same 5-tuple (protocol, source address, destination 1758 address, source port, destination port) which is the most common 1759 selector for flow based QoS. Thus, separation of the level of QoS 1760 between media streams is not possible. That is however possible for 1761 session based multiplexing, where each media stream for which a 1762 separate QoS handling is desired can be in a different RTP session 1763 that can be sent over different 5-tuples. 1765 6.4.2. NAT and Firewall Traversal 1767 In today's network there exist a large number of middleboxes. The 1768 ones that normally have most impact on RTP are Network Address 1769 Translators (NAT) and Firewalls (FW). 1771 Below we analyze and comment on the impact of requiring more 1772 underlying transport flows in the presence of NATs and Firewalls: 1774 End-Point Port Consumption: A given IP address only has 65536 1775 available local ports per transport protocol for all consumers of 1776 ports that exist on the machine. This is normally never an issue 1777 for an end-user machine. It can become an issue for servers that 1778 handle large number of simultaneous streams. However, if the 1779 application uses ICE to authenticate STUN requests, a server can 1780 serve multiple endpoints from the same local port, and use the 1781 whole 5-tuple (source and destination address, source and 1782 destination port, protocol) as identifier of flows after having 1783 securely bound them to the remote endpoint address using the STUN 1784 request. In theory the minimum number of media server ports 1785 needed are the maximum number of simultaneous RTP Sessions a 1786 single endpoint may use. In practice, implementation will 1787 probably benefit from using more server ports to simplify 1788 implementation or avoid performance bottlenecks. 1790 NAT State: If an endpoint sits behind a NAT, each flow it generates 1791 to an external address will result in a state that has to be kept 1792 in the NAT. That state is a limited resource. In home or Small 1793 Office/Home Office (SOHO) NATs, memory or processing are usually 1794 the most limited resources. For large scale NATs serving many 1795 internal endpoints, available external ports are typically the 1796 scarce resource. Port limitations is primarily a problem for 1797 larger centralised NATs where endpoint independent mapping 1798 requires each flow to use one port for the external IP address. 1799 This affects the maximum number of internal users per external IP 1800 address. However, it is worth pointing out that a real-time video 1801 conference session with audio and video is likely using less than 1802 10 UDP flows, compared to certain web applications that can use 1803 100+ TCP flows to various servers from a single browser instance. 1805 NAT Traversal Excess Time: Making the NAT/FW traversal takes a 1806 certain amount of time for each flow. It also takes time in a 1807 phase of communication between accepting to communicate and the 1808 media path being established which is fairly critical. The best 1809 case scenario for how much extra time it takes after finding the 1810 first valid candidate pair following the specified ICE procedures 1811 are: 1.5*RTT + Ta*(Additional_Flows-1), where Ta is the pacing 1812 timer, which ICE specifies to be no smaller than 20 ms. That 1813 assumes a message in one direction, and then an immediate 1814 triggered check back. The reason it isn't more is that ICE first 1815 finds one candidate pair that works prior to attempting to 1816 establish multiple flows. Thus, there is no extra time until one 1817 has found a working candidate pair. Based on that working pair 1818 the needed extra time is to in parallel establish the, in most 1819 cases 2-3, additional flows. However, packet loss causes extra 1820 delays, at least 100 ms which is the minimal retransmission timer 1821 for ICE. 1823 NAT Traversal Failure Rate: Due to the need to establish more than a 1824 single flow through the NAT, there is some risk that establishing 1825 the first flow succeeds but that one or more of the additional 1826 flows fail. The risk that this happens is hard to quantify, but 1827 it should be fairly low as one flow from the same interfaces has 1828 just been successfully established. Thus only rare events such as 1829 NAT resource overload, or selecting particular port numbers that 1830 are filtered etc, should be reasons for failure. 1832 Deep Packet Inspection and Multiple Streams: Firewalls differ in how 1833 deeply they inspect packets. There exist some potential that 1834 deeply inspecting firewalls will have similar legacy issues with 1835 multiple SSRCs as some stack implementations. 1837 SSRC multiplexing keeps the additional media streams within one RTP 1838 Session and does not introduce any additional NAT traversal 1839 complexities per media stream. This can be compared with normally 1840 one or two additional transport flows per RTP session when using 1841 session multiplexing. Additional lower layer transport flows will be 1842 required, unless an explicit de-multiplexing layer is added between 1843 RTP and the transport protocol. A proposal for how to multiplex 1844 multiple RTP sessions over the same single lower layer transport 1845 exist in [I-D.westerlund-avtcore-transport-multiplexing]. 1847 6.4.3. Multicast 1849 Multicast groups provides a powerful semantics for a number of real- 1850 time applications, especially the ones that desire broadcast-like 1851 behaviours with one endpoint transmitting to a large number of 1852 receivers, like in IPTV. But that same semantics do result in a 1853 certain number of limitations. 1855 One limitation is that for any group, sender side adaptation to the 1856 actual receiver properties causes degradation for all participants to 1857 what is supported by the receiver with the worst conditions among the 1858 group participants. In most cases this is not acceptable. Instead 1859 various receiver based solutions are employed to ensure that the 1860 receivers achieve best possible performance. By using scalable 1861 encoding and placing each scalability layer in a different multicast 1862 group, the receiver can control the amount of traffic it receives. 1863 To have each scalability layer on a different multicast group, one 1864 RTP session per multicast group is used. 1866 RTP can't function correctly if media streams sent over different 1867 multicast groups where considered part of the same RTP session. 