| Network Working Group | M. Westerlund |
| Internet-Draft | B. Burman |
| Intended status: Standards Track | P. Sandgren |
| Expires: April 26, 2012 | Ericsson |
| October 24, 2011 |
RTCP SDES Item SRCNAME to Label Individual Sources
draft-westerlund-avtext-rtcp-sdes-srcname-00
This document defines a new SDES item called SRCNAME which uniquely identifies a single media source, like a camera or a microphone. That way anyone receiving the SDES information from a set of interlinked RTP sessions can determine which SSRCs are related to the same source. It can equally be used to label SSRC multiplexed related streams, such as FEC or Retransmission streams related to the original source stream in the same session. In addition the new SDES item is also defined for usage with the SDP source specific media attribute ("a=ssrc") enabling an end-point to declare and learn the source bindings ahead of receiving RTP/RTCP packets through signalling.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on April 26, 2012.
Copyright (c) 2011 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
RTP has always been a protocol that supports multiple participants each sending their own media streams in RTP sessions. Previously many implementations have aimed only at point to point voice over IP with a single source in each end-point. Even client implementations aimed at video conferences have often been built with the assumption around central mixers that only deliver a single media stream per media type. However, more advanced client implementations may transmit multiple streams in the same RTP session and there may be tight relations between different streams and their SSRCs. For example, a client with several cameras that uses simulcast to send streams with different encodings of the video from each camera have the need of conveying the relation of the streams to the receiver. A similar example is a client with several cameras that uses SVC multi-session transmission [RFC6190] and also here the receiver needs to know which streams relate to which video source. Other examples of tight relations are a retransmission stream and its original stream as well as the case of forward error correction (FEC) where a client can send source streams and associated repair streams.
CNAME is not sufficient to express this relation although that is commonly inferred from end-points that have only one media stream per media type. The primary use of CNAME in multi-source usages is instead to indicate which end-point and what synchronization context a particular media stream relates to and that usually means that all streams sent from a client have the same CNAME. We are neither relying on using the same SSRC for all streams related to a particular media source as it is not robust against SSRC collision and forces potentially cascading SSRC changes between sessions. Also, using the same SSRC is not possible when SSRC-multiplexing is used.
A common solution to convey the relation between streams is to use SDP attributes. Session-multiplexed streams can be associated with an attribute that groups different RTP sessions and SSRC-multiplexed streams can be grouped at the media level for each RTP session. For example, in Forward Error Correction Grouping Semantics in the Session Description Protocol [RFC5956] an SDP media level attribute called "ssrc-group:FEC-FR" is used for grouping FEC associations when the different streams from a source are SSRC-multiplexed in the same RTP session. Using SDP attributes may work fine in the case when the receivers of the streams also get an SDP describing the bindings of all the streams, but that is not always the case. One such example is a conference session where clients are communicating with each other via an RTP Translator. The RTP Translator forwards all RTP and RTCP traffic from a client to all other clients and the clients can be prepared to receive any number of streams of certain specified media. When a new client joins the session the other clients may not be notified with a SIP Update including a new SDP, instead the clients will detect the new client's streams via RTP and RTCP. In this case there is no way for a client to identify if certain streams are related to each other since that information only was included in the SDP.
RTP Retransmission Payload Format [RFC4588] describes a solution for finding the association between original streams and retransmission streams when SSRC-multiplexing is used. The association can be resolved when the receiver receives a retransmission packet matching a retransmission request sent earlier. However, the RFC continues with describing that this mechanism might fail if there are two outstanding requests for the same packet sequence number in two different original streams of a session. Therefore, to avoid ambiguity in unicast a receiver MUST NOT have two outstanding requests for the same packet sequence number in two different original streams before the association is resolved. For multicast, however, this ambiguity cannot be avoided and SSRC-multiplexing of original and retransmission streams is therefore prohibited in multicast. By defining a solution for one to one mapping between an original stream and any supporting streams this issue can be avoided in the future. Note: This document does not update RFC 4588 to use this solution, but it may be done in the future.
