idnits 2.17.1 draft-ietf-avtcore-multiplex-guidelines-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 : ---------------------------------------------------------------------------- ** There is 1 instance of too long lines in the document, the longest one being 13 characters in excess of 72. ** 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 427: '... Transmission of scalable video [RFC6190]. This is NOT RECOMMENDED...' RFC 2119 keyword, line 538: '... MUST be unique across signalling co...' RFC 2119 keyword, line 740: '... SHOULD be carried in a separate RTP...' RFC 2119 keyword, line 743: '...nd video streams SHOULD NOT be carried...' RFC 2119 keyword, line 828: '...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 (January 13, 2014) is 3753 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Unused Reference: 'I-D.ietf-avtext-multiple-clock-rates' is defined on line 1757, but no explicit reference was found in the text == Unused Reference: 'RFC5583' is defined on line 1885, but no explicit reference was found in the text == Unused Reference: 'RFC6285' is defined on line 1907, but no explicit reference was found in the text == Outdated reference: A later version (-13) exists of draft-ietf-avtcore-multi-media-rtp-session-03 == Outdated reference: A later version (-10) exists of draft-ietf-avtcore-rtp-security-options-09 == Outdated reference: A later version (-11) exists of draft-ietf-avtext-multiple-clock-rates-10 == Outdated reference: A later version (-17) exists of draft-ietf-mmusic-msid-02 == Outdated reference: A later version (-54) exists of draft-ietf-mmusic-sdp-bundle-negotiation-05 == Outdated reference: A later version (-14) exists of draft-ietf-payload-rtp-howto-09 -- Obsolete informational reference (is this intentional?): RFC 2326 (Obsoleted by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: 2 errors (**), 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: July 17, 2014 C. Perkins 6 University of Glasgow 7 H. Alvestrand 8 Google 9 January 13, 2014 11 Guidelines for using the Multiplexing Features of RTP to Support 12 Multiple Media Streams 13 draft-ietf-avtcore-multiplex-guidelines-02 15 Abstract 17 The Real-time Transport Protocol (RTP) is a flexible protocol that 18 can be used in a wide range of applications, networks, and system 19 topologies. That flexibility makes for wide applicability, but can 20 complicate the application design process. One particular design 21 question that has received much attention is how to support multiple 22 media streams in RTP. This memo discusses the available options and 23 design trade-offs, and provides guidelines on how to use the 24 multiplexing features of RTP to support multiple media streams. 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 July 17, 2014. 43 Copyright Notice 45 Copyright (c) 2014 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 . . . . . . . . . . . . . . . . . . . . . . . . 3 61 2. Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 4 62 2.1. Terminology . . . . . . . . . . . . . . . . . . . . . . . 4 63 2.2. Subjects Out of Scope . . . . . . . . . . . . . . . . . . 6 64 3. Reasons for Multiplexing and Grouping RTP Media Streams . . . 6 65 4. RTP Multiplexing Points . . . . . . . . . . . . . . . . . . . 7 66 4.1. RTP Session . . . . . . . . . . . . . . . . . . . . . . . 8 67 4.2. Synchronisation Source (SSRC) . . . . . . . . . . . . . . 9 68 4.3. Contributing Source (CSRC) . . . . . . . . . . . . . . . 10 69 4.4. RTP Payload Type . . . . . . . . . . . . . . . . . . . . 11 70 5. RTP Topologies and Issues . . . . . . . . . . . . . . . . . . 12 71 5.1. Point to Point . . . . . . . . . . . . . . . . . . . . . 12 72 5.2. Translators & Gateways . . . . . . . . . . . . . . . . . 13 73 5.3. Point to Multipoint Using Multicast . . . . . . . . . . . 13 74 5.4. Point to Multipoint Using an RTP Transport Translator . . 14 75 5.5. Point to Multipoint Using an RTP Mixer . . . . . . . . . 15 76 6. RTP Multiplexing: When to Use Multiple RTP Sessions . . . . . 15 77 6.1. RTP and RTCP Protocol Considerations . . . . . . . . . . 16 78 6.1.1. The RTP Specification . . . . . . . . . . . . . . . . 16 79 6.1.2. Multiple SSRCs in a Session . . . . . . . . . . . . . 18 80 6.1.3. Handling Varying Sets of Senders . . . . . . . . . . 19 81 6.1.4. Cross Session RTCP Requests . . . . . . . . . . . . . 19 82 6.1.5. Binding Related Sources . . . . . . . . . . . . . . . 19 83 6.1.6. Forward Error Correction . . . . . . . . . . . . . . 21 84 6.1.7. Transport Translator Sessions . . . . . . . . . . . . 21 85 6.2. Interworking Considerations . . . . . . . . . . . . . . . 21 86 6.2.1. Types of Interworking . . . . . . . . . . . . . . . . 22 87 6.2.2. RTP Translator Interworking . . . . . . . . . . . . . 22 88 6.2.3. Gateway Interworking . . . . . . . . . . . . . . . . 22 89 6.2.4. Multiple SSRC Legacy Considerations . . . . . . . . . 23 90 6.3. Network Considerations . . . . . . . . . . . . . . . . . 24 91 6.3.1. Quality of Service . . . . . . . . . . . . . . . . . 24 92 6.3.2. NAT and Firewall Traversal . . . . . . . . . . . . . 25 93 6.3.3. Multicast . . . . . . . . . . . . . . . . . . . . . . 26 94 6.3.4. Multiplexing multiple RTP Session on a Single 95 Transport . . . . . . . . . . . . . . . . . . . . . . 27 97 6.4. Security and Key Management Considerations . . . . . . . 27 98 6.4.1. Security Context Scope . . . . . . . . . . . . . . . 27 99 6.4.2. Key Management for Multi-party session . . . . . . . 28 100 6.4.3. Complexity Implications . . . . . . . . . . . . . . . 28 101 7. Archetypes . . . . . . . . . . . . . . . . . . . . . . . . . 29 102 7.1. Single SSRC per Session . . . . . . . . . . . . . . . . . 29 103 7.2. Multiple SSRCs of the Same Media Type . . . . . . . . . . 31 104 7.3. Multiple Sessions for one Media type . . . . . . . . . . 32 105 7.4. Multiple Media Types in one Session . . . . . . . . . . . 34 106 7.5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . 35 107 8. Summary considerations and guidelines . . . . . . . . . . . . 35 108 8.1. Guidelines . . . . . . . . . . . . . . . . . . . . . . . 35 109 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 36 110 10. Security Considerations . . . . . . . . . . . . . . . . . . . 37 111 11. References . . . . . . . . . . . . . . . . . . . . . . . . . 37 112 11.1. Normative References . . . . . . . . . . . . . . . . . . 37 113 11.2. Informative References . . . . . . . . . . . . . . . . . 37 114 Appendix A. Dismissing Payload Type Multiplexing . . . . . . . . 41 115 Appendix B. Proposals for Future Work . . . . . . . . . . . . . 43 116 Appendix C. Signalling considerations . . . . . . . . . . . . . 43 117 C.1. Signalling Aspects . . . . . . . . . . . . . . . . . . . 44 118 C.1.1. Session Oriented Properties . . . . . . . . . . . . . 44 119 C.1.2. SDP Prevents Multiple Media Types . . . . . . . . . . 44 120 C.1.3. Signalling Media Stream Usage . . . . . . . . . . . . 45 121 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 45 123 1. Introduction 125 The Real-time Transport Protocol (RTP) [RFC3550] is a commonly used 126 protocol for real-time media transport. It is a protocol that 127 provides great flexibility and can support a large set of different 128 applications. RTP has several multiplexing points designed for 129 different purposes. These enable support of multiple media streams 130 and switching between different encoding or packetization of the 131 media. By using multiple RTP sessions, sets of media streams can be 132 structured for efficient processing or identification. Thus the 133 question for any RTP application designer is how to best use the RTP 134 session, the SSRC and the payload type to meet the application's 135 needs. 137 The purpose of this document is to provide clear information about 138 the possibilities of RTP when it comes to multiplexing. The RTP 139 application designer needs to understand the implications that come 140 from a particular usage of the RTP multiplexing points. The document 141 will recommend against some usages as being unsuitable, in general or 142 for particular purposes. 144 RTP was from the beginning designed for multiple participants in a 145 communication session. This is not restricted to multicast, as some 146 believe, but also provides functionality over unicast, using either 147 multiple transport flows below RTP or a network node that re- 148 distributes the RTP packets. The re-distributing node can for 149 example be a transport translator (relay) that forwards the packets 150 unchanged, a translator performing media or protocol translation in 151 addition to forwarding, or an RTP mixer that creates new sources from 152 the received streams. In addition, multiple streams can occur when a 153 single endpoint have multiple media sources, like multiple cameras or 154 microphones that need to be sent simultaneously. 156 This document has been written due to increased interest in more 157 advanced usage of RTP, resulting in questions regarding the most 158 appropriate RTP usage. The limitations in some implementations, RTP/ 159 RTCP extensions, and signalling has also been exposed. It is 160 expected that some limitations will be addressed by updates or new 161 extensions resolving the shortcomings. The authors also hope that 162 clarification on the usefulness of some functionalities in RTP will 163 result in more complete implementations in the future. 165 The document starts with some definitions and then goes into the 166 existing RTP functionalities around multiplexing. Both the desired 167 behaviour and the implications of a particular behaviour depend on 168 which topologies are used, which requires some consideration. This 169 is followed by a discussion of some choices in multiplexing behaviour 170 and their impacts. Some archetypes of RTP usage are discussed. 171 Finally, some recommendations and examples are provided. 173 2. Definitions 175 2.1. Terminology 177 The following terms and abbreviations are used in this document: 179 Endpoint: A single entity sending or receiving RTP packets. It can 180 be decomposed into several functional blocks, but as long as it 181 behaves a single RTP stack entity it is classified as a single 182 endpoint. 184 Multiparty: A communication situation including multiple endpoints. 185 In this document it will be used to refer to situations where more 186 than two endpoints communicate. 188 Media Source: The source of a stream of data of one Media Type, It 189 can either be a single media capturing device such as a video 190 camera, a microphone, or a specific output of a media production 191 function, such as an audio mixer, or some video editing function. 193 Sending data from a Media Source can cause multiple RTP sources to 194 send multiple Media Streams. 196 Media Stream: A sequence of RTP packets using a single SSRC that 197 together carries part or all of the content of a specific Media 198 Type from a specific sender source within a given RTP session. 200 RTP Source: The originator or source of a particular Media Stream. 201 Identified using an SSRC in a particular RTP session. An RTP 202 source is the source of a single media stream, and is associated 203 with a single endpoint and a single Media Source. An RTP Source 204 is just called a Source in RFC 3550. 206 RTP Sink: A recipient of a Media Stream. The Media Sink is 207 identified using one or more SSRCs. There can be more than one 208 RTP Sink for one RTP source. 210 CNAME: "Canonical name" - identifier associated with one or more RTP 211 sources from a single endpoint. Defined in the RTP specification 212 [RFC3550]. A CNAME identifies a synchronisation context. A CNAME 213 is associated with a single endpoint, although some RTP nodes will 214 use an endpoint's CNAME on that endpoints behalf. An endpoint can 215 use multiple CNAMEs. A CNAME is intended to be globally unique 216 and stable for the full duration of a communication session. 217 [RFC6222][I-D.ietf-avtcore-6222bis] gives updated guidelines for 218 choosing CNAMEs. 220 Media Type: Audio, video, text or data whose form and meaning are 221 defined by a specific real-time application. 223 Multiplexing: The operation of taking multiple entities as input, 224 aggregating them onto some common resource while keeping the 225 individual entities addressable such that they can later be fully 226 and unambiguously separated (de-multiplexed) again. 228 RTP Session: As defined by [RFC3550], the endpoints belonging to the 229 same RTP Session are those that share a single SSRC space. That 230 is, those endpoints can see an SSRC identifier transmitted by any 231 one of the other endpoints. An endpoint can receive an SSRC 232 either as SSRC or as CSRC in RTP and RTCP packets. Thus, the RTP 233 Session scope is decided by the endpoints' network interconnection 234 topology, in combination with RTP and RTCP forwarding strategies 235 deployed by endpoints and any interconnecting middle nodes. 237 RTP Session Group: One or more RTP sessions that are used together 238 to perform some function. Examples are multiple RTP sessions used 239 to carry different layers of a layered encoding. In an RTP 240 Session Group, CNAMEs are assumed to be valid across all RTP 241 sessions, and designate synchronisation contexts that can cross 242 RTP sessions. 