1868 First of all the different layers needs different SSRCs or the 1869 sequence number space seen for a receiver of any sub set of the 1870 layers would have sender side holes. Thus triggering packet loss 1871 reactions. Also any RTCP reporting of such a session would be non 1872 consistent and making it difficult for the sender to determine the 1873 sessions actual state. 1875 Thus it appears easiest and most straightforward to use multiple RTP 1876 sessions. In addition, the transport flow considerations in 1877 multicast are a bit different from unicast. First of all there is no 1878 shortage of port space, as each multicast group has its own port 1879 space. 1881 6.4.4. Multiplexing multiple RTP Session on a Single Transport 1883 For applications that doesn't need flow based QoS and like to save 1884 ports and NAT/FW traversal costs and where usage of multiple media 1885 types in one RTP session is not suitable, there is a proposal for how 1886 to achieve multiplexing of multiple RTP sessions over the same lower 1887 layer transport [I-D.westerlund-avtcore-transport-multiplexing]. 1888 Using such a solution would allow session multiplexing without most 1889 of the perceived downsides of additional RTP sessions creating a need 1890 for additional transport flows. 1892 6.5. Security Aspects 1894 When dealing with point-to-point, 2-member RTP sessions only, there 1895 are few security issues that are relevant to the choice of having one 1896 RTP session or multiple RTP sessions. However, there are a few 1897 aspects of multiparty sessions that might warrant consideration. 1899 6.5.1. Security Context Scope 1901 When using SRTP [RFC3711] the security context scope is important and 1902 can be a necessary differentiation in some applications. As SRTP's 1903 crypto suites (so far) is built around symmetric keys, the receiver 1904 will need to have the same key as the sender. This results in that 1905 no one in a multi-party session can be certain that a received packet 1906 really was sent by the claimed sender or by another party having 1907 access to the key. In most cases this is a sufficient security 1908 property, but there are a few cases where this does create 1909 situations. 1911 The first case is when someone leaves a multi-party session and one 1912 wants to ensure that the party that left can no longer access the 1913 media streams. This requires that everyone re-keys without 1914 disclosing the keys to the excluded party. 1916 A second case is when using security as an enforcing mechanism for 1917 differentiation. Take for example a scalable layer or a high quality 1918 simulcast version which only premium users are allowed to access. 1919 The mechanism preventing a receiver from getting the high quality 1920 stream can be based on the stream being encrypted with a key that 1921 user can't access without paying premium, having the key-management 1922 limit access to the key. 1924 SRTP [RFC3711] has not special functions for dealing with different 1925 sets of master keys for different SSRCs. The key-management 1926 functions has different capabilities to establish different set of 1927 keys, normally on a per end-point basis. DTLS-SRTP [RFC5764] and 1928 Security Descriptions [RFC4568] for example establish different keys 1929 for outgoing and incoming traffic from an end-point. This key usage 1930 must be written into the cryptographic context, possibly associated 1931 with different SSRCs. 1933 6.5.2. Key Management for Multi-party session 1935 Performing key-management for multi-party session can be a challenge. 1936 This section considers some of the issues. 1938 Multi-party sessions, such as transport translator based sessions and 1939 multicast sessions, cannot use Security Description [RFC4568] nor 1940 DTLS-SRTP [RFC5764] without an extension as each endpoint provides 1941 its set of keys. In centralised conference, the signalling 1942 counterpart is a conference server and the media plane unicast 1943 counterpart (to which DTLS messages would be sent) is the transport 1944 translator. Thus an extension like Encrypted Key Transport 1945 [I-D.ietf-avt-srtp-ekt] is needed or a MIKEY [RFC3830] based solution 1946 that allows for keying all session participants with the same master 1947 key. 1949 6.5.3. Complexity Implications 1951 The usage of security functions can surface complexity implications 1952 of the choice of multiplexing and topology. This becomes especially 1953 evident in RTP topologies having any type of middlebox that processes 1954 or modifies RTP/RTCP packets. Where there is very small overhead for 1955 an RTP translator or mixer to rewrite an SSRC value in the RTP packet 1956 of an unencrypted session, the cost of doing it when using 1957 cryptographic security functions is higher. For example if using 1958 SRTP [RFC3711], the actual security context and exact crypto key are 1959 determined by the SSRC field value. If one changes it, the 1960 encryption and authentication tag must be performed using another 1961 key. Thus changing the SSRC value implies a decryption using the old 1962 SSRC and its security context followed by an encryption using the new 1963 one. 1965 7. Arch-Types 1967 This section discusses some arch-types of how RTP multiplexing can be 1968 used in applications to achieve certain goals and a summary of their 1969 implications. For each arch-type there is discussion of benefits and 1970 downsides. 1972 7.1. Single SSRC per Session 1974 In this arch-type each endpoint in a point-to-point session has only 1975 a single SSRC, thus the RTP session contains only two SSRCs, one 1976 local and one remote. This session can be used both unidirectional, 1977 i.e. only a single media stream or bi-directional, i.e. both 1978 endpoints have one media stream each. If the application needs 1979 additional media flows between the endpoints, they will have to 1980 establish additional RTP sessions. 1982 The Pros: 1984 1. This arch-type has great legacy interoperability potential as it 1985 will not tax any RTP stack implementations. 1987 2. The signalling has good possibilities to negotiate and describe 1988 the exact formats and bit-rates for each media stream, especially 1989 using today's tools in SDP. 1991 3. It does not matter if usage or purpose of the media stream is 1992 signalled on media stream level or session level as there is no 1993 difference. 