To enable an RTP session participant to determine the close relation of different streams without the above mentioned problems, a new method for identifying such sources is needed. RTP [RFC3550] defines the Source Description RTCP Packet (SDES), which contains one or more chunks, each of which is composed of SDES items describing the SSRC identified in that chunk. None of the present SDES items is, however, suitable for uniquely identifying a media source.
Therefore we propose that one defines a new SDES item called the SRCNAME which with a unique label identifies a single media source, like a camera or a microphone. The source may also be a particular media mix or conceptual stream, such as the "most active speaker" output by a RTP mixer performing stream switching. That way anyone receiving the SDES information from a set of interlinked RTP sessions or multiple SSRCs in the same session can determine which SSRCs are the same source. Connecting streams with SRCNAME can be done irrespective of which multiplexing type is used and it solves the problems with the current solutions described above.
It is, however, possible that a receiver will receive the RTP streams before receiving SDES packets with all SRCNAME items and that would mean that the receiver cannot make the connections between SSRCs and SRCNAMEs when starting to receive the media. "Source-Specific Media Attributes in the Session Description Protocol (SDP)" [RFC5576] defines a way of declaring different attributes for SSRCs in each session in SDP and if a new source attribute is added to this framework it would be suitable for conveying the connections between SSRCs and SRCNAMEs before the media communication starts. Thus, in addition to the new SDES item we also define a new SDP source-specific media attribute called srcname, which enables an end-point to declare and learn the source bindings ahead of receiving RTP/RTCP packets. Of course, this new SDP source attribute will not be useful for the case described above when clients did not get updates with new client's stream bindings, but it will be useful in most other cases.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119].
Source Descriptions are a method that should work with all RTP topologies (assuming that any intermediary node is supporting this item) and existing RTP extensions. Thus we propose that one defines a new SDES item called the SRCNAME which with a unique identifier identifies a single media source, like a camera, a microphone, a particular media mix, or conceptual stream. That way anyone receiving the SDES information from a set of interlinked RTP sessions or SSRCs in a single session can determine which SSRCs are related to the same source.
The SRCNAME is RECOMMENDED to be per communication session unique random identifiers generated according to "Guidelines for Choosing RTP Control Protocol (RTCP) Canonical Names (CNAMEs)" [RFC6222] with the addition that a local counter enumerating the sources on the host also is concatenated to the key in step 4 prior to calculating the hash. The SRCNAME included in an RTCP packet MUST fulfill the requirements Section 6.5 in RTP [RFC3550] puts on SDES item values in general. These requirements is that it is a UTF-8 [RFC3629] string that have a maximum length of 255 bytes.
This SRCNAME's relation to CNAME is the following. CNAME represents an end-point and a synchronization context. If the different sources identified by SRCNAMEs should be played out synchronized when receiving them in a multi-stream context, then the sources need to be in the same synchronization context. Thus in all cases, all SSRCs with the same SRCNAME will have the same CNAME. A given CNAME may contain multiple sets of sources using different SRCNAMEs.
The SDES SRCNAME item follows the same format as the other SDES items defined in RTP [RFC3550]:
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | SRCNAME=TBA1 | length | source name ... +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
When using the SRCNAME SDES item it is equally important to CNAME, thus it is RECOMMENDED to be included in all full compound RTCP packets being sent. It MAY also be included in non-compound packets in cases where the implementation believes that there might be new receivers needing the information.
"Source-Specific Media Attributes in the Session Description Protocol (SDP)" [RFC5576] defines a way of declaring attributes for SSRC in each session in SDP. With a new SDES item, one can use this framework to define how also the SRCNAME can be provided in the SDP for each SSRC in each RTP session, thus enabling an end-point to declare and learn the source bindings ahead of receiving RTP/RTCP packets.
Hence, we propose a new SDP source attribute called srcname with the following structure:
a=ssrc:<ssrc-id> srcname:<srcname>
The srcname value MUST be identical to the SRCNAME value the media sender will send in the SDES SRCNAME item in the SDES RTCP packets.