244 Source: Term that ought not be used alone. An RTP Source, as 245 identified by its SSRC, is the source of a single Media Stream; a 246 Media Source can be the source of mutiple Media Streams. 248 SSRC: A 32-bit unsigned integer used as identifier for a RTP Source. 250 CSRC: Contributing Source, A SSRC identifier used in a context, like 251 the RTP headers CSRC list, where it is clear that the Media Source 252 is not the source of the media stream, instead only a contributor 253 to the Media Stream. 255 Signalling: The process of configuring endpoints to participate in 256 one or more RTP sessions. 258 2.2. Subjects Out of Scope 260 This document is focused on issues that affect RTP. Thus, issues 261 that involve signalling protocols, such as whether SIP, Jingle or 262 some other protocol is in use for session configuration, the 263 particular syntaxes used to define RTP session properties, or the 264 constraints imposed by particular choices in the signalling 265 protocols, are mentioned only as examples in order to describe the 266 RTP issues more precisely. 268 This document assumes the applications will use RTCP. While there 269 are such applications that don't send RTCP, they do not conform to 270 the RTP specification, and thus can be regarded as reusing the RTP 271 packet format but not implementing the RTP protocol. 273 3. Reasons for Multiplexing and Grouping RTP Media Streams 275 The reasons why an endpoint might choose to send multiple media 276 streams are widespread. In the below discussion, please keep in mind 277 that the reasons for having multiple media streams vary and include 278 but are not limited to the following: 280 o Multiple Media Sources 282 o Multiple Media Streams might be needed to represent one Media 283 Source (for instance when using layered encodings) 285 o A Retransmission stream might repeat the content of another Media 286 Stream 288 o An FEC stream might provide material that can be used to repair 289 another Media Stream 291 o Alternative Encodings, for instance different codecs for the same 292 audio stream 294 o Alternative formats, for instance multiple resolutions of the same 295 video stream 297 For each of these, it is necessary to decide if each additional media 298 stream gets its own SSRC multiplexed within a RTP Session, or if it 299 is necessary to use additional RTP sessions to group the media 300 streams. The choice between these made due to one reason might not 301 be the choice suitable for another reason. In the above list, the 302 different items have different levels of maturity in the discussion 303 on how to solve them. The clearest understanding is associated with 304 multiple media sources of the same media type. However, all warrant 305 discussion and clarification on how to deal with them. As the 306 discussion below will show, in reality we cannot choose a single one 307 of the two solutions. To utilise RTP well and as efficiently as 308 possible, both are needed. The real issue is finding the right 309 guidance on when to create RTP sessions and when additional SSRCs in 310 an RTP session is the right choice. 312 4. RTP Multiplexing Points 314 This section describes the multiplexing points present in the RTP 315 protocol that can be used to distinguish media streams and groups of 316 media streams. Figure 1 outlines the process of demultiplexing 317 incoming RTP streams: 319 | 320 | packets 321 +-- v 322 | +------------+ 323 | | Socket | 324 | +------------+ 325 | || || 326 RTP | RTP/ || |+-----> SCTP ( ...and any other protocols) 327 Session | RTCP || +------> STUN (multiplexed using same port) 328 +-- || 329 +-- || 330 | (split by SSRC) 331 | || || || 332 | || || || 333 Media | +--+ +--+ +--+ 334 Streams | |PB| |PB| |PB| Jitter buffer, process RTCP, FEC, etc. 335 | +--+ +--+ +--+ 336 +-- | | | 337 (pick rending context based on PT) 338 +-- | / | 339 | +---+ | 340 | / | | 341 Payload | +--+ +--+ +--+ 342 Formats | |CR| |CR| |CR| Codecs and rendering 343 | +--+ +--+ +--+ 344 +-- 346 Figure 1: RTP Demultiplexing Process 348 4.1. RTP Session 350 An RTP Session is the highest semantic layer in the RTP protocol, and 351 represents an association between a group of communicating endpoints. 352 The set of participants that form an RTP session is defined as those 353 that share a single synchronisation source space [RFC3550]. That is, 354 if a group of participants are each aware of the synchronisation 355 source identifiers belonging to the other participants, then those 356 participants are in a single RTP session. A participant can become 357 aware of a synchronisation source identifier by receiving an RTP 358 packet containing it in the SSRC field or CSRC list, by receiving an 359 RTCP packet mentioning it in an SSRC field, or through signalling 360 (e.g., the SDP "a=ssrc:" attribute). Thus, the scope of an RTP 361 session is determined by the participants' network interconnection 362 topology, in combination with RTP and RTCP forwarding strategies 363 deployed by the endpoints and any middleboxes, and by the signalling. 365 RTP does not contain a session identifier. Rather, it relies on the 366 underlying transport layer to separate different sessions, and on the 367 signalling to identify sessions in a manner that is meaningful to the 368 application. The signalling layer might give sessions an explicit 369 identifier, or their identification might be implicit based on the 370 addresses and ports used. Accordingly, a single RTP Session can have 371 multiple associated identifiers, explicit and implicit, belonging to 372 different contexts. For example, when running RTP on top of UDP/IP, 373 an RTP endpoint can identify and delimit an RTP Session from other 374 RTP Sessions using the UDP source and destination IP addresses and 375 UDP port numbers. Another example is when using SDP grouping 376 framework [RFC5888] which uses an identifier per "m="-line; if there 377 is a one-to-one mapping between "m="-lines and RTP sessions, that 378 grouping framework identifier will identify an RTP Session. 380 RTP sessions are globally unique, but their identity can only be 381 determined by the communication context at an endpoint of the 382 session, or by a middlebox that is aware of the session context. The 383 relationship between RTP sessions depending on the underlying 384 application, transport, and signalling protocol. The RTP protocol 385 makes no normative statements about the relationship between 386 different RTP sessions, however the applications that use more than 387 one RTP session will have some higher layer understanding of the 388 relationship between the sessions they create. 390 4.2. Synchronisation Source (SSRC) 392 A synchronisation source (SSRC) identifies an RTP source or an RTP 393 sink. Every endpoint will have at least one synchronisation source 394 identifier, even if it does not send media (endpoints that are only 395 RTP sinks still send RTCP, and use their synchronisation source 396 identifier in the RTCP packets they send). An endpoint can have 397 multiple synchronisation sources identifiers if it contains multiple 398 RTP sources (i.e., if it sends multiple media streams). Endpoints 399 that are both RTP sources and RTP sinks use the same synchronisation 400 sources in both roles. At any given time, a RTP source has one and 401 only one SSRC - although that can change over the lifetime of the RTP 402 source or sink. 404 The synchronisation Source identifier is a 32-bit unsigned integer. 405 It is present in every RTP and RTCP packet header, and in the payload 406 of some RTCP packet types. It can also be present in SDP signalling. 407 Unless pre-signalled using the SDP "a=ssrc:" attribute [RFC5576], the 408 synchronisation source identifier is chosen at random. It is not 409 dependent on the network address of the endpoint, and is intended to 410 be unique within an RTP session. Synchronisation source identifier 411 collisions can occur, and are handled as specified in [RFC3550] and 412 [RFC5576], resulting in the synchronisation source identifier of the 413 affecting RTP sources and/or sinks changing. An RTP source that 414 changes its RTP Session identifier (e.g. source transport address) 415 during a session has to choose a new SSRC identifier to avoid being 416 interpreted as looped source. 418 Synchronisation source identifiers that belong to the same 419 synchronisation context (i.e., that represent media streams that can 420 be synchronised using information in RTCP SR packets) are indicated 421 by use of identical CNAME chunks in corresponding RTCP SDES packets. 422 SDP signalling can also be used to provide explicit grouping of 423 synchronisation sources [RFC5576]. 425 In some cases, the same SSRC Identifier value is used to relate 426 streams in two different RTP Sessions, such as in Multi-Session 427 Transmission of scalable video [RFC6190]. This is NOT RECOMMENDED 428 since there is no guarantee of uniqueness in SSRC values across 429 RTP sessions. 431 Note that RTP sequence number and RTP timestamp are scoped by the 432 synchronisation source. Each RTP source will have a different 433 synchronisation source, and the corresponding media stream will have 434 a separate RTP sequence number and timestamp space. 436 An SSRC identifier is used by different type of sources as well as 437 sinks: 439 Real Media Source: Connected to a "physical" media source, for 440 example a camera or microphone. 442 Processed Media Source: A source with some attributed property 443 generated by some network node, for example a filtering function 444 in an RTP mixer that provides the most active speaker based on 445 some criteria, or a mix representing a set of other sources. 447 RTP Sink: A source that does not generate any RTP media stream in 448 itself (e.g. an endpoint or middlebox only receiving in an RTP 449 session). It still needs a sender SSRC for use as source in RTCP 450 reports. 452 Note that a endpoint that generates more than one media type, e.g. a 453 conference participant sending both audio and video, need not (and 454 commonly does not) use the same SSRC value across RTP sessions. RTCP 455 Compound packets containing the CNAME SDES item is the designated 456 method to bind an SSRC to a CNAME, effectively cross-correlating 457 SSRCs within and between RTP Sessions as coming from the same 458 endpoint. The main property attributed to SSRCs associated with the 459 same CNAME is that they are from a particular synchronisation context 460 and can be synchronised at playback. 462 An RTP receiver receiving a previously unseen SSRC value will 463 interpret it as a new source. It might in fact be a previously 464 existing source that had to change SSRC number due to an SSRC 465 conflict. However, the originator of the previous SSRC ought to have 466 ended the conflicting source by sending an RTCP BYE for it prior to 467 starting to send with the new SSRC, so the new SSRC is anyway 468 effectively a new source. 470 4.3. Contributing Source (CSRC) 472 The Contributing Source (CSRC) is not a separate identifier. Rather 473 a synchronisation source identifier is listed as a CSRC in the RTP 474 header of a packet generated by an RTP mixer if the corresponding 475 SSRC was in the header of one of the packets that contributed to the 476 mix. 478 It is not possible, in general, to extract media represented by an 479 individual CSRC since it is typically the result of a media mixing 480 (merge) operation by an RTP mixer on the individual media streams 481 corresponding to the CSRC identifiers. The exception is the case 482 when only a single CSRC is indicated as this represent forwarding of 483 a media stream, possibly modified. The RTP header extension for 484 Mixer-to-Client Audio Level Indication [RFC6465] expands on the 485 receivers information about a packet with a CSRC list. Due to these 486 restrictions, CSRC will not be considered a fully qualified 487 multiplexing point and will be disregarded in the rest of this 488 document. 490 4.4. RTP Payload Type 492 Each Media Stream utilises one or more RTP payload formats. An RTP 493 payload format describes how the output of a particular media codec 494 is framed and encoded into RTP packets. The payload format used is 495 identified by the payload type field in the RTP data packet header. 496 The combination therefore identifies a specific Media Stream encoding 497 format. The format definition can be taken from [RFC3551] for 498 statically allocated payload types, but ought to be explicitly 499 defined in signalling, such as SDP, both for static and dynamic 500 Payload Types. The term "format" here includes whatever can be 501 described by out-of-band signalling means. In SDP, the term "format" 502 includes media type, RTP timestamp sampling rate, codec, codec 503 configuration, payload format configurations, and various robustness 504 mechanisms such as redundant encodings [RFC2198]. 506 The payload type is scoped by sending endpoint within an RTP Session. 