1995 4. It is possible to control security association per RTP session 1996 with current key-management. 1998 The Cons: 2000 a. The number of required RTP sessions cannot really be higher, 2001 which has the implications: 2003 * Linear growth of the amount of NAT/FW state with number of 2004 media streams. 2006 * Increased delay and resource consumption from NAT/FW 2007 traversal. 2009 * Likely larger signalling message and signalling processing 2010 requirement due to the amount of session related information. 2012 * Higher potential for a single media stream to fail during 2013 transport between the endpoints. 2015 b. When the number of RTP sessions grows, the amount of explicit 2016 state for relating media stream also grows, linearly or possibly 2017 exponentially, depending on how the application needs to relate 2018 media streams. 2020 c. The port consumption may become a problem for centralised 2021 services, where the central node's port consumption grows rapidly 2022 with the number of sessions. 2024 d. For applications where the media streams are highly dynamic in 2025 their usage, i.e. entering and leaving, the amount of signalling 2026 can grow high. Issues arising from the timely establishment of 2027 additional RTP sessions can also arise. 2029 e. Cross session RTCP requests needs is likely to exist and may 2030 cause issues. 2032 f. If the same SSRC value is reused in multiple RTP sessions rather 2033 than being randomly chosen, interworking with applications that 2034 uses another multiplexing structure than this application will 2035 have issues and require SSRC translation. 2037 g. Cannot be used with Any Source Multicast (ASM) as one cannot 2038 guarantee that only two endpoints participate as packet senders. 2039 Using SSM, it is possible to restrict to these requirements if no 2040 RTCP feedback is used. 2042 h. For most security mechanisms, each RTP session or transport flow 2043 requires individual key-management and security association 2044 establishment thus increasing the overhead. 2046 i. Does not support multiparty session within a session. Instead 2047 each multi-party participant will require an individual RTP 2048 session to a given endpoint, even if a central node is used. 2050 RTP applications that need to inter-work with legacy RTP 2051 applications, like VoIP and video conferencing, can potentially 2052 benefit from this structure. However, a large number of media 2053 descriptions in SDP can also run into issues with existing 2054 implementations. For any application needing a larger number of 2055 media flows, the overhead can become very significant. This 2056 structure is also not suitable for multi-party sessions, as any given 2057 media stream from each participant, although having same usage in the 2058 application, must have its own RTP session. In addition, the dynamic 2059 behaviour that can arise in multi-party applications can tax the 2060 signalling system and make timely media establishment more difficult. 2062 7.2. Multiple SSRCs of the Same Media Type 2064 In this arch-type, each RTP session serves only a single media type. 2065 The RTP session can contain multiple media streams, either from a 2066 single endpoint or due to multiple endpoints. This commonly creates 2067 a low number of RTP sessions, typically only two one for audio and 2068 one for video with a corresponding need for two listening ports when 2069 using RTP and RTCP multiplexing. 2071 The Pros: 2073 1. Low number of RTP sessions needed compared to single SSRC case. 2074 This implies: 2076 * Reduced NAT/FW state 2078 * Lower NAT/FW Traversal Cost in both processing and delay. 2080 2. Allows for early de-multiplexing in the processing chain in RTP 2081 applications where all media streams of the same type have the 2082 same usage in the application. 2084 3. Works well with media type de-composite endpoints. 2086 4. Enables Flow-based QoS with different prioritisation between 2087 media types. 2089 5. For applications with dynamic usage of media streams, i.e. they 2090 come and go frequently, having much of the state associated with 2091 the RTP session rather than an individual SSRC can avoid the need 2092 for in-session signalling of meta-information about each SSRC. 2094 6. Low overhead for security association establishment. 2096 The Cons: 2098 a. May have some need for cross session RTCP requests for things 2099 that affect both media types in an asynchronous way. 2101 b. Some potential for concern with legacy implementations that does 2102 not support the RTP specification fully when it comes to handling 2103 multiple SSRC per endpoint. 2105 c. Will not be able to control security association for sets of 2106 media streams within the same media type with today's key- 2107 management mechanisms, only between SDP media descriptions. 2109 For RTP applications where all media streams of the same media type 2110 share same usage, this structure provides efficiency gains in amount 2111 of network state used and provides more faith sharing with other 2112 media flows of the same type. At the same time, it is still 2113 maintaining almost all functionalities when it comes to negotiation 2114 in the signalling of the properties for the individual media type and 2115 also enabling flow based QoS prioritisation between media types. It 2116 handles multi-party session well, independently of multicast or 2117 centralised transport distribution, as additional sources can 2118 dynamically enter and leave the session. 2120 7.3. Multiple Sessions for one Media type 2122 In this arch-type one goes one step further than in the above 2123 (Section 7.2) by using multiple RTP sessions also for a single media 2124 type. The main reason for going in this direction is that the RTP 2125 application needs separation of the media streams due to their usage. 2126 Some typical reasons for going to this arch-type are scalability over 2127 multicast, simulcast, need for extended QoS prioritisation of media 2128 streams due to their usage in the application, or the need for fine 2129 granular signalling using today's tools. 2131 The Pros: 2133 1. More suitable for Multicast usage where receivers can 2134 individually select which RTP sessions they want to participate 2135 in, assuming each RTP session has its own multicast group. 2137 2. Detailed indication of the application's usage of the media 2138 stream, where multiple different usages exist. 