FormalABNF syntax [RFC5234] for the "srcname" attribute:
srcname-attr = "srcname:" srcname ssrcname = byte-string ; The definition of "byte-string" is in RFC 4566. attribute =/ srcname-attr ; The definition of "attribute" is in RFC 4566.
This section shows SDP examples of declaring the SRCNAME in SDP. Only the relevant parts of the SDP are shown to improve readability. Please note that the below examples are all hypothetical as no decision has yet been to use the SRCNAME mechanism with the respective example.
In this use case the end-point is a client with a single audio source and two video sources and it uses simulcast for sending different encodings of the same video source. This example is based on Using Simulcast in RTP sessions [I-D.westerlund-avtcore-rtp-simulcast]. The following SDP describes this.
s=Simulcast enabled client m=audio 49200 RTP/AVP 96 a=rtpmap:96 G719/48000/2 a=ssrc:521923924 cname:alice@foo.example.com a=ssrc:521923924 srcname:2b:45:c7:12:83:e6 a=mid:1 m=video 49300 RTP/AVP 96 a=rtpmap:96 H264/90000 a=fmtp:96 profile-level-id=42c01e a=imageattr:* send [x=640,y=360] recv [x=640,y=360] [x=320,y=180] a=ssrc:192392452 cname:alice@foo.example.com a=ssrc:192392452 srcname:a3:d3:4b:f1:22:12 a=ssrc:834753488 cname:alice@foo.example.com a=ssrc:834753488 srcname:7a:39:a9:3e:28:f7 a=mid:2 a=content:main m=video 49400 RTP/AVP 96 a=rtpmap:96 H264/90000 a=fmtp:96 profile-level-id=42c00d a=imageattr:96 send [x=320,y=180] a=ssrc:239245219 cname:alice@foo.example.com a=ssrc:239245219 srcname:a3:d3:4b:f1:22:12 a=ssrc:734623563 cname:alice@foo.example.com a=ssrc:734623563 srcname:7a:39:a9:3e:28:f7 a=mid:3 a=sendonly
The audio session is proposing to use one stereo stream of G.719 and the video sessions are proposing to send two different encodings of each video source, one with the resolution 640x360 and one with 320x180. The end-point also declares the SSRCs it intends to use with bindings to CNAME and SRCNAME, enabling the receiver of the SDP to bind together the video streams that originates from the same video camera.
The use of the srcname attribute in the SDP is optional and the information can be retrieved from RTCP reporting, but it will then not be possible to correctly relate the video sources until the first RTCP report is received.
Here an example is shown of a client that uses SVC with multi-session transmission as described in RTP Payload Format for Scalable Video Coding [RFC6190]. RTP Payload Format for Scalable Video Coding [RFC6190] only describes examples for a client with one video source and the decoder dependencies of the different sessions are grouped using the Session grouping DDP attribute as defined in Signaling Media Decoding Dependency in the Session Description Protocol (SDP) [RFC5583] and implicitly CNAME.
However, if a client has two video sources and wishes to use multi-session transmission and send streams from both sources in each session an additional grouping mechanism is needed to group the different streams in the different sessions. SRCNAME is suitable for this and here we show an example where the DDP attribute groups the different sessions and the SRCNAME is used to relate the different SSRCs in each RTP session to one of the two video sources.