507 All synchronisation sources sent from an single endpoint share the 508 same payload types definitions. The RTP Payload Type is designed 509 such that only a single Payload Type is valid at any time instant in 510 the RTP source's RTP timestamp time line, effectively time- 511 multiplexing different Payload Types if any change occurs. The 512 payload type used can change on a per-packet basis for an SSRC, for 513 example a speech codec making use of generic comfort noise [RFC3389]. 514 If there is a true need to send multiple Payload Types for the same 515 SSRC that are valid for the same instant, then redundant encodings 516 [RFC2198] can be used. Several additional constraints than the ones 517 mentioned above need to be met to enable this use, one of which is 518 that the combined payload sizes of the different Payload Types ought 519 not exceed the transport MTU. 521 Other aspects of RTP payload format use are described in RTP Payload 522 HowTo [I-D.ietf-payload-rtp-howto]. 524 The payload type is not a multiplexing point at the RTP layer (see 525 Appendix A for a detailed discussion of why using the payload type as 526 an RTP multiplexing point does not work). The RTP payload type is, 527 however, used to determine how to render a media stream, and so can 528 be viewed as selecting a rendering context. The rendering context 529 can be defined by the signalling, and the RTP payload type number is 530 sometimes used to associate an RTP media stream with the signalling. 531 This association is possible provided unique RTP payload type numbers 532 are used in each context. For example, an RTP media stream can be 533 associated with an SDP "m=" line by comparing the RTP payload type 534 numbers used by the media stream with payload types signalled in the 535 "a=rtpmap:" lines in the media sections of the SDP. If RTP media 536 streams are being associated with signalling contexts based on the 537 RTP payload type, then the assignment of RTP payload type numbers 538 MUST be unique across signalling contexts; if the same RTP payload 539 format configuration is used in multiple contexts, then a different 540 RTP payload type number has to be assigned in each context to ensure 541 uniqueness. If the RTP payload type number is not being used to 542 associated RTP media streams with a signalling context, then the same 543 RTP payload type number can be used to indicate the exact same RTP 544 payload format configuration in multiple contexts. 546 5. RTP Topologies and Issues 548 The impact of how RTP multiplexing is performed will in general vary 549 with how the RTP Session participants are interconnected, described 550 by RTP Topology [RFC5117] and its intended successor 551 [I-D.westerlund-avtcore-rtp-topologies-update]. 553 5.1. Point to Point 555 Even the most basic use case, denoted Topo-Point-to-Point in 556 [I-D.westerlund-avtcore-rtp-topologies-update], raises a number of 557 considerations that are discussed in detail below (Section 6). They 558 range over such aspects as: 560 o Does my communication peer support RTP as defined with multiple 561 SSRCs? 563 o Do I need network differentiation in form of QoS? 565 o Can the application more easily process and handle the media 566 streams if they are in different RTP sessions? 568 o Do I need to use additional media streams for RTP retransmission 569 or FEC. 571 o etc. 573 The point to point topology can contain one to many RTP sessions with 574 one to many media sources per session, each having one or more RTP 575 sources per media source. 577 5.2. Translators & Gateways 579 A point to point communication can end up in a situation when the 580 peer it is communicating with is not compatible with the other peer 581 for various reasons: 583 o No common media codec for a media type thus requiring transcoding 585 o Different support for multiple RTP sources and RTP sessions 587 o Usage of different media transport protocols, i.e RTP or other. 589 o Usage of different transport protocols, e.g. UDP, DCCP, TCP 591 o Different security solutions, e.g. IPsec, TLS, DTLS, SRTP with 592 different keying mechanisms. 594 In many situations this is resolved by the inclusion of a translator 595 between the two peers, as described by Topo-PtP-Translator in 596 [I-D.westerlund-avtcore-rtp-topologies-update]. The translator's 597 main purpose is to make the peer look to the other peer like 598 something it is compatible with. There can also be other reasons 599 than compatibility to insert a translator in the form of a middlebox 600 or gateway, for example a need to monitor the media streams. If the 601 stream transport characteristics are changed by the translator, 602 appropriate media handling can require thorough understanding of the 603 application logic, specifically any congestion control or media 604 adaptation. 606 5.3. Point to Multipoint Using Multicast 608 The Point to Multi-point topology is using Multicast to interconnect 609 the session participants. This includes both Topo-ASM and Topo-SSM 610 in [I-D.westerlund-avtcore-rtp-topologies-update]. 612 Special considerations need to be made as multicast is a one to many 613 distribution system. For example, the only practical method for 614 adapting the bit-rate sent towards a given receiver for large groups 615 is to use a set of multicast groups, where each multicast group 616 represents a particular bit-rate. Otherwise the whole group gets 617 media adapted to the participant with the worst conditions. The 618 media encoding is either scalable, where multiple layers can be 619 combined, or simulcast, where a single version is selected. By 620 either selecting or combing multicast groups, the receiver can 621 control the bit-rate sent on the path to itself. It is also common 622 that streams that improve transport robustness are sent in their own 623 multicast group to allow for interworking with legacy or to support 624 different levels of protection. 626 The result of this is some common behaviours for RTP multicast: 628 1. Multicast applications use a group of RTP sessions, not one. 629 Each endpoint will need to be a member of a number of RTP 630 sessions in order to perform well. 632 2. Within each RTP session, the number of RTP Sinks is likely to be 633 much larger than the number of RTP sources. 635 3. Multicast applications need signalling functions to identify the 636 relationships between RTP sessions. 638 4. Multicast applications need signalling functions to identify the 639 relationships between SSRCs in different RTP sessions. 641 All multicast configurations share a signalling requirement; all of 642 the participants will need to have the same RTP and payload type 643 configuration. Otherwise, A could for example be using payload type 644 97 as the video codec H.264 while B thinks it is MPEG-2. It is to be 645 noted that SDP offer/answer [RFC3264] is not appropriate for ensuring 646 this property. The signalling aspects of multicast are not explored 647 further in this memo. 649 Security solutions for this type of group communications are also 650 challenging. First of all the key-management and the security 651 protocol needs to support group communication. Source authentication 652 requires special solutions. For more discussion on this please 653 review Options for Securing RTP Sessions 654 [I-D.ietf-avtcore-rtp-security-options]. 656 5.4. Point to Multipoint Using an RTP Transport Translator 658 This mode is described as Topo-Translator in 659 [I-D.westerlund-avtcore-rtp-topologies-update]. 661 Transport Translators (Relays) result in an RTP session situation 662 that is very similar to how an ASM group RTP session would behave. 664 One of the most important aspects with the simple relay is that it is 665 only rewriting transport headers, no RTP modifications nor media 666 transcoding occur. The most obvious downside of this basic relaying 667 is that the translator has no control over how many streams need to 668 be delivered to a receiver. Nor can it simply select to deliver only 669 certain streams, as this creates session inconsistencies: If the 670 translator temporarily stops a stream, this prevents some receivers 671 from reporting on it. From the sender's perspective it will look 672 like a transport failure. Applications needing to stop or switch 673 streams in the central node ought to consider using an RTP mixer to 674 avoid this issue. 676 The Transport Translator has the same signalling requirement as 677 multicast: All participants need to have the same payload type 678 configuration. Most of the ASM security issues also arise here. 679 Some alternatives when it comes to solution do exist, as there exists 680 a central node to communicate with, one that also can enforce some 681 security policies depending on the level of trust placed in the node. 683 5.5. Point to Multipoint Using an RTP Mixer 685 A mixer, described by Topo-Mixer in 686 [I-D.westerlund-avtcore-rtp-topologies-update], is a centralised node 687 that selects or mixes content in a conference to optimise the RTP 688 session so that each endpoint only needs connect to one entity, the 689 mixer. The media sent from the mixer to the endpoint can be 690 optimised in different ways. These optimisations include methods 691 like only choosing media from the currently most active speaker or 692 mixing together audio so that only one audio stream is needed. 694 Mixers have some downsides, the first is that the mixer has to be a 695 trusted node as they repacketize the media, and can perform media 696 transformation operations. When using SRTP, both media operations 697 and repacketization requires that the mixer verifies integrity, 698 decrypts the content, performs the operation and forms new RTP 699 packets, encrypts and integrity-protects them. This applies to all 700 types of mixers. The second downside is that all these operations 701 and optimisations of the session requires processing. How much 702 depends on the implementation, as will become evident below. 704 A mixer, unlike a pure transport translator, is always application 705 specific: the application logic for stream mixing or stream selection 706 has to be embedded within the mixer, and controlled using application 707 specific signalling. The implementation of a mixer can take several 708 different forms, as discussed below. 710 A Mixer can also contain translator functionalities, like a media 711 transcoder to adjust the media bit-rate or codec used for a 712 particular RTP media stream. 714 6. RTP Multiplexing: When to Use Multiple RTP Sessions 715 Using multiple media streams is a well supported feature of RTP. 716 However, it can be unclear for most implementers or people writing 717 RTP/RTCP applications or extensions attempting to apply multiple 718 streams when it is most appropriate to add an additional SSRC in an 719 existing RTP session and when it is better to use multiple RTP 720 sessions. This section tries to discuss the various considerations 721 needed. The next section then concludes with some guidelines. 723 6.1. RTP and RTCP Protocol Considerations 725 This section discusses RTP and RTCP aspects worth considering when 726 selecting between using an additional SSRC and Multiple RTP sessions. 728 6.1.1. The RTP Specification 730 RFC 3550 contains some recommendations and a bullet list with 5 731 arguments for different aspects of RTP multiplexing. Let's review 732 Section 5.2 of [RFC3550], reproduced below: 734 "For efficient protocol processing, the number of multiplexing points 735 should be minimised, as described in the integrated layer processing 736 design principle [ALF]. In RTP, multiplexing is provided by the 737 destination transport address (network address and port number) which 738 is different for each RTP session. For example, in a teleconference 739 composed of audio and video media encoded separately, each medium 740 SHOULD be carried in a separate RTP session with its own destination 741 transport address. 743 Separate audio and video streams SHOULD NOT be carried in a single 744 RTP session and demultiplexed based on the payload type or SSRC 745 fields. Interleaving packets with different RTP media types but 746 using the same SSRC would introduce several problems: 748 1. If, say, two audio streams shared the same RTP session and the 749 same SSRC value, and one were to change encodings and thus 750 acquire a different RTP payload type, there would be no general 751 way of identifying which stream had changed encodings. 753 2. An SSRC is defined to identify a single timing and sequence 754 number space. Interleaving multiple payload types would require 755 different timing spaces if the media clock rates differ and would 756 require different sequence number spaces to tell which payload 757 type suffered packet loss. 759 3. The RTCP sender and receiver reports (see Section 6.4) can only 760 describe one timing and sequence number space per SSRC and do not 761 carry a payload type field. 763 4. An RTP mixer would not be able to combine interleaved streams of 764 incompatible media into one stream. 766 5. Carrying multiple media in one RTP session precludes: the use of 767 different network paths or network resource allocations if 768 appropriate; reception of a subset of the media if desired, for 769 example just audio if video would exceed the available bandwidth; 770 and receiver implementations that use separate processes for the 771 different media, whereas using separate RTP sessions permits 772 either single- or multiple-process implementations. 