2140 3. Less need for SSRC specific explicit signalling for each media 2141 stream and thus reduced need for explicit and timely signalling. 2143 4. Enables detailed QoS prioritisation for flow based mechanisms. 2145 5. Works well with de-composite endpoints. 2147 6. Handles dynamic usage of media streams well. 2149 7. For transport translator based multi-party sessions, this 2150 structure allows for improved control of which type of media 2151 streams an endpoint receives. 2153 8. The scope for who is included in a security association can be 2154 structured around the different RTP sessions, thus enabling such 2155 functionality with existing key-management. 2157 The Cons: 2159 a. Increases the amount of RTP sessions compared to Multiple SSRCs 2160 of the Same Media Type. 2162 b. Increased amount of session configuration state. 2164 c. May need synchronised cross-session RTCP requests and require 2165 some consideration due to this. 2167 d. For media streams that are part of scalability, simulcast or 2168 transport robustness it will be needed to bind sources, which 2169 must support multiple RTP sessions. 2171 e. Some potential for concern with legacy implementations that does 2172 not support the RTP specification fully when it comes to handling 2173 multiple SSRC per endpoint. 2175 f. Higher overhead for security association establishment. 2177 g. If the applications need finer control than on media type level 2178 over which session participants that are included in different 2179 sets of security associations, most of today's key-management 2180 will have difficulties establishing such a session. 2182 For more complex RTP applications that have several different usages 2183 for media streams of the same media type and / or uses scalability or 2184 simulcast, this solution can enable those functions at the cost of 2185 increased overhead associated with the additional sessions. This 2186 type of structure is suitable for more advanced applications as well 2187 as multicast based applications requiring differentiation to 2188 different participants. 2190 7.4. Multiple Media Types in one Session 2192 This arch-type is to use a single RTP session for multiple different 2193 media types, like audio and video, and possibly also transport 2194 robustness mechanisms like FEC or Retransmission. Each media stream 2195 will use its own SSRC and a given SSRC value from a particular 2196 endpoint will never use the SSRC for more than a single media type. 2198 The Pros: 2200 1. Single RTP session which implies: 2202 * Minimal NAT/FW state. 2204 * Minimal NAT/FW Traversal Cost. 2206 * Fate-sharing for all media flows. 2208 2. Enables separation of the different media types based on the 2209 payload types so media type specific endpoint or central 2210 processing can still be supported despite single session. 2212 3. Can handle dynamic allocations of media streams well on an RTP 2213 level. Depends on the application's needs for explicit 2214 indication of the stream usage and how timely that can be 2215 signalled. 2217 4. Minimal overhead for security association establishment. 2219 The Cons: 2221 a. Less suitable for interworking with other applications that uses 2222 individual RTP sessions per media type or multiple sessions for a 2223 single media type, due to need of SSRC translation. 2225 b. Negotiation of bandwidth for the different media types is 2226 currently not possible in SDP. This requires SDP extensions to 2227 enable payload or source specific bandwidth. Likely to be a 2228 problem due to media type asymmetry in required bandwidth. 2230 c. Not suitable for de-composite end-points as it requires higher 2231 bandwidth and processing. 2233 d. Flow based QoS cannot provide separate treatment to some media 2234 streams compared to other in the single RTP session. 2236 e. If there is significant asymmetry between the media streams RTCP 2237 reporting needs, there are some challenges in configuration and 2238 usage to avoid wasting RTCP reporting on the media stream that 2239 does not need that frequent reporting. 2241 f. Not suitable for applications where some receivers like to 2242 receive only a subset of the media streams, especially if 2243 multicast or transport translator is being used. 2245 g. Additional concern with legacy implementations that does not 2246 support the RTP specification fully when it comes to handling 2247 multiple SSRC per endpoint, as also multiple simultaneous media 2248 types needs to be handled. 2250 h. If the applications need finer control over which session 2251 participants that are included in different sets of security 2252 associations, most key-management will have difficulties 2253 establishing such a session. 2255 The analysis in this document and considerations in ??? implies that 2256 this is suitable only in a set of restricted use cases. The aspect 2257 in the above list that can be most difficult to judge long term is 2258 likely the potential need for interworking with other applications 2259 and services. 2261 7.5. Summary 2263 There are some clear relations between these arch-types. Both the 2264 "single SSRC per RTP session" and the "multiple media types in one 2265 session" are cases which require full explicit signalling of the 2266 media stream relations. However, they operate on two different 2267 levels where the first primarily enables session level binding, and 2268 the second needs to do it all on SSRC level. From another 2269 perspective, the two solutions are the two extreme points when it 2270 comes to number of RTP sessions required. 2272 The two other arch-types "Multiple SSRCs of the Same Media Type" and 2273 "Multiple Sessions for one Media Type" are examples of two other 2274 cases that first of all allows for some implicit mapping of the role 2275 or usage of the media streams based on which RTP session they appear 2276 in. It thus potentially allows for less signalling and in particular 2277 reduced need for real-time signalling in dynamic sessions. They also 2278 represent points in between the first two when it comes to amount of 2279 RTP sessions established, i.e. representing an attempt to reduce the 2280 amount of sessions as much as possible without compromising the 2281 functionality the session provides both on network level and on 2282 signalling level. 2284 8. Summary considerations and guidelines 2286 8.1. Guidelines 2288 This section contains a number of recommendations for implementors or 2289 specification writers when it comes to handling multi-stream. 