s=SVC MST client
a=group:DDP L1 L2 L3
m=video 20000 RTP/AVP 96
a=rtpmap:96 H264/90000
a=fmtp:96 profile-level-id=4de00a; packetization-mode=1;
mst-mode=NI-TC; sprop-parameter-sets={sps0},{pps0};
a=ssrc:743947584 cname:bob@foo.example.com
a=ssrc:743947584 srcname:7e:83:c1:82:e8:a6
a=ssrc:283894947 cname:bob@foo.example.com
a=ssrc:283894947 srcname:b3:8d:f1:18:c5:84
a=mid:L1
m=video 20002 RTP/AVP 97
a=rtpmap:97 H264-SVC/90000
a=fmtp:97 profile-level-id=53000c; packetization-mode=1;
mst-mode=NI-T; sprop-parameter-sets={sps1},{pps1};
a=ssrc:492784823 cname:bob@foo.example.com
a=ssrc:492784823 srcname:7e:83:c1:82:e8:a6
a=ssrc:892362397 cname:bob@foo.example.com
a=ssrc:892362397 srcname:b3:8d:f1:18:c5:84
a=mid:L2
a=depend:97 lay L1:96
m=video 20004 RTP/AVP 98
a=rtpmap:98 H264-SVC/90000
a=fmtp:98 profile-level-id=53001F; packetization-mode=1;
mst-mode=NI-T; sprop-parameter-sets={sps2},{pps2};
a=ssrc:184562894 cname:bob@foo.example.com
a=ssrc:184562894 srcname:7e:83:c1:82:e8:a6
a=ssrc:305605682 cname:bob@foo.example.com
a=ssrc:305605682 srcname:b3:8d:f1:18:c5:84
a=mid:L3
a=depend:98 lay L1:96 L2:97
Thus, the client declares that it will send two video streams in each RTP session and the receiver is then able to relate the streams in the different sessions by using the SRCNAME binding. Without the SRCNAME binding it would not be possible for the receiver to know which streams belong to the same source.
This use case shows how SRCNAME can be used to connect retransmission streams to the original streams in the case of SSRC multiplexed RTP retransmission [RFC4588]. This is included to exemplify how RTP retransmission could be updated to provide explicit bindings between the source and the repair stream, but just an example and not a specification.
s=SSRC-multiplexed retransmission client m=audio 49200 RTP/AVP 96 a=rtpmap:96 G719/48000/2 a=ssrc:521923924 cname:carol@foo.example.com a=ssrc:521923924 srcname:88:3a:93:c1:3f:71 a=mid:1 m=video 49300 RTP/AVP 96 97 a=rtpmap:96 H264/90000 a=rtcp-fb:96 nack a=fmtp:96 profile-level-id=42c01e a=rtpmap:97 rtx/90000 a=fmtp:97 apt=96;rtx-time=200 a=ssrc:192392452 cname:carol@foo.example.com a=ssrc:192392452 srcname:7b:6e:23:8b:31:a8 a=ssrc:834753488 cname:carol@foo.example.com a=ssrc:834753488 srcname:7b:6e:23:8b:31:a8 a=ssrc:682394013 cname:carol@foo.example.com a=ssrc:682394013 srcname:c4:98:d9:1a:fc:58 a=ssrc:284576129 cname:carol@foo.example.com a=ssrc:284576129 srcname:c4:98:d9:1a:fc:58 a=mid:2
The client proposes to send two original video streams in the video session and a retransmission stream for each one of them. The retransmission streams are associated with the respective original stream by using the same SRCNAME and a receiver would then know which original stream a certain retransmission stream is associated with. This solves the ambiguity problem when SSRC-multiplexing is used for retransmission and it enables SSRC-multiplexing of original and retransmission streams to be used also in multicast sessions.
Forward Error Correction Grouping Semantics in the Session Description Protocol [RFC5956] defines two SDP attributes for grouping the associated source and FEC-based repair streams. One can be used for grouping different RTP sessions and the other can be used for grouping SSRCs in the same RTP session, i.e. when session- respective SSRC-multiplexing is used. However, it may be advantageous to SSRC-multiplex the source streams in one RTP session and the repair streams in another since that gives a receiver the possibility to reject the repair session in case it does not support the proposed FEC. In this case the above mentioned grouping attributes cannot be used to associate the repair streams with the respective source stream since grouping of SSRCs cannot be made across RTP sessions. The following example shows how SRCNAME can be used for that.