774 Using a different SSRC for each medium but sending them in the same 775 RTP session would avoid the first three problems but not the last 776 two. 778 On the other hand, multiplexing multiple related sources of the same 779 medium in one RTP session using different SSRC values is the norm for 780 multicast sessions. The problems listed above don't apply: an RTP 781 mixer can combine multiple audio sources, for example, and the same 782 treatment is applicable for all of them. It might also be 783 appropriate to multiplex streams of the same medium using different 784 SSRC values in other scenarios where the last two problems do not 785 apply." 787 Let's consider one argument at a time. The first is an argument for 788 using different SSRC for each individual media stream, which is very 789 applicable. 791 The second argument is advocating against using payload type 792 multiplexing, which still stands as can been seen by the extensive 793 list of issues found in Appendix A. 795 The third argument is yet another argument against payload type 796 multiplexing. 798 The fourth is an argument against multiplexing media streams that 799 require different handling into the same session. As we saw in the 800 discussion of RTP mixers, the RTP mixer has to embed application 801 logic in order to handle streams anyway; the separation of streams 802 according to stream type is just another piece of application logic, 803 which might or might not be appropriate for a particular application. 804 A type of application that can mix different media sources "blindly" 805 is the audio only "telephone" bridge; most other type of application 806 needs application-specific logic to perform the mix correctly. 808 The fifth argument discusses network aspects that we will discuss 809 more below in Section 6.3. It also goes into aspects of 810 implementation, like decomposed endpoints where different processes 811 or inter-connected devices handle different aspects of the whole 812 multi-media session. 814 A summary of RFC 3550's view on multiplexing is to use unique SSRCs 815 for anything that is its own media/packet stream, and to use 816 different RTP sessions for media streams that don't share a media 817 type. This document supports the first point; it is very valid. The 818 later is one thing which is further discussed in this document as 819 something the application developer needs to make a conscious choice 820 for, but where imposing a single solution on all usages of RTP is 821 inappropriate. 823 6.1.1.1. Different Media Types: Recommendations 825 The above quote from RTP [RFC3550] includes a strong recommendation: 827 "For example, in a teleconference composed of audio and video 828 media encoded separately, each medium SHOULD be carried in a 829 separate RTP session with its own destination transport address." 831 It was identified in "Why RTP Sessions Should Be Content Neutral" 832 [I-D.alvestrand-rtp-sess-neutral] that the above statement is poorly 833 supported by any of the motivations provided in the RTP 834 specification. This has resulted in the creation of a specification 835 Multiple Media Types in an RTP Session specification 836 [I-D.ietf-avtcore-multi-media-rtp-session] which intends to update 837 this recommendation. That document has a detailed analysis of the 838 potential issues in having multiple media types in the same RTP 839 session. This document tries to provide an moreover arching 840 consideration regarding the usage of RTP session and considers 841 multiple media types in one RTP session as possible choice for the 842 RTP application designer. 844 6.1.2. Multiple SSRCs in a Session 846 Using multiple SSRCs in an RTP session at one endpoint requires 847 resolving some unclear aspects of the RTP specification. These could 848 potentially lead to some interoperability issues as well as some 849 potential significant inefficencies. These are further discussed in 850 "RTP Considerations for Endpoints Sending Multiple Media Streams" 851 [I-D.lennox-avtcore-rtp-multi-stream]. A application designer needs 852 to consider these issues and the impact availability or lack of the 853 optimization in the endpoints has on their application. 855 If an application will become affected by the issues described, using 856 Multiple RTP sessions can mitigate these issues. 858 6.1.3. Handling Varying Sets of Senders 860 In some applications, the set of simultaneously active sources varies 861 within a larger set of session members. A receiver can then possibly 862 try to use a set of decoding chains that is smaller than the number 863 of senders, switching the decoding chains between different senders. 864 As each media decoding chain can contain state, either the receiver 865 needs to either be able to save the state of swapped-out senders, or 866 the sender needs to be able to send data that permits the receiver to 867 reinitialise when it resumes activity. 869 This behaviour will cause similar issues independent of Additional 870 SSRC or Multiple RTP session. 872 6.1.4. Cross Session RTCP Requests 874 There currently exists no functionality to make truly synchronised 875 and atomic RTCP messages with some type of request semantics across 876 multiple RTP Sessions. Instead, separate RTCP messages will have to 877 be sent in each session. This gives streams in the same RTP session 878 a slight advantage as RTCP messages for different streams in the same 879 session can be sent in a compound RTCP packet, thus providing an 880 atomic operation if different modifications of different streams are 881 requested at the same time. 883 When using multiple RTP sessions, the RTCP timing rules in the 884 sessions and the transport aspects, such as packet loss and jitter, 885 prevents a receiver from relying on atomic operations, forcing it to 886 use more robust and forgiving mechanisms. 888 6.1.5. Binding Related Sources 890 A common problem in a number of various RTP extensions has been how 891 to bind related RTP sources and their media streams together. This 892 issue is common to both using additional SSRCs and Multiple RTP 893 sessions. 895 The solutions can be divided into some groups, RTP/RTCP based, 896 Signalling based (SDP), grouping related RTP sessions, and grouping 897 SSRCs within an RTP session. Most solutions are explicit, but some 898 implicit methods have also been applied to the problem. 900 The SDP-based signalling solutions are: 902 SDP Media Description Grouping: The SDP Grouping Framework [RFC5888] 903 uses various semantics to group any number of media descriptions. 904 These has previously been considered primarily as grouping RTP 905 sessions, but this might change. 907 SDP SSRC grouping: Source-Specific Media Attributes in SDP [RFC5576] 908 includes a solution for grouping SSRCs the same way as the 909 Grouping framework groupes Media Descriptions. 911 SDP MSID grouping: Media Stream Identifiers [I-D.ietf-mmusic-msid] 912 includes a solution for grouping SSRCs that is independent of 913 their allocation to RTP sessions. 915 This supports a lot of use cases. All these solutions have 916 shortcomings in cases where the session's dynamic properties are such 917 that it is difficult or resource consuming to keep the list of 918 related SSRCs up to date. 920 Within RTP/RTCP based solutions when binding to a endpoint or 921 synchronization context, i.e. the CNAME has not be sufficient and 922 one has multiple RTP sessions has been to using the same SSRC value 923 across all the RTP sessions. RTP Retransmission [RFC4588] is 924 multiple RTP session mode, Generic FEC [RFC5109], as well as the RTP 925 payload format for Scalable Video Coding [RFC6190] in Multi Session 926 Transmission (MST) mode uses this method. This method clearly works 927 but might have some downside in RTP sessions with many participating 928 SSRCs. The birthday paradox ensures that if you populate a single 929 session with 9292 SSRCs at random, the chances are approximately 1% 930 that at least one collision will occur. When a collision occur this 931 will force one to change SSRC in all RTP sessions and thus 932 resynchronizing all of them instead of only the single media stream 933 having the collision. 935 It can be noted that Section 8.3 of the RTP Specification [RFC3550] 936 recommends using a single SSRC space across all RTP sessions for 937 layered coding. 939 Another solution that has been applied to binding SSRCs has been an 940 implicit method used by RTP Retransmission [RFC4588] when doing 941 retransmissions in the same RTP session as the source RTP media 942 stream. This issues an RTP retransmission request, and then await a 943 new SSRC carrying the RTP retransmission payload and where that SSRC 944 is from the same CNAME. This limits a requestor to having only one 945 outstanding request on any new source SSRCs per endpoint. 947 There exists no RTP/RTCP based mechanism capable of supporting 948 explicit association accross multiple RTP sessions as well within an 949 RTP session. A proposed solution for handling this issue is 950 [I-D.westerlund-avtext-rtcp-sdes-srcname]. If accepted, this can 951 potentially also be part of an SDP based solution also by reusing the 952 same identifiers and name space. 954 6.1.6. Forward Error Correction 956 There exist a number of Forward Error Correction (FEC) based schemes 957 for how to reduce the packet loss of the original streams. Most of 958 the FEC schemes will protect a single source flow. The protection is 959 achieved by transmitting a certain amount of redundant information 960 that is encoded such that it can repair one or more packet losses 961 over the set of packets they protect. This sequence of redundant 962 information also needs to be transmitted as its own media stream, or 963 in some cases instead of the original media stream. Thus many of 964 these schemes create a need for binding related flows as discussed 965 above. Looking at the history of these schemes, there are schemes 966 using multiple SSRCs and schemes using multiple RTP sessions, and 967 some schemes that support both modes of operation. 969 Using multiple RTP sessions supports the case where some set of 970 receivers might not be able to utilise the FEC information. By 971 placing it in a separate RTP session, it can easily be ignored. 973 In usages involving multicast, having the FEC information on its own 974 multicast group, and therefore in its own RTP session, allows for 975 flexibility. This is especially useful when receivers see very 976 heterogeneous packet loss rates. Those receivers that are not seeing 977 packet loss don't need to join the multicast group with the FEC data, 978 and so avoid the overhead of receiving unnecessary FEC packets, for 979 example. 981 6.1.7. Transport Translator Sessions 983 A basic Transport Translator relays any incoming RTP and RTCP packets 984 to the other participants. The main difference between Additional 985 SSRCs and Multiple RTP Sessions resulting from this use case is that 986 with Additional SSRCs it is not possible for a particular session 987 participant to decide to receive a subset of media streams. When 988 using separate RTP sessions for the different sets of media streams, 989 a single participant can choose to leave one of the sessions but not 990 the other. 992 6.2. Interworking Considerations 994 There are several different kinds of interworking, and this section 995 discusses two related ones. The interworking between different 996 applications and the implications of potentially different choices of 997 usage of RTP's multiplexing points. The second topic relates to what 998 limitations have to be considered working with some legacy 999 applications. 1001 6.2.1. Types of Interworking 1003 It is not uncommon that applications or services of similar usage, 1004 especially the ones intended for interactive communication, encounter 1005 a situation where one want to interconnect two or more of these 1006 applications. 1008 In these cases one ends up in a situation where one might use a 1009 gateway to interconnect applications. This gateway then needs to 1010 change the multiplexing structure or adhere to limitations in each 1011 application. 1013 There are two fundamental approaches to gatewaying: RTP Translator 1014 interworking (RTP bridging), where the gateway acts as an RTP 1015 Translator, and the two applications are members of the same RTP 1016 session, and Gateway Interworking (with RTP termination), where there 1017 are independent RTP sessions running from each interconnected 1018 application to the gateway. 1020 6.2.2. RTP Translator Interworking 1022 From an RTP perspective the RTP Translator approach could work if all 1023 the applications are using the same codecs with the same payload 1024 types, have made the same multiplexing choices, have the same 1025 capabilities in number of simultaneous media streams combined with 1026 the same set of RTP/RTCP extensions being supported. Unfortunately 1027 this might not always be true. 1029 When one is gatewaying via an RTP Translator, a natural requirement 1030 is that the two applications being interconnected need to use the 1031 same approach to multiplexing. Furthermore, if one of the 1032 applications is capable of working in several modes (such as being 1033 able to use Additional SSRCs or Multiple RTP sessions at will), and 1034 the other one is not, successful interconnection depends on locking 1035 the more flexible application into the operating mode where 1036 interconnection can be successful, even if no participants using the 1037 less flexible application are present when the RTP sessions are being 1038 created. 