2291 Do not Require the same SSRC across Sessions: As discussed in 2292 Section 6.2.5 there exist drawbacks in using the same SSRC in 2293 multiple RTP sessions as a mechanism to bind related media streams 2294 together. It is instead recommended that a mechanism to 2295 explicitly signal the relation is used, either in RTP/RTCP or in 2296 the used signalling mechanism that establishes the RTP session(s). 2298 Use SSRC multiplexing for additional Media Sources: In the cases an 2299 RTP endpoint needs to transmit additional media streams of the 2300 same media type in the application, with the same processing 2301 requirements at the network and RTP layers, it is recommended to 2302 send them as additional SSRCs in the same RTP session. For 2303 example a telepresence room where there are three cameras, and 2304 each camera captures 2 persons sitting at the table, sending each 2305 camera as its own SSRC within a single RTP session is recommended. 2307 Use additional RTP sessions for streams with different requirements: 2308 When media streams have different processing requirements from the 2309 network or the RTP layer at the endpoints, it is recommended that 2310 the different types of streams are put in different RTP sessions. 2311 This includes the case where different participants want different 2312 subsets of the set of RTP streams. 2314 When using Session Multiplexing use grouping: When using Session 2315 Multiplexing solutions, it is recommended to be explicitly group 2316 the involved RTP sessions using the signalling mechanism, for 2317 example The Session Description Protocol (SDP) Grouping Framework. 2318 [RFC5888], using some appropriate grouping semantics. 2320 RTP/RTCP Extensions May Support SSRC and Session Multiplexing: When 2321 defining an RTP or RTCP extension, the creator needs to consider 2322 if this extension is applicable in both SSRC multiplexed and 2323 Session multiplexed usages. Any extension intended to be generic 2324 is recommended to support both. Applications that are not as 2325 generally applicable will have to consider if interoperability is 2326 better served by defining a single solution or providing both 2327 options. 2329 Transport Support Extensions: When defining new RTP/RTCP extensions 2330 intended for transport support, like the retransmission or FEC 2331 mechanisms, they are recommended to include support for both SSRC 2332 and Session multiplexing so that application developers can choose 2333 freely from the set of mechanisms without concerning themselves 2334 with which of the multiplexing choices a particular solution 2335 supports. 2337 9. IANA Considerations 2339 This document makes no request of IANA. 2341 Note to RFC Editor: this section may be removed on publication as an 2342 RFC. 2344 10. Security Considerations 2346 There is discussion of the security implications of choosing SSRC vs 2347 Session multiplexing in Section 6.5. 2349 11. References 2351 11.1. Normative References 2353 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 2354 Jacobson, "RTP: A Transport Protocol for Real-Time 2355 Applications", STD 64, RFC 3550, July 2003. 2357 11.2. Informative References 2359 [ALF] Clark, D. and D. Tennenhouse, "Architectural 2360 Considerations for a New Generation of Protocols", SIGCOMM 2361 Symposium on Communications Architectures and 2362 Protocols (Philadelphia, Pennsylvania), pp. 200--208, IEEE 2363 Computer Communications Review, Vol. 20(4), 2364 September 1990. 2366 [I-D.alvestrand-rtp-sess-neutral] 2367 Alvestrand, H., "Why RTP Sessions Should Be Content 2368 Neutral", draft-alvestrand-rtp-sess-neutral-01 (work in 2369 progress), June 2012. 2371 [I-D.ietf-avt-srtp-ekt] 2372 Wing, D., McGrew, D., and K. Fischer, "Encrypted Key 2373 Transport for Secure RTP", draft-ietf-avt-srtp-ekt-03 2374 (work in progress), October 2011. 2376 [I-D.ietf-avtcore-rtp-security-options] 2377 Westerlund, M. and C. Perkins, "Options for Securing RTP 2378 Sessions", draft-ietf-avtcore-rtp-security-options-00 2379 (work in progress), July 2012. 2381 [I-D.ietf-avtext-multiple-clock-rates] 2382 Petit-Huguenin, M. and G. Zorn, "Support for Multiple 2383 Clock Rates in an RTP Session", 2384 draft-ietf-avtext-multiple-clock-rates-05 (work in 2385 progress), May 2012. 2387 [I-D.ietf-mmusic-sdp-bundle-negotiation] 2388 Holmberg, C. and H. Alvestrand, "Multiplexing Negotiation 2389 Using Session Description Protocol (SDP) Port Numbers", 2390 draft-ietf-mmusic-sdp-bundle-negotiation-00 (work in 2391 progress), February 2012. 2393 [I-D.ietf-payload-rtp-howto] 2394 Westerlund, M., "How to Write an RTP Payload Format", 2395 draft-ietf-payload-rtp-howto-02 (work in progress), 2396 July 2012. 2398 [I-D.lennox-avtcore-rtp-multi-stream] 2399 Lennox, J. and M. Westerlund, "Real-Time Transport 2400 Protocol (RTP) Considerations for Endpoints Sending 2401 Multiple Media Streams", 2402 draft-lennox-avtcore-rtp-multi-stream-00 (work in 2403 progress), July 2012. 2405 [I-D.lennox-mmusic-sdp-source-selection] 2406 Lennox, J. and H. Schulzrinne, "Mechanisms for Media 2407 Source Selection in the Session Description Protocol 2408 (SDP)", draft-lennox-mmusic-sdp-source-selection-04 (work 2409 in progress), March 2012. 2411 [I-D.westerlund-avtcore-max-ssrc] 2412 Westerlund, M., Burman, B., and F. Jansson, "Multiple 2413 Synchronization sources (SSRC) in RTP Session Signaling", 2414 draft-westerlund-avtcore-max-ssrc-01 (work in progress), 2415 April 2012. 2417 [I-D.westerlund-avtcore-multi-media-rtp-session] 2418 Westerlund, M., Perkins, C., and J. Lennox, "Multiple 2419 Media Types in an RTP Session", 2420 draft-westerlund-avtcore-multi-media-rtp-session-00 (work 2421 in progress), July 2012. 2423 [I-D.westerlund-avtcore-transport-multiplexing] 2424 Westerlund, M. and C. Perkins, "Multiple RTP Sessions on a 2425 Single Lower-Layer Transport", 2426 draft-westerlund-avtcore-transport-multiplexing-02 (work 2427 in progress), March 2012. 2429 [I-D.westerlund-avtext-rtcp-sdes-srcname] 2430 Westerlund, M., Burman, B., and P. Sandgren, "RTCP SDES 2431 Item SRCNAME to Label Individual Sources", 2432 draft-westerlund-avtext-rtcp-sdes-srcname-00 (work in 2433 progress), October 2011. 2435 [RFC2198] Perkins, C., Kouvelas, I., Hodson, O., Hardman, V., 2436 Handley, M., Bolot, J., Vega-Garcia, A., and S. Fosse- 2437 Parisis, "RTP Payload for Redundant Audio Data", RFC 2198, 2438 September 1997. 