s=FEC client a=group:FEC-FR 1 2 m=video 49200 RTP/AVP 100 a=rtpmap:100 MP2T/90000 a=ssrc:847612849 cname:dave@foo.example.com a=ssrc:847612849 srcname:45:a8:f4:19:b4:c3 a=ssrc:558237845 cname:dave@foo.example.com a=ssrc:558237845 srcname:b8:58:29:c7:2f:9e a=mid:1 m=application 49300 RTP/AVP 101 a=rtpmap:101 1d-interleaved-parityfec/90000 a=fmtp:101 L=5; D=10; repair-window=200000 a=ssrc:389572053 cname:dave@foo.example.com a=ssrc:389572053 srcname:45:a8:f4:19:b4:c3 a=ssrc:185729479 cname:dave@foo.example.com a=ssrc:185729479 srcname:b8:58:29:c7:2f:9e a=mid:2
In this example the client proposes to send two video streams in one session and two repair streams in the other session. The repair streams are associated with the respective video stream by using the same SRCNAME. When receiving either this SDP or the SDES SRCNAME packets a receiver can make the connection between the source streams and the repair streams. Even a client not receiving the SDP will be able to do the association if it has established one RTP session for receiving source streams and another for receiving repair streams.
The SDP offer/answer procedures for the a=ssrc is specified in Source-Specific Media Attributes in the Session Description Protocol (SDP) [RFC5576].
Clients not supporting SRCNAME will not have the possibility to bind different streams to a specific media source, since they will not understand the SRCNAME SDES item. However, sending SRCNAME SDES items to a client not supporting it should not impose any problems since all clients should be prepared that new SDES items may be specified according to RTP [RFC3550].
According to the definition of SDP attributes in SDP: Session Description Protocol [RFC4566], if an attribute is received that is not understood, it MUST be ignored by the receiver. So a receiver not supporting the ssrc attribute will simply ignore it.
Source-Specific Media Attributes in the Session Description Protocol (SDP) [RFC5576] defines rules of how new source attributes should be registered, which means that a receiver supporting RFC5576 should be prepared that new source attributes may be defined. This means that a user supporting some of the source attributes should not have any problems when the user receives an SDP with unknown source attributes.
Following the guidelines in SDP [RFC4566], in The Session Description Protocol (SDP) Grouping Framework [RFC5888], and in RTP [RFC3550], the IANA is requested to register:
The SDES SRCNAMEs being opaque identifiers could potentially carry additional meanings or function as overt channel. If the SRCNAME would be permanent between sessions, they have the potential for compromising the users’ privacy as they can be tracked between sessions. See Guidelines for Choosing RTP Control Protocol (RTCP) Canonical Names (CNAMEs) [RFC6222] for more discussion.
A third party modification of the srcname labels either in the RTCP SDES items or in the SDP a=ssrc attribute can cause service disruption. By modifying labels the wrong streams could be associated, with potentially serious effects including media disruptions. If streams that are to be associated aren't associated, then another type of failures occur. To prevent modification, insertion or deletion of the srcname labels the carrying channel needs to be protected by integrity protection and source authentication. For RTCP various solutions exist, such as SRTP [RFC3711], DTLS [RFC4347], IPsec [RFC4301]. For protecting the SDP the signalling channel needs to provide protection. For SIP S/MIME [RFC3261] are the ideal, and hop by hopTLS [RFC5246] provides at least some protection, although not perfect. For SDP's retrieved using RTSP DESCRIBE [RFC2326] TLS would be the RECOMMENDED solution.
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. |
| [RFC3629] | Yergeau, F., "UTF-8, a transformation format of ISO 10646", STD 63, RFC 3629, November 2003. |
| [RFC3550] | Schulzrinne, H., Casner, S., Frederick, R. and V. Jacobson, "RTP: A Transport Protocol for Real-Time Applications", STD 64, RFC 3550, July 2003. |
| [RFC5234] | Crocker, D. and P. Overell, "Augmented BNF for Syntax Specifications: ABNF", STD 68, RFC 5234, January 2008. |
| [RFC5576] | Lennox, J., Ott, J. and T. Schierl, "Source-Specific Media Attributes in the Session Description Protocol (SDP)", RFC 5576, June 2009. |
| [RFC6222] | Begen, A., Perkins, C. and D. Wing, "Guidelines for Choosing RTP Control Protocol (RTCP) Canonical Names (CNAMEs)", RFC 6222, April 2011. |