1040 6.2.3. Gateway Interworking 1042 When one terminates RTP sessions at the gateway, there are certain 1043 tasks that the gateway has to carry out: 1045 o Generating appropriate RTCP reports for all media streams 1046 (possibly based on incoming RTCP reports), originating from SSRCs 1047 controlled by the gateway. 1049 o Handling SSRC collision resolution in each application's RTP 1050 sessions. 1052 o Signalling, choosing and policing appropriate bit-rates for each 1053 session. 1055 If either of the applications has any security applied, e.g. in the 1056 form of SRTP, the gateway needs to be able to decrypt incoming 1057 packets and re-encrypt them in the other application's security 1058 context. This is necessary even if all that's needed is a simple 1059 remapping of SSRC numbers. If this is done, the gateway also needs 1060 to be a member of the security contexts of both sides, of course. 1062 Other tasks a gateway might need to apply include transcoding (for 1063 incompatible codec types), rescaling (for incompatible video size 1064 requirements), suppression of content that is known not to be handled 1065 in the destination application, or the addition or removal of 1066 redundancy coding or scalability layers to fit the need of the 1067 destination domain. 1069 From the above, we can see that the gateway needs to have an intimate 1070 knowledge of the application requirements; a gateway is by its nature 1071 application specific, not a commodity product. 1073 This fact reveals the potential for these gateways to block evolution 1074 of the applications by blocking unknown RTP and RTCP extensions that 1075 the regular application has been extended with. 1077 If one uses security functions, like SRTP, they can as seen above 1078 incur both additional risk due to the gateway needing to be in 1079 security association between the endpoints, unless the gateway is on 1080 the transport level, and additional complexities in form of the 1081 decrypt-encrypt cycles needed for each forwarded packet. SRTP, due 1082 to its keying structure, also requires that each RTP session needs 1083 different master keys, as use of the same key in two RTP sessions can 1084 result in two-time pads that completely breaks the confidentiality of 1085 the packets. 1087 6.2.4. Multiple SSRC Legacy Considerations 1088 Historically, the most common RTP use cases have been point to point 1089 Voice over IP (VoIP) or streaming applications, commonly with no more 1090 than one media source per endpoint and media type (typically audio 1091 and video). Even in conferencing applications, especially voice 1092 only, the conference focus or bridge has provided a single stream 1093 with a mix of the other participants to each participant. It is also 1094 common to have individual RTP sessions between each endpoint and the 1095 RTP mixer, meaning that the mixer functions as an RTP-terminating 1096 gateway. 1098 When establishing RTP sessions that can contain endpoints that aren't 1099 updated to handle multiple streams following these recommendations, a 1100 particular application can have issues with multiple SSRCs within a 1101 single session. These issues include: 1103 1. Need to handle more than one stream simultaneously rather than 1104 replacing an already existing stream with a new one. 1106 2. Be capable of decoding multiple streams simultaneously. 1108 3. Be capable of rendering multiple streams simultaneously. 1110 This indicates that gateways attempting to interconnect to this class 1111 of devices has to make sure that only one media stream of each type 1112 gets delivered to the endpoint if it's expecting only one, and that 1113 the multiplexing format is what the device expects. It is highly 1114 unlikely that RTP translator-based interworking can be made to 1115 function successfully in such a context. 1117 6.3. Network Considerations 1119 The multiplexing choice has impact on network level mechanisms that 1120 need to be considered by the implementor. 1122 6.3.1. Quality of Service 1124 When it comes to Quality of Service mechanisms, they are either flow 1125 based or marking based. RSVP [RFC2205] is an example of a flow based 1126 mechanism, while Diff-Serv [RFC2474] is an example of a Marking based 1127 one. For a marking based scheme, the method of multiplexing will not 1128 affect the possibility to use QoS. 1130 However, for a flow based scheme there is a clear difference between 1131 the methods. Additional SSRC will result in all media streams being 1132 part of the same 5-tuple (protocol, source address, destination 1133 address, source port, destination port) which is the most common 1134 selector for flow based QoS. Thus, separation of the level of QoS 1135 between media streams is not possible. That is however possible when 1136 using multiple RTP sessions, where each media stream for which a 1137 separate QoS handling is desired can be in a different RTP session 1138 that can be sent over different 5-tuples. 1140 6.3.2. NAT and Firewall Traversal 1142 In today's network there exist a large number of middleboxes. The 1143 ones that normally have most impact on RTP are Network Address 1144 Translators (NAT) and Firewalls (FW). 1146 Below we analyze and comment on the impact of requiring more 1147 underlying transport flows in the presence of NATs and Firewalls: 1149 End-Point Port Consumption: A given IP address only has 65536 1150 available local ports per transport protocol for all consumers of 1151 ports that exist on the machine. This is normally never an issue 1152 for an end-user machine. It can become an issue for servers that 1153 handle large number of simultaneous streams. However, if the 1154 application uses ICE to authenticate STUN requests, a server can 1155 serve multiple endpoints from the same local port, and use the 1156 whole 5-tuple (source and destination address, source and 1157 destination port, protocol) as identifier of flows after having 1158 securely bound them to the remote endpoint address using the STUN 1159 request. In theory the minimum number of media server ports 1160 needed are the maximum number of simultaneous RTP Sessions a 1161 single endpoint can use. In practice, implementation will 1162 probably benefit from using more server ports to simplify 1163 implementation or avoid performance bottlenecks. 1165 NAT State: If an endpoint sits behind a NAT, each flow it generates 1166 to an external address will result in a state that has to be kept 1167 in the NAT. That state is a limited resource. In home or Small 1168 Office/Home Office (SOHO) NATs, memory or processing are usually 1169 the most limited resources. For large scale NATs serving many 1170 internal endpoints, available external ports are likely the scarce 1171 resource. Port limitations is primarily a problem for larger 1172 centralised NATs where endpoint independent mapping requires each 1173 flow to use one port for the external IP address. This affects 1174 the maximum number of internal users per external IP address. 1175 However, it is worth pointing out that a real-time video 1176 conference session with audio and video is likely using less than 1177 10 UDP flows, compared to certain web applications that can use 1178 100+ TCP flows to various servers from a single browser instance. 1180 NAT Traversal Excess Time: Making the NAT/FW traversal takes a 1181 certain amount of time for each flow. It also takes time in a 1182 phase of communication between accepting to communicate and the 1183 media path being established which is fairly critical. The best 1184 case scenario for how much extra time it takes after finding the 1185 first valid candidate pair following the specified ICE procedures 1186 are: 1.5*RTT + Ta*(Additional_Flows-1), where Ta is the pacing 1187 timer, which ICE specifies to be no smaller than 20 ms. That 1188 assumes a message in one direction, and then an immediate 1189 triggered check back. The reason it isn't more, is that ICE first 1190 finds one candidate pair that works prior to attempting to 1191 establish multiple flows. Thus, there is no extra time until one 1192 has found a working candidate pair. Based on that working pair 1193 the needed extra time is to in parallel establish the, in most 1194 cases 2-3, additional flows. However, packet loss causes extra 1195 delays, at least 100 ms, which is the minimal retransmission timer 1196 for ICE. 1198 NAT Traversal Failure Rate: Due to the need to establish more than a 1199 single flow through the NAT, there is some risk that establishing 1200 the first flow succeeds but that one or more of the additional 1201 flows fail. The risk that this happens is hard to quantify, but 1202 ought to be fairly low as one flow from the same interfaces has 1203 just been successfully established. Thus only rare events such as 1204 NAT resource overload, or selecting particular port numbers that 1205 are filtered etc, ought to be reasons for failure. 1207 Deep Packet Inspection and Multiple Streams: Firewalls differ in how 1208 deeply they inspect packets. There exist some potential that 1209 deeply inspecting firewalls will have similar legacy issues with 1210 multiple SSRCs as some stack implementations. 1212 Additional SSRC keeps the additional media streams within one RTP 1213 Session and transport flow and does not introduce any additional NAT 1214 traversal complexities per media stream. This can be compared with 1215 normally one or two additional transport flows per RTP session when 1216 using multiple RTP sessions. Additional lower layer transport flows 1217 will be needed, unless an explicit de-multiplexing layer is added 1218 between RTP and the transport protocol. A proposal for how to 1219 multiplex multiple RTP sessions over the same single lower layer 1220 transport exist in [I-D.westerlund-avtcore-transport-multiplexing]. 1222 6.3.3. Multicast 1224 Multicast groups provides a powerful semantics for a number of real- 1225 time applications, especially the ones that desire broadcast-like 1226 behaviours with one endpoint transmitting to a large number of 1227 receivers, like in IPTV. But that same semantics do result in a 1228 certain number of limitations. 1230 One limitation is that for any group, sender side adaptation to the 1231 actual receiver properties causes degradation for all participants to 1232 what is supported by the receiver with the worst conditions among the 1233 group participants. In most cases this is not acceptable. Instead 1234 various receiver based solutions are employed to ensure that the 1235 receivers achieve best possible performance. By using scalable 1236 encoding and placing each scalability layer in a different multicast 1237 group, the receiver can control the amount of traffic it receives. 1238 To have each scalability layer on a different multicast group, one 1239 RTP session per multicast group is used. 1241 In addition, the transport flow considerations in multicast are a bit 1242 different from unicast; NATs are not useful in the multicast 1243 environment, meaning that the entire port range of each multicast 1244 address is available for distinguishing between RTP sessions. 1246 Thus it appears easiest and most straightforward to use multiple RTP 1247 sessions for sending different media flows used for adapting to 1248 network conditions. 1250 6.3.4. Multiplexing multiple RTP Session on a Single Transport 1252 For applications that don't need flow based QoS and like to save 1253 ports and NAT/FW traversal costs and where usage of multiple media 1254 types in one RTP session is not suitable, there is a proposal for how 1255 to achieve multiplexing of multiple RTP sessions over the same lower 1256 layer transport [I-D.westerlund-avtcore-transport-multiplexing]. 1257 Using such a solution would allow Multiple RTP session without most 1258 of the perceived downsides of Multiple RTP sessions creating a need 1259 for additional transport flows, but this solution would require 1260 support from all functions that handle RTP packets, including 1261 firewalls. 1263 6.4. Security and Key Management Considerations 1265 When dealing with point-to-point, 2-member RTP sessions only, there 1266 are few security issues that are relevant to the choice of having one 1267 RTP session or multiple RTP sessions. However, there are a few 1268 aspects of multiparty sessions that might warrant consideration. For 1269 general information of possible methods of securing RTP, please 1270 review RTP Security Options [I-D.ietf-avtcore-rtp-security-options]. 