2440 [RFC2205] Braden, B., Zhang, L., Berson, S., Herzog, S., and S. 2441 Jamin, "Resource ReSerVation Protocol (RSVP) -- Version 1 2442 Functional Specification", RFC 2205, September 1997. 2444 [RFC2326] Schulzrinne, H., Rao, A., and R. Lanphier, "Real Time 2445 Streaming Protocol (RTSP)", RFC 2326, April 1998. 2447 [RFC2474] Nichols, K., Blake, S., Baker, F., and D. Black, 2448 "Definition of the Differentiated Services Field (DS 2449 Field) in the IPv4 and IPv6 Headers", RFC 2474, 2450 December 1998. 2452 [RFC2974] Handley, M., Perkins, C., and E. Whelan, "Session 2453 Announcement Protocol", RFC 2974, October 2000. 2455 [RFC3261] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, 2456 A., Peterson, J., Sparks, R., Handley, M., and E. 2457 Schooler, "SIP: Session Initiation Protocol", RFC 3261, 2458 June 2002. 2460 [RFC3264] Rosenberg, J. and H. Schulzrinne, "An Offer/Answer Model 2461 with Session Description Protocol (SDP)", RFC 3264, 2462 June 2002. 2464 [RFC3389] Zopf, R., "Real-time Transport Protocol (RTP) Payload for 2465 Comfort Noise (CN)", RFC 3389, September 2002. 2467 [RFC3551] Schulzrinne, H. and S. Casner, "RTP Profile for Audio and 2468 Video Conferences with Minimal Control", STD 65, RFC 3551, 2469 July 2003. 2471 [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and K. 2472 Norrman, "The Secure Real-time Transport Protocol (SRTP)", 2473 RFC 3711, March 2004. 2475 [RFC3830] Arkko, J., Carrara, E., Lindholm, F., Naslund, M., and K. 2476 Norrman, "MIKEY: Multimedia Internet KEYing", RFC 3830, 2477 August 2004. 2479 [RFC4103] Hellstrom, G. and P. Jones, "RTP Payload for Text 2480 Conversation", RFC 4103, June 2005. 2482 [RFC4566] Handley, M., Jacobson, V., and C. Perkins, "SDP: Session 2483 Description Protocol", RFC 4566, July 2006. 2485 [RFC4568] Andreasen, F., Baugher, M., and D. Wing, "Session 2486 Description Protocol (SDP) Security Descriptions for Media 2487 Streams", RFC 4568, July 2006. 2489 [RFC4588] Rey, J., Leon, D., Miyazaki, A., Varsa, V., and R. 2490 Hakenberg, "RTP Retransmission Payload Format", RFC 4588, 2491 July 2006. 2493 [RFC4607] Holbrook, H. and B. Cain, "Source-Specific Multicast for 2494 IP", RFC 4607, August 2006. 2496 [RFC5104] Wenger, S., Chandra, U., Westerlund, M., and B. Burman, 2497 "Codec Control Messages in the RTP Audio-Visual Profile 2498 with Feedback (AVPF)", RFC 5104, February 2008. 2500 [RFC5117] Westerlund, M. and S. Wenger, "RTP Topologies", RFC 5117, 2501 January 2008. 2503 [RFC5576] Lennox, J., Ott, J., and T. Schierl, "Source-Specific 2504 Media Attributes in the Session Description Protocol 2505 (SDP)", RFC 5576, June 2009. 2507 [RFC5583] Schierl, T. and S. Wenger, "Signaling Media Decoding 2508 Dependency in the Session Description Protocol (SDP)", 2509 RFC 5583, July 2009. 2511 [RFC5760] Ott, J., Chesterfield, J., and E. Schooler, "RTP Control 2512 Protocol (RTCP) Extensions for Single-Source Multicast 2513 Sessions with Unicast Feedback", RFC 5760, February 2010. 2515 [RFC5761] Perkins, C. and M. Westerlund, "Multiplexing RTP Data and 2516 Control Packets on a Single Port", RFC 5761, April 2010. 2518 [RFC5764] McGrew, D. and E. Rescorla, "Datagram Transport Layer 2519 Security (DTLS) Extension to Establish Keys for the Secure 2520 Real-time Transport Protocol (SRTP)", RFC 5764, May 2010. 2522 [RFC5888] Camarillo, G. and H. Schulzrinne, "The Session Description 2523 Protocol (SDP) Grouping Framework", RFC 5888, June 2010. 2525 [RFC6190] Wenger, S., Wang, Y., Schierl, T., and A. Eleftheriadis, 2526 "RTP Payload Format for Scalable Video Coding", RFC 6190, 2527 May 2011. 2529 [RFC6222] Begen, A., Perkins, C., and D. Wing, "Guidelines for 2530 Choosing RTP Control Protocol (RTCP) Canonical Names 2531 (CNAMEs)", RFC 6222, April 2011. 2533 [RFC6285] Ver Steeg, B., Begen, A., Van Caenegem, T., and Z. Vax, 2534 "Unicast-Based Rapid Acquisition of Multicast RTP 2535 Sessions", RFC 6285, June 2011. 2537 [RFC6465] Ivov, E., Marocco, E., and J. Lennox, "A Real-time 2538 Transport Protocol (RTP) Header Extension for Mixer-to- 2539 Client Audio Level Indication", RFC 6465, December 2011. 2541 Appendix A. Dismissing Payload Type Multiplexing 2543 This section documents a number of reasons why using the payload type 2544 as a multiplexing point for most things related to multiple streams 2545 is unsuitable. If one attempts to use Payload type multiplexing 2546 beyond it's defined usage, that has well known negative effects on 2547 RTP. To use Payload type as the single discriminator for multiple 2548 streams implies that all the different media streams are being sent 2549 with the same SSRC, thus using the same timestamp and sequence number 2550 space. This has many effects: 2552 1. Putting restraint on RTP timestamp rate for the multiplexed 2553 media. For example, media streams that use different RTP 2554 timestamp rates cannot be combined, as the timestamp values need 2555 to be consistent across all multiplexed media frames. Thus 2556 streams are forced to use the same rate. When this is not 2557 possible, Payload Type multiplexing cannot be used. 2559 2. Many RTP payload formats may fragment a media object over 2560 multiple packets, like parts of a video frame. These payload 2561 formats need to determine the order of the fragments to 2562 correctly decode them. Thus it is important to ensure that all 2563 fragments related to a frame or a similar media object are 2564 transmitted in sequence and without interruptions within the 2565 object. This can relatively simple be solved on the sender side 2566 by ensuring that the fragments of each media stream are sent in 2567 sequence. 2569 3. Some media formats require uninterrupted sequence number space 2570 between media parts. These are media formats where any missing 2571 RTP sequence number will result in decoding failure or invoking 2572 of a repair mechanism within a single media context. The text/ 2573 T140 payload format [RFC4103] is an example of such a format. 2574 These formats will need a sequence numbering abstraction 2575 function between RTP and the individual media stream before 2576 being used with Payload Type multiplexing. 