| Network Working Group | P. Saint-Andre |
| Internet-Draft | Cisco Systems, Inc. |
| Intended status: Standards Track | S. Ibarra |
| Expires: June 15, 2014 | AG Projects |
| E. Ivov | |
| Jitsi | |
| December 12, 2013 