1272 6.4.1. Security Context Scope 1274 When using SRTP [RFC3711] the security context scope is important and 1275 can be a necessary differentiation in some applications. As SRTP's 1276 crypto suites (so far) are built around symmetric keys, the receiver 1277 will need to have the same key as the sender. This results in that 1278 no one in a multi-party session can be certain that a received packet 1279 really was sent by the claimed sender or by another party having 1280 access to the key. In most cases this is a sufficient security 1281 property, but there are a few cases where this does create issues. 1283 The first case is when someone leaves a multi-party session and one 1284 wants to ensure that the party that left can no longer access the 1285 media streams. This requires that everyone re-keys without 1286 disclosing the keys to the excluded party. 1288 A second case is when using security as an enforcing mechanism for 1289 differentiation. Take for example a scalable layer or a high quality 1290 simulcast version which only premium users are allowed to access. 1291 The mechanism preventing a receiver from getting the high quality 1292 stream can be based on the stream being encrypted with a key that 1293 user can't access without paying premium, having the key-management 1294 limit access to the key. 1296 SRTP [RFC3711] has no special functions for dealing with different 1297 sets of master keys for different SSRCs. The key-management 1298 functions have different capabilities to establish different set of 1299 keys, normally on a per endpoint basis. For example, DTLS-SRTP 1300 [RFC5764] and Security Descriptions [RFC4568] establish different 1301 keys for outgoing and incoming traffic from an endpoint. This key 1302 usage has to be written into the cryptographic context, possibly 1303 associated with different SSRCs. 1305 6.4.2. Key Management for Multi-party session 1307 Performing key-management for multi-party session can be a challenge. 1308 This section considers some of the issues. 1310 Multi-party sessions, such as transport translator based sessions and 1311 multicast sessions, cannot use Security Description [RFC4568] nor 1312 DTLS-SRTP [RFC5764] without an extension as each endpoint provides 1313 its set of keys. In centralised conferences, the signalling 1314 counterpart is a conference server and the media plane unicast 1315 counterpart (to which DTLS messages would be sent) is the transport 1316 translator. Thus an extension like Encrypted Key Transport 1317 [I-D.ietf-avt-srtp-ekt] is needed or a MIKEY [RFC3830] based solution 1318 that allows for keying all session participants with the same master 1319 key. 1321 6.4.3. Complexity Implications 1323 The usage of security functions can surface complexity implications 1324 of the choice of multiplexing and topology. This becomes especially 1325 evident in RTP topologies having any type of middlebox that processes 1326 or modifies RTP/RTCP packets. Where there is very small overhead for 1327 an RTP translator or mixer to rewrite an SSRC value in the RTP packet 1328 of an unencrypted session, the cost of doing it when using 1329 cryptographic security functions is higher. For example if using 1330 SRTP [RFC3711], the actual security context and exact crypto key are 1331 determined by the SSRC field value. If one changes it, the 1332 encryption and authentication tag needs to be performed using another 1333 key. Thus changing the SSRC value implies a decryption using the old 1334 SSRC and its security context followed by an encryption using the new 1335 one. 1337 7. Archetypes 1339 This section discusses some archetypes of how RTP multiplexing can be 1340 used in applications to achieve certain goals and a summary of their 1341 implications. For each archetype there is discussion of benefits and 1342 downsides. 1344 7.1. Single SSRC per Session 1346 In this archetype each endpoint in a point-to-point session has only 1347 a single SSRC, thus the RTP session contains only two SSRCs, one 1348 local and one remote. This session can be used both unidirectional, 1349 i.e. only a single media stream or bi-directional, i.e. both 1350 endpoints have one media stream each. If the application needs 1351 additional media flows between the endpoints, they will have to 1352 establish additional RTP sessions. 1354 The Pros: 1356 1. This archetype has great legacy interoperability potential as it 1357 will not tax any RTP stack implementations. 1359 2. The signalling has good possibilities to negotiate and describe 1360 the exact formats and bit-rates for each media stream, especially 1361 using today's tools in SDP. 1363 3. It does not matter if usage or purpose of the media stream is 1364 signalled on media stream level or session level as there is no 1365 difference. 1367 4. It is possible to control security association per RTP media 1368 stream with current key-management, since each media stream is 1369 directly related to an RTP session, and the keying operates on a 1370 per-session basis. 1372 The Cons: 1374 a. The number of RTP sessions grows directly in proportion with the 1375 number of media streams, which has the implications: 1377 * Linear growth of the amount of NAT/FW state with number of 1378 media streams. 1380 * Increased delay and resource consumption from NAT/FW 1381 traversal. 1383 * Likely larger signalling message and signalling processing 1384 requirement due to the amount of session related information. 1386 * Higher potential for a single media stream to fail during 1387 transport between the endpoints. 1389 b. When the number of RTP sessions grows, the amount of explicit 1390 state for relating media stream also grows, linearly or possibly 1391 exponentially, depending on how the application needs to relate 1392 media streams. 1394 c. The port consumption might become a problem for centralised 1395 services, where the central node's port consumption grows rapidly 1396 with the number of sessions. 1398 d. For applications where the media streams are highly dynamic in 1399 their usage, i.e. entering and leaving, the amount of signalling 1400 can grow high. Issues arising from the timely establishment of 1401 additional RTP sessions can also arise. 1403 e. Cross session RTCP requests might be needed, and the fact that 1404 they're impossible can cause issues. 1406 f. If the same SSRC value is reused in multiple RTP sessions rather 1407 than being randomly chosen, interworking with applications that 1408 uses another multiplexing structure than this application will 1409 require SSRC translation. 1411 g. Cannot be used with Any Source Multicast (ASM) as one cannot 1412 guarantee that only two endpoints participate as packet senders. 1413 Using SSM, it is possible to restrict to these requirements if no 1414 RTCP feedback is injected back into the SSM group. 1416 h. For most security mechanisms, each RTP session or transport flow 1417 requires individual key-management and security association 1418 establishment thus increasing the overhead. 1420 RTP applications that need to inter-work with legacy RTP 1421 applications, like most deployed VoIP and video conferencing 1422 solutions, can potentially benefit from this structure. However, a 1423 large number of media descriptions in SDP can also run into issues 1424 with existing implementations. For any application needing a larger 1425 number of media flows, the overhead can become very significant. 1426 This structure is also not suitable for multi-party sessions, as any 1427 given media stream from each participant, although having same usage 1428 in the application, needs its own RTP session. In addition, the 1429 dynamic behaviour that can arise in multi-party applications can tax 1430 the signalling system and make timely media establishment more 1431 difficult. 1433 7.2. Multiple SSRCs of the Same Media Type 1435 In this archetype, each RTP session serves only a single media type. 1436 The RTP session can contain multiple media streams, either from a 1437 single endpoint or from multiple endpoints. This commonly creates a 1438 low number of RTP sessions, typically only one for audio and one for 1439 video, with a corresponding need for two listening ports when using 1440 RTP/RTCP multiplexing. 1442 The Pros: 1444 1. Low number of RTP sessions needed compared to single SSRC case. 1445 This implies: 1447 * Reduced NAT/FW state 1449 * Lower NAT/FW Traversal Cost in both processing and delay. 1451 2. Allows for early de-multiplexing in the processing chain in RTP 1452 applications where all media streams of the same type have the 1453 same usage in the application. 1455 3. Works well with media type de-composite endpoints. 1457 4. Enables Flow-based QoS with different prioritisation between 1458 media types. 1460 5. For applications with dynamic usage of media streams, i.e. they 1461 come and go frequently, having much of the state associated with 1462 the RTP session rather than an individual SSRC can avoid the need 1463 for in-session signalling of meta-information about each SSRC. 1465 6. Low overhead for security association establishment. 1467 The Cons: 1469 a. May have some need for cross session RTCP requests for things 1470 that affect both media types in an asynchronous way. 1472 b. Some potential for concern with legacy implementations that does 1473 not support the RTP specification fully when it comes to handling 1474 multiple SSRC per endpoint. 1476 c. Will not be able to control security association for sets of 1477 media streams within the same media type with today's key- 1478 management mechanisms, unless these are split into different RTP 1479 sessions. 1481 For RTP applications where all media streams of the same media type 1482 share same usage, this structure provides efficiency gains in amount 1483 of network state used and provides more fate sharing with other media 1484 flows of the same type. At the same time, it is still maintaining 1485 almost all functionalities when it comes to negotiation in the 1486 signalling of the properties for the individual media type and also 1487 enabling flow based QoS prioritisation between media types. It 1488 handles multi-party session well, independently of multicast or 1489 centralised transport distribution, as additional sources can 1490 dynamically enter and leave the session. 1492 7.3. Multiple Sessions for one Media type 1494 In this archetype one goes one step further than in the above 1495 (Section 7.2) by using multiple RTP sessions also for a single media 1496 type, but still not as far as having a single SSRC per RTP session. 1497 The main reason for going in this direction is that the RTP 1498 application needs separation of the media streams due to their usage. 1499 Some typical reasons for going to this archetype are scalability over 1500 multicast, simulcast, need for extended QoS prioritisation of media 1501 streams due to their usage in the application, or the need for fine- 1502 grained signalling using today's tools. 1504 The Pros: 1506 1. More suitable for Multicast usage where receivers can 1507 individually select which RTP sessions they want to participate 1508 in, assuming each RTP session has its own multicast group. 1510 2. Indication of the application's usage of the media stream, where 1511 multiple different usages exist. 1513 3. Less need for SSRC specific explicit signalling for each media 1514 stream and thus reduced need for explicit and timely signalling. 1516 4. Enables detailed QoS prioritisation for flow based mechanisms. 1518 5. Works well with de-composite endpoints. 1520 6. Handles dynamic usage of media streams well. 1522 7. For transport translator based multi-party sessions, this 1523 structure allows for improved control of which type of media 1524 streams an endpoint receives. 1526 8. The scope for who is included in a security association can be 1527 structured around the different RTP sessions, thus enabling such 1528 functionality with existing key-management. 1530 The Cons: 1532 a. Increases the amount of RTP sessions compared to Multiple SSRCs 1533 of the Same Media Type. 1535 b. Increased amount of session configuration state. 1537 c. May need synchronised cross-session RTCP requests and require 1538 some consideration due to this. 1540 d. For media streams that are part of scalability, simulcast or 1541 transport robustness it will be needed to bind sources, which 1542 need to support multiple RTP sessions. 1544 e. Some potential for concern with legacy implementations that does 1545 not support the RTP specification fully when it comes to handling 1546 multiple SSRC per endpoint. 1548 f. Higher overhead for security association establishment. 1550 g. If the applications need finer control than on media type level 1551 over which session participants that are included in different 1552 sets of security associations, most of today's key-management 1553 will have difficulties establishing such a session. 1555 For more complex RTP applications that have several different usages 1556 for media streams of the same media type and / or uses scalability or 1557 simulcast, this solution can enable those functions at the cost of 1558 increased overhead associated with the additional sessions. This 1559 type of structure is suitable for more advanced applications as well 1560 as multicast based applications requiring differentiation to 1561 different participants. 