2578 4. Sending multiple streams in the same sequence number space makes 2579 it impossible to determine which Payload Type and thus which 2580 stream a packet loss relates to. 2582 5. If RTP Retransmission [RFC4588] is used and there is a loss, it 2583 is possible to ask for the missing packet(s) by SSRC and 2584 sequence number, not by Payload Type. If only some of the 2585 Payload Type multiplexed streams are of interest, there is no 2586 way of telling which missing packet(s) belong to the interesting 2587 stream(s) and all lost packets must be requested, wasting 2588 bandwidth. 2590 6. The current RTCP feedback mechanisms are built around providing 2591 feedback on media streams based on stream ID (SSRC), packet 2592 (sequence numbers) and time interval (RTP Timestamps). There is 2593 almost never a field to indicate which Payload Type is reported, 2594 so sending feedback for a specific media stream is difficult 2595 without extending existing RTCP reporting. 2597 7. The current RTCP media control messages [RFC5104] specification 2598 is oriented around controlling particular media flows, i.e. 2599 requests are done addressing a particular SSRC. Such mechanisms 2600 would need to be redefined to support Payload Type multiplexing. 2602 8. The number of payload types are inherently limited. 2603 Accordingly, using Payload Type multiplexing limits the number 2604 of streams that can be multiplexed and does not scale. This 2605 limitation is exacerbated if one uses solutions like RTP and 2606 RTCP multiplexing [RFC5761] where a number of payload types are 2607 blocked due to the overlap between RTP and RTCP. 2609 9. At times, there is a need to group multiplexed streams and this 2610 is currently possible for RTP Sessions and for SSRC, but there 2611 is no defined way to group Payload Types. 2613 10. It is currently not possible to signal bandwidth requirements 2614 per media stream when using Payload Type Multiplexing. 2616 11. Most existing SDP media level attributes cannot be applied on a 2617 per Payload Type level and would require re-definition in that 2618 context. 2620 12. A legacy endpoint that doesn't understand the indication that 2621 different RTP payload types are different media streams may be 2622 slightly confused by the large amount of possibly overlapping or 2623 identically defined RTP Payload Types. 2625 Appendix B. Proposals for Future Work 2627 The above discussion and guidelines indicates that a small set of 2628 extension mechanisms could greatly improve the situation when it 2629 comes to using multiple streams independently of Session multiplexing 2630 or SSRC multiplexing. These extensions are: 2632 Media Source Identification: A Media source identification that can 2633 be used to bind together media streams that are related to the 2634 same media source. A proposal 2635 [I-D.westerlund-avtext-rtcp-sdes-srcname] exist for a new SDES 2636 item SRCNAME that also can be used with the a=ssrc SDP attribute 2637 to provide signalling layer binding information. 2639 SSRC limitations within RTP sessions: By providing a signalling 2640 solution that allows the signalling peers to explicitly express 2641 both support and limitations on how many simultaneous media 2642 streams an endpoint can handle within a given RTP Session. That 2643 ensures that usage of SSRC multiplexing occurs when supported and 2644 without overloading an endpoint. This extension is proposed in 2645 [I-D.westerlund-avtcore-max-ssrc]. 2647 Appendix C. RTP Specification Clarifications 2649 This section describes a number of clarifications to the RTP 2650 specifications that are likely necessary for aligned behaviour when 2651 RTP sessions contain more SSRCs than one local and one remote. 2653 All of the below proposals are under consideration in 2654 [I-D.lennox-avtcore-rtp-multi-stream]. 2656 C.1. RTCP Reporting from all SSRCs 2658 When one has multiple SSRC in an RTP node, all these SSRC must send 2659 some RTP or RTCP packet as long as the SSRC exist. It is not 2660 sufficient that only one SSRC in the node sends report blocks on the 2661 incoming RTP streams; any SSRC that intends to remain in the session 2662 must send some packets to avoid timing out according to the rules in 2663 RFC 3550 section 6.3.5. 2665 It has been hypothesised that a third party monitor may be confused 2666 by not necessarily being able to determine that all these SSRC are in 2667 fact co-located and originate from the same stack instance; if this 2668 hypothesis is true, this may argue for having all the sources send 2669 full reception reports, even though they are reporting the same 2670 packet delivery. 2672 The contrary argument is that such double reporting may confuse the 2673 third party monitor even more by making it seem that utilisation of 2674 the last-hop link to the recipient is (number of SSRCs) times higher 2675 than what it actually is. 2677 C.2. RTCP Self-reporting 2679 For any RTP node that sends more than one SSRC, there is the question 2680 if SSRC1 needs to report its reception of SSRC2 and vice versa. The 2681 reason that they in fact need to report on all other local streams as 2682 being received is report consistency. The hypothetical third party 2683 monitor that considers the full matrix of media streams and all known 2684 SSRC reports on these media streams would detect a gap in the reports 2685 which could be a transport issue unless identified as in fact being 2686 sources from the same node. 2688 C.3. Combined RTCP Packets 2690 When a node contains multiple SSRCs, it is questionable if an RTCP 2691 compound packet can only contain RTCP packets from a single SSRC or 2692 if multiple SSRCs can include their packets in a joint compound 2693 packet. The high level question is a matter for any receiver 2694 processing on what to expect. In addition to that question there is 2695 the issue of how to use the RTCP timer rules in these cases, as the 2696 existing rules are focused on determining when a single SSRC can 2697 send. 2699 Appendix D. Signalling considerations 2701 Signalling is not an architectural consideration for RTP itself, so 2702 this discussion has been moved to an appendix. However, it is hugely 2703 important for anyone building complete applications, so it is 2704 deserving of discussion. 2706 The issues raised here need to be addressed in the WGs that deal with 2707 signalling; they cannot be addressed by tweaking, extending or 2708 profiling RTP. 2710 D.1. Signalling Aspects 2712 There exist various signalling solutions for establishing RTP 2713 sessions. Many are SDP [RFC4566] based, however SDP functionality is 2714 also dependent on the signalling protocols carrying the SDP. Where 2715 RTSP [RFC2326] and SAP [RFC2974] both use SDP in a declarative 2716 fashion, while SIP [RFC3261] uses SDP with the additional definition 2717 of Offer/Answer [RFC3264]. The impact on signalling and especially 2718 SDP needs to be considered as it can greatly affect how to deploy a 2719 certain multiplexing point choice. 