1563 7.4. Multiple Media Types in one Session 1565 This archetype is to use a single RTP session for multiple different 1566 media types, like audio and video, and possibly also transport 1567 robustness mechanisms like FEC or Retransmission. Each media stream 1568 will use its own SSRC and a given SSRC value from a particular 1569 endpoint will never use the SSRC for more than a single media type. 1571 The Pros: 1573 1. Single RTP session which implies: 1575 * Minimal NAT/FW state. 1577 * Minimal NAT/FW Traversal Cost. 1579 * Fate-sharing for all media flows. 1581 2. Enables separation of the different media types based on the 1582 payload types so media type specific endpoint or central 1583 processing can still be supported despite single session. 1585 3. Can handle dynamic allocations of media streams well on an RTP 1586 level. Depends on the application's needs for explicit 1587 indication of the stream usage and how timely that can be 1588 signalled. 1590 4. Minimal overhead for security association establishment. 1592 The Cons: 1594 a. Less suitable for interworking with other applications that uses 1595 individual RTP sessions per media type or multiple sessions for a 1596 single media type, due to need of SSRC translation. 1598 b. Negotiation of bandwidth for the different media types is 1599 currently not possible in SDP. This requires SDP extensions to 1600 enable payload or source specific bandwidth. Likely to be a 1601 problem due to media type asymmetry in needed bandwidth. 1603 c. Not suitable for de-composite endpoints. 1605 d. Flow based QoS cannot provide separate treatment to some media 1606 streams compared to others in the single RTP session. 1608 e. If there is significant asymmetry between the media streams' RTCP 1609 reporting needs, there are some challenges in configuration and 1610 usage to avoid wasting RTCP reporting on the media stream that 1611 does not need that frequent reporting. 1613 f. Not suitable for applications where some receivers like to 1614 receive only a subset of the media streams, especially if 1615 multicast or transport translator is being used. 1617 g. Additional concern with legacy implementations that do not 1618 support the RTP specification fully when it comes to handling 1619 multiple SSRC per endpoint, as also multiple simultaneous media 1620 types needs to be handled. 1622 h. If the applications need finer control over which session 1623 participants that are included in different sets of security 1624 associations, most key-management will have difficulties 1625 establishing such a session. 1627 7.5. Summary 1629 There are some clear relations between these archetypes. Both the 1630 "single SSRC per RTP session" and the "multiple media types in one 1631 session" are cases which require full explicit signalling of the 1632 media stream relations. However, they operate on two different 1633 levels where the first primarily enables session level binding, and 1634 the second needs to do it all on SSRC level. From another 1635 perspective, the two solutions are the two extreme points when it 1636 comes to number of RTP sessions needed. 1638 The two other archetypes "Multiple SSRCs of the Same Media Type" and 1639 "Multiple Sessions for one Media Type" are examples of two other 1640 cases that first of all allows for some implicit mapping of the role 1641 or usage of the media streams based on which RTP session they appear 1642 in. It thus potentially allows for less signalling and in particular 1643 reduced need for real-time signalling in dynamic sessions. They also 1644 represent points in between the first two when it comes to amount of 1645 RTP sessions established, i.e. representing an attempt to reduce the 1646 amount of sessions as much as possible without compromising the 1647 functionality the session provides both on network level and on 1648 signalling level. 1650 8. Summary considerations and guidelines 1652 8.1. Guidelines 1654 This section contains a number of recommendations for implementors or 1655 specification writers when it comes to handling multi-stream. 1657 Do not Require the same SSRC across Sessions: As discussed in 1658 Section 6.1.5 there exist drawbacks in using the same SSRC in 1659 multiple RTP sessions as a mechanism to bind related media streams 1660 together. It is instead suggested that a mechanism to explicitly 1661 signal the relation is used, either in RTP/RTCP or in the used 1662 signalling mechanism that establishes the RTP session(s). 1664 Use additional SSRCs additional Media Sources: In the cases where an 1665 RTP endpoint needs to transmit additional media streams of the 1666 same media type in the application, with the same processing 1667 requirements at the network and RTP layers, it is suggested to 1668 send them as additional SSRCs in the same RTP session. For 1669 example a telepresence room where there are three cameras, and 1670 each camera captures 2 persons sitting at the table, sending each 1671 camera as its own SSRC within a single RTP session is suggested. 1673 Use additional RTP sessions for streams with different requirements: 1674 When media streams have different processing requirements from the 1675 network or the RTP layer at the endpoints, it is suggested that 1676 the different types of streams are put in different RTP sessions. 1677 This includes the case where different participants want different 1678 subsets of the set of RTP streams. 1680 When using multiple RTP Sessions use grouping: When using Multiple 1681 RTP session solutions, it is suggested to explicitly group the 1682 involved RTP sessions when needed using the signalling mechanism, 1683 for example The Session Description Protocol (SDP) Grouping 1684 Framework. [RFC5888], using some appropriate grouping semantics. 1686 RTP/RTCP Extensions May Support Additional SSRCs as well as Multiple RTP sessions: 1687 When defining an RTP or RTCP extension, the creator needs to 1688 consider if this extension is applicable to usage with additional 1689 SSRCs and Multiple RTP sessions. Any extension intended to be 1690 generic is suggested to support both. Applications that are not 1691 as generally applicable will have to consider if interoperability 1692 is better served by defining a single solution or providing both 1693 options. 1695 Transport Support Extensions: When defining new RTP/RTCP extensions 1696 intended for transport support, like the retransmission or FEC 1697 mechanisms, they are expected to include support for both 1698 additional SSRCs and multiple RTP sessions so that application 1699 developers can choose freely from the set of mechanisms without 1700 concerning themselves with which of the multiplexing choices a 1701 particular solution supports. 1703 9. IANA Considerations 1704 This document makes no request of IANA. 1706 Note to RFC Editor: this section can be removed on publication as an 1707 RFC. 1709 10. Security Considerations 1711 There is discussion of the security implications of choosing SSRC vs 1712 Multiple RTP session in Section 6.4. 1714 11. References 1716 11.1. Normative References 1718 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 1719 Jacobson, "RTP: A Transport Protocol for Real-Time 1720 Applications", STD 64, RFC 3550, July 2003. 1722 11.2. Informative References 1724 [ALF] Clark, D. and D. Tennenhouse, "Architectural 1725 Considerations for a New Generation of Protocols", SIGCOMM 1726 Symposium on Communications Architectures and Protocols 1727 (Philadelphia, Pennsylvania), pp. 200--208, IEEE Computer 1728 Communications Review, Vol. 20(4), September 1990. 1730 [I-D.alvestrand-rtp-sess-neutral] 1731 Alvestrand, H., "Why RTP Sessions Should Be Content 1732 Neutral", draft-alvestrand-rtp-sess-neutral-01 (work in 1733 progress), June 2012. 1735 [I-D.ietf-avt-srtp-ekt] 1736 Wing, D., McGrew, D., and K. Fischer, "Encrypted Key 1737 Transport for Secure RTP", draft-ietf-avt-srtp-ekt-03 1738 (work in progress), October 2011. 1740 [I-D.ietf-avtcore-6222bis] 1741 Begen, A., Perkins, C., Wing, D., and E. Rescorla, 1742 "Guidelines for Choosing RTP Control Protocol (RTCP) 1743 Canonical Names (CNAMEs)", draft-ietf-avtcore-6222bis-06 1744 (work in progress), July 2013. 1746 [I-D.ietf-avtcore-multi-media-rtp-session] 1747 Westerlund, M., Perkins, C., and J. Lennox, "Sending 1748 Multiple Types of Media in a Single RTP Session", draft- 1749 ietf-avtcore-multi-media-rtp-session-03 (work in 1750 progress), July 2013. 1752 [I-D.ietf-avtcore-rtp-security-options] 1753 Westerlund, M. and C. Perkins, "Options for Securing RTP 1754 Sessions", draft-ietf-avtcore-rtp-security-options-09 1755 (work in progress), November 2013. 1757 [I-D.ietf-avtext-multiple-clock-rates] 1758 Petit-Huguenin, M. and G. Zorn, "Support for Multiple 1759 Clock Rates in an RTP Session", draft-ietf-avtext- 1760 multiple-clock-rates-10 (work in progress), September 1761 2013. 1763 [I-D.ietf-mmusic-msid] 1764 Alvestrand, H., "Cross Session Stream Identification in 1765 the Session Description Protocol", draft-ietf-mmusic- 1766 msid-02 (work in progress), November 2013. 1768 [I-D.ietf-mmusic-sdp-bundle-negotiation] 1769 Holmberg, C., Alvestrand, H., and C. Jennings, 1770 "Multiplexing Negotiation Using Session Description 1771 Protocol (SDP) Port Numbers", draft-ietf-mmusic-sdp- 1772 bundle-negotiation-05 (work in progress), October 2013. 1774 [I-D.ietf-payload-rtp-howto] 1775 Westerlund, M., "How to Write an RTP Payload Format", 1776 draft-ietf-payload-rtp-howto-09 (work in progress), 1777 October 2013. 1779 [I-D.lennox-avtcore-rtp-multi-stream] 1780 Lennox, J., Westerlund, M., Wu, W., and C. Perkins, "RTP 1781 Considerations for Endpoints Sending Multiple Media 1782 Streams", draft-lennox-avtcore-rtp-multi-stream-02 (work 1783 in progress), February 2013. 1785 [I-D.lennox-mmusic-sdp-source-selection] 1786 Lennox, J. and H. Schulzrinne, "Mechanisms for Media 1787 Source Selection in the Session Description Protocol 1788 (SDP)", draft-lennox-mmusic-sdp-source-selection-05 (work 1789 in progress), October 2012. 1791 [I-D.westerlund-avtcore-max-ssrc] 1792 Westerlund, M., Burman, B., and F. Jansson, "Multiple 1793 Synchronization sources (SSRC) in RTP Session Signaling", 1794 draft-westerlund-avtcore-max-ssrc-02 (work in progress), 1795 July 2012. 1797 [I-D.westerlund-avtcore-rtp-topologies-update] 1798 Westerlund, M. and S. Wenger, "RTP Topologies", draft- 1799 westerlund-avtcore-rtp-topologies-update-02 (work in 1800 progress), February 2013. 1802 [I-D.westerlund-avtcore-transport-multiplexing] 1803 Westerlund, M. and C. Perkins, "Multiplexing Multiple RTP 1804 Sessions onto a Single Lower-Layer Transport", draft- 1805 westerlund-avtcore-transport-multiplexing-07 (work in 1806 progress), October 2013. 1808 [I-D.westerlund-avtext-rtcp-sdes-srcname] 1809 Westerlund, M., "RTCP Source Description Item SRCNAME to 1810 Label Individual Media Sources", draft-westerlund-avtext- 1811 rtcp-sdes-srcname-03 (work in progress), October 2013. 1813 [RFC2198] Perkins, C., Kouvelas, I., Hodson, O., Hardman, V., 1814 Handley, M., Bolot, J., Vega-Garcia, A., and S. Fosse- 1815 Parisis, "RTP Payload for Redundant Audio Data", RFC 2198, 1816 September 1997. 1818 [RFC2205] Braden, B., Zhang, L., Berson, S., Herzog, S., and S. 1819 Jamin, "Resource ReSerVation Protocol (RSVP) -- Version 1 1820 Functional Specification", RFC 2205, September 1997. 1822 [RFC2326] Schulzrinne, H., Rao, A., and R. Lanphier, "Real Time 1823 Streaming Protocol (RTSP)", RFC 2326, April 1998. 1825 [RFC2474] Nichols, K., Blake, S., Baker, F., and D.L. Black, 1826 "Definition of the Differentiated Services Field (DS 1827 Field) in the IPv4 and IPv6 Headers", RFC 2474, December 1828 1998. 1830 [RFC2974] Handley, M., Perkins, C., and E. Whelan, "Session 1831 Announcement Protocol", RFC 2974, October 2000. 1833 [RFC3261] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, 1834 A., Peterson, J., Sparks, R., Handley, M., and E. 1835 Schooler, "SIP: Session Initiation Protocol", RFC 3261, 1836 June 2002. 1838 [RFC3264] Rosenberg, J. and H. Schulzrinne, "An Offer/Answer Model 1839 with Session Description Protocol (SDP)", RFC 3264, June 1840 2002. 1842 [RFC3389] Zopf, R., "Real-time Transport Protocol (RTP) Payload for 1843 Comfort Noise (CN)", RFC 3389, September 2002. 1845 [RFC3551] Schulzrinne, H. and S. Casner, "RTP Profile for Audio and 1846 Video Conferences with Minimal Control", STD 65, RFC 3551, 1847 July 2003. 1849 [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and K. 1850 Norrman, "The Secure Real-time Transport Protocol (SRTP)", 1851 RFC 3711, March 2004. 1853 [RFC3830] Arkko, J., Carrara, E., Lindholm, F., Naslund, M., and K. 1854 Norrman, "MIKEY: Multimedia Internet KEYing", RFC 3830, 1855 August 2004. 1857 [RFC4103] Hellstrom, G. and P. Jones, "RTP Payload for Text 1858 Conversation", RFC 4103, June 2005. 1860 [RFC4566] Handley, M., Jacobson, V., and C. Perkins, "SDP: Session 1861 Description Protocol", RFC 4566, July 2006. 1863 [RFC4568] Andreasen, F., Baugher, M., and D. Wing, "Session 1864 Description Protocol (SDP) Security Descriptions for Media 1865 Streams", RFC 4568, July 2006. 1867 [RFC4588] Rey, J., Leon, D., Miyazaki, A., Varsa, V., and R. 1868 Hakenberg, "RTP Retransmission Payload Format", RFC 4588, 1869 July 2006. 1871 [RFC5104] Wenger, S., Chandra, U., Westerlund, M., and B. Burman, 1872 "Codec Control Messages in the RTP Audio-Visual Profile 1873 with Feedback (AVPF)", RFC 5104, February 2008. 1875 [RFC5109] Li, A., "RTP Payload Format for Generic Forward Error 1876 Correction", RFC 5109, December 2007. 1878 [RFC5117] Westerlund, M. and S. Wenger, "RTP Topologies", RFC 5117, 1879 January 2008. 1881 [RFC5576] Lennox, J., Ott, J., and T. Schierl, "Source-Specific 1882 Media Attributes in the Session Description Protocol 1883 (SDP)", RFC 5576, June 2009. 1885 [RFC5583] Schierl, T. and S. Wenger, "Signaling Media Decoding 1886 Dependency in the Session Description Protocol (SDP)", RFC 1887 5583, July 2009. 1889 [RFC5761] Perkins, C. and M. Westerlund, "Multiplexing RTP Data and 1890 Control Packets on a Single Port", RFC 5761, April 2010. 1892 [RFC5764] McGrew, D. and E. Rescorla, "Datagram Transport Layer 1893 Security (DTLS) Extension to Establish Keys for the Secure 1894 Real-time Transport Protocol (SRTP)", RFC 5764, May 2010. 1896 [RFC5888] Camarillo, G. and H. Schulzrinne, "The Session Description 1897 Protocol (SDP) Grouping Framework", RFC 5888, June 2010. 1899 [RFC6190] Wenger, S., Wang, Y.-K., Schierl, T., and A. 1900 Eleftheriadis, "RTP Payload Format for Scalable Video 1901 Coding", RFC 6190, May 2011. 1903 [RFC6222] Begen, A., Perkins, C., and D. Wing, "Guidelines for 1904 Choosing RTP Control Protocol (RTCP) Canonical Names 1905 (CNAMEs)", RFC 6222, April 2011. 1907 [RFC6285] Ver Steeg, B., Begen, A., Van Caenegem, T., and Z. Vax, 1908 "Unicast-Based Rapid Acquisition of Multicast RTP 1909 Sessions", RFC 6285, June 2011. 1911 [RFC6465] Ivov, E., Marocco, E., and J. Lennox, "A Real-time 1912 Transport Protocol (RTP) Header Extension for Mixer-to- 1913 Client Audio Level Indication", RFC 6465, December 2011. 1915 Appendix A. Dismissing Payload Type Multiplexing 1917 This section documents a number of reasons why using the payload type 1918 as a multiplexing point for most things related to multiple streams 1919 is unsuitable. If one attempts to use Payload type multiplexing 1920 beyond it's defined usage, that has well known negative effects on 1921 RTP. To use Payload type as the single discriminator for multiple 1922 streams implies that all the different media streams are being sent 1923 with the same SSRC, thus using the same timestamp and sequence number 1924 space. This has many effects: 1926 1. Putting restraint on RTP timestamp rate for the multiplexed 1927 media. For example, media streams that use different RTP 1928 timestamp rates cannot be combined, as the timestamp values need 1929 to be consistent across all multiplexed media frames. Thus 1930 streams are forced to use the same rate. When this is not 1931 possible, Payload Type multiplexing cannot be used. 1933 2. Many RTP payload formats can fragment a media object over 1934 multiple packets, like parts of a video frame. These payload 1935 formats need to determine the order of the fragments to 1936 correctly decode them. Thus it is important to ensure that all 1937 fragments related to a frame or a similar media object are 1938 transmitted in sequence and without interruptions within the 1939 object. This can relatively simple be solved on the sender side 1940 by ensuring that the fragments of each media stream are sent in 1941 sequence. 1943 3. Some media formats require uninterrupted sequence number space 1944 between media parts. These are media formats where any missing 1945 RTP sequence number will result in decoding failure or invoking 1946 of a repair mechanism within a single media context. The text/ 1947 T140 payload format [RFC4103] is an example of such a format. 1948 These formats will need a sequence numbering abstraction 1949 function between RTP and the individual media stream before 1950 being used with Payload Type multiplexing. 1952 4. Sending multiple streams in the same sequence number space makes 1953 it impossible to determine which Payload Type and thus which 1954 stream a packet loss relates to. 1956 5. If RTP Retransmission [RFC4588] is used and there is a loss, it 1957 is possible to ask for the missing packet(s) by SSRC and 1958 sequence number, not by Payload Type. If only some of the 1959 Payload Type multiplexed streams are of interest, there is no 1960 way of telling which missing packet(s) belong to the interesting 1961 stream(s) and all lost packets need be requested, wasting 1962 bandwidth. 1964 6. The current RTCP feedback mechanisms are built around providing 1965 feedback on media streams based on stream ID (SSRC), packet 1966 (sequence numbers) and time interval (RTP Timestamps). There is 1967 almost never a field to indicate which Payload Type is reported, 1968 so sending feedback for a specific media stream is difficult 1969 without extending existing RTCP reporting. 1971 7. The current RTCP media control messages [RFC5104] specification 1972 is oriented around controlling particular media flows, i.e. 1973 requests are done addressing a particular SSRC. Such mechanisms 1974 would need to be redefined to support Payload Type multiplexing. 1976 8. The number of payload types are inherently limited. 1977 Accordingly, using Payload Type multiplexing limits the number 1978 of streams that can be multiplexed and does not scale. This 1979 limitation is exacerbated if one uses solutions like RTP and 1980 RTCP multiplexing [RFC5761] where a number of payload types are 1981 blocked due to the overlap between RTP and RTCP. 1983 9. At times, there is a need to group multiplexed streams and this 1984 is currently possible for RTP Sessions and for SSRC, but there 1985 is no defined way to group Payload Types. 1987 10. It is currently not possible to signal bandwidth requirements 1988 per media stream when using Payload Type Multiplexing. 1990 11. Most existing SDP media level attributes cannot be applied on a 1991 per Payload Type level and would require re-definition in that 1992 context. 1994 12. A legacy endpoint that doesn't understand the indication that 1995 different RTP payload types are different media streams might be 1996 slightly confused by the large amount of possibly overlapping or 1997 identically defined RTP Payload Types. 1999 Appendix B. Proposals for Future Work 2001 The above discussion and guidelines indicates that a small set of 2002 extension mechanisms could greatly improve the situation when it 2003 comes to using multiple streams independently of Multiple RTP session 2004 or Additional SSRC. These extensions are: 2006 Media Source Identification: A Media source identification that can 2007 be used to bind together media streams that are related to the 2008 same media source. A proposal 2009 [I-D.westerlund-avtext-rtcp-sdes-srcname] exist for a new SDES 2010 item SRCNAME that also can be used with the a=ssrc SDP attribute 2011 to provide signalling layer binding information. 2013 MSID: A Media Stream identification scheme that can be used to 2014 signal relationships between SSRCs that can be in the same or in 2015 different RTP sessions. Described in [I-D.ietf-mmusic-msid] 2017 SSRC limitations within RTP sessions: By providing a signalling 2018 solution that allows the signalling peers to explicitly express 2019 both support and limitations on how many simultaneous media 2020 streams an endpoint can handle within a given RTP Session. That 2021 ensures that usage of Additional SSRC occurs when supported and 2022 without overloading an endpoint. This extension is proposed in 2023 [I-D.westerlund-avtcore-max-ssrc]. 2025 Appendix C. Signalling considerations 2027 Signalling is not an architectural consideration for RTP itself, so 2028 this discussion has been moved to an appendix. However, it is hugely 2029 important for anyone building complete applications, so it is 2030 deserving of discussion. 2032 The issues raised here need to be addressed in the WGs that deal with 2033 signalling; they cannot be addressed by tweaking, extending or 2034 profiling RTP. 2036 C.1. Signalling Aspects 2038 There exist various signalling solutions for establishing RTP 2039 sessions. Many are SDP [RFC4566] based, however SDP functionality is 2040 also dependent on the signalling protocols carrying the SDP. Where 2041 RTSP [RFC2326] and SAP [RFC2974] both use SDP in a declarative 2042 fashion, while SIP [RFC3261] uses SDP with the additional definition 2043 of Offer/Answer [RFC3264]. The impact on signalling and especially 2044 SDP needs to be considered as it can greatly affect how to deploy a 2045 certain multiplexing point choice. 2047 C.1.1. Session Oriented Properties 2049 One aspect of the existing signalling is that it is focused around 2050 sessions, or at least in the case of SDP the media description. 2051 There are a number of things that are signalled on a session level/ 2052 media description but those are not necessarily strictly bound to an 2053 RTP session and could be of interest to signal specifically for a 2054 particular media stream (SSRC) within the session. The following 2055 properties have been identified as being potentially useful to signal 2056 not only on RTP session level: 2058 o Bitrate/Bandwidth exist today only at aggregate or a common any 2059 media stream limit, unless either codec-specific bandwidth 2060 limiting or RTCP signalling using TMMBR is used. 2062 o Which SSRC that will use which RTP Payload Types (this will be 2063 visible from the first media packet, but is sometimes useful to 2064 know before packet arrival). 2066 Some of these issues are clearly SDP's problem rather than RTP 2067 limitations. However, if the aim is to deploy an solution using 2068 additional SSRCs that contains several sets of media streams with 2069 different properties (encoding/packetization parameter, bit-rate, 2070 etc), putting each set in a different RTP session would directly 2071 enable negotiation of the parameters for each set. If insisting on 2072 additional SSRC only, a number of signalling extensions are needed to 2073 clarify that there are multiple sets of media streams with different 2074 properties and that they need in fact be kept different, since a 2075 single set will not satisfy the application's requirements. 2077 For some parameters, such as resolution and framerate, a SSRC-linked 2078 mechanism has been proposed: 2079 [I-D.lennox-mmusic-sdp-source-selection]. 2081 C.1.2. SDP Prevents Multiple Media Types 2082 SDP chose to use the m= line both to delineate an RTP session and to 2083 specify the top level of the MIME media type; audio, video, text, 2084 image, application. This media type is used as the top-level media 2085 type for identifying the actual payload format bound to a particular 2086 payload type using the rtpmap attribute. This binding has to be 2087 loosened in order to use SDP to describe RTP sessions containing 2088 multiple MIME top level types. 2090 There is an accepted WG item in the MMUSIC WG to define how multiple 2091 media lines describe a single underlying transport 2092 [I-D.ietf-mmusic-sdp-bundle-negotiation] and thus it becomes possible 2093 in SDP to define one RTP session with media types having different 2094 MIME top level types. 2096 C.1.3. Signalling Media Stream Usage 2098 Media streams being transported in RTP has some particular usage in 2099 an RTP application. This usage of the media stream is in many 2100 applications so far implicitly signalled. For example, an 2101 application might choose to take all incoming audio RTP streams, mix 2102 them and play them out. However, in more advanced applications that 2103 use multiple media streams there will be more than a single usage or 2104 purpose among the set of media streams being sent or received. RTP 2105 applications will need to signal this usage somehow. The signalling 2106 used will have to identify the media streams affected by their RTP- 2107 level identifiers, which means that they have to be identified either 2108 by their session or by their SSRC + session. 2110 In some applications, the receiver cannot utilise the media stream at 2111 all before it has received the signalling message describing the 2112 media stream and its usage. In other applications, there exists a 2113 default handling that is appropriate. 2115 If all media streams in an RTP session are to be treated in the same 2116 way, identifying the session is enough. If SSRCs in a session are to 2117 be treated differently, signalling needs to identify both the session 2118 and the SSRC. 2120 If this signalling affects how any RTP central node, like an RTP 2121 mixer or translator that selects, mixes or processes streams, treats 2122 the streams, the node will also need to receive the same signalling 2123 to know how to treat media streams with different usage in the right 2124 fashion. 2126 Authors' Addresses 2127 Magnus Westerlund 2128 Ericsson 2129 Farogatan 6 2130 SE-164 80 Kista 2131 Sweden 2133 Phone: +46 10 714 82 87 2134 Email: magnus.westerlund@ericsson.com 2136 Bo Burman 2137 Ericsson 2138 Farogatan 6 2139 SE-164 80 Kista 2140 Sweden 2142 Phone: +46 10 714 13 11 2143 Email: bo.burman@ericsson.com 2145 Colin Perkins 2146 University of Glasgow 2147 School of Computing Science 2148 Glasgow G12 8QQ 2149 United Kingdom 2151 Email: csp@csperkins.org 2153 Harald Tveit Alvestrand 2154 Google 2155 Kungsbron 2 2156 Stockholm 11122 2157 Sweden 2159 Email: harald@alvestrand.no