2721 D.1.1. Session Oriented Properties 2723 One aspect of the existing signalling is that it is focused around 2724 sessions, or at least in the case of SDP the media description. 2725 There are a number of things that are signalled on a session level/ 2726 media description but those are not necessarily strictly bound to an 2727 RTP session and could be of interest to signal specifically for a 2728 particular media stream (SSRC) within the session. The following 2729 properties have been identified as being potentially useful to signal 2730 not only on RTP session level: 2732 o Bitrate/Bandwidth exist today only at aggregate or a common any 2733 media stream limit, unless either codec-specific bandwidth 2734 limiting or RTCP signalling using TMMBR is used. 2736 o Which SSRC that will use which RTP Payload Types (this will be 2737 visible from the first media packet, but is sometimes useful to 2738 know before packet arrival). 2740 Some of these issues are clearly SDP's problem rather than RTP 2741 limitations. However, if the aim is to deploy an SSRC multiplexed 2742 solution that contains several sets of media streams with different 2743 properties (encoding/packetization parameter, bit-rate, etc), putting 2744 each set in a different RTP session would directly enable negotiation 2745 of the parameters for each set. If insisting on SSRC multiplexing 2746 only, a number of signalling extensions are needed to clarify that 2747 there are multiple sets of media streams with different properties 2748 and that they shall in fact be kept different, since a single set 2749 will not satisfy the application's requirements. 2751 For some parameters, such as resolution and framerate, a SSRC-linked 2752 mechanism has been proposed: 2753 [I-D.lennox-mmusic-sdp-source-selection]. 2755 D.1.2. SDP Prevents Multiple Media Types 2757 SDP chose to use the m= line both to delineate an RTP session and to 2758 specify the top level of the MIME media type; audio, video, text, 2759 image, application. This media type is used as the top-level media 2760 type for identifying the actual payload format bound to a particular 2761 payload type using the rtpmap attribute. This binding has to be 2762 loosened in order to use SDP to describe RTP sessions containing 2763 multiple MIME top level types. 2765 There is an accepted WG item in the MMUSIC WG to define how multiple 2766 media lines describe a single underlying transport 2767 [I-D.ietf-mmusic-sdp-bundle-negotiation] and thus it becomes possible 2768 in SDP to define one RTP session with media types having different 2769 MIME top level types. 2771 D.1.3. Signalling Media Stream Usage 2773 Media streams being transported in RTP has some particular usage in 2774 an RTP application. This usage of the media stream is in many 2775 applications so far implicitly signalled. For example, an 2776 application may choose to take all incoming audio RTP streams, mix 2777 them and play them out. However, in more advanced applications that 2778 use multiple media streams there will be more than a single usage or 2779 purpose among the set of media streams being sent or received. RTP 2780 applications will need to signal this usage somehow. The signalling 2781 used will have to identify the media streams affected by their RTP- 2782 level identifiers, which means that they have to be identified either 2783 by their session or by their SSRC + session. 2785 In some applications, the receiver cannot utilise the media stream at 2786 all before it has received the signalling message describing the 2787 media stream and its usage. In other applications, there exists a 2788 default handling that is appropriate. 2790 If all media streams in an RTP session are to be treated in the same 2791 way, identifying the session is enough. If SSRCs in a session are to 2792 be treated differently, signalling must identify both the session and 2793 the SSRC. 2795 If this signalling affects how any RTP central node, like an RTP 2796 mixer or translator that selects, mixes or processes streams, treats 2797 the streams, the node will also need to receive the same signalling 2798 to know how to treat media streams with different usage in the right 2799 fashion. 2801 Appendix E. Changes from -01 to -02 2803 o Added Harald Alvestrand as co-author. 2805 o Removed unused term "Media aggregate". 2807 o Added term "RTP session group", noted that CNAMEs are assumed to 2808 bind across the sessions of an RTP session group, and used it when 2809 appropriate (TODO) 2811 o Moved discussion of signalling aspects to appendix 2813 o Removed all suggestion that PT can be a multiplexing point 2815 o Normalised spelling of "endpoint" to follow RFC 3550 and not use a 2816 hyphen. 2818 o Added CNAME to definition list. 2820 o Added term "Media Sink" for the thing that is identified by a 2821 listen-only SSRC. 2823 o Added term "RTP source" for the thing that transmits one media 2824 stream, separating it from "Media Source". [[OUTSTANDING: Whether 2825 to use "RTP Source" or "Media Sender" here]] 2827 o Rewrote section on distributed endpoint, noting that this, like 2828 any endpoint that wants a subset of a set of RTP streams, needs 2829 multiple RTP sessions. 2831 o Removed all substantive references to the undefined term "purpose" 2832 from the main body of the document when it referred to the purpose 2833 of an RTP stream. 2835 o Moved the summary section of section 6 to the guidelines section 2836 that it most closely supports. 2838 o 2840 Authors' Addresses 2842 Magnus Westerlund 2843 Ericsson 2844 Farogatan 6 2845 SE-164 80 Kista 2846 Sweden 2848 Phone: +46 10 714 82 87 2849 Email: magnus.westerlund@ericsson.com 2851 Bo Burman 2852 Ericsson 2853 Farogatan 6 2854 SE-164 80 Kista 2855 Sweden 2857 Phone: +46 10 714 13 11 2858 Email: bo.burman@ericsson.com 2860 Colin Perkins 2861 University of Glasgow 2862 School of Computing Science 2863 Glasgow G12 8QQ 2864 United Kingdom 2866 Email: csp@csperkins.org 2868 Harald Tveit Alvestrand 2869 Google 2870 Kungsbron 2 2871 Stockholm, 11122 2872 Sweden 2874 Phone: 2875 Fax: 2876 Email: harald@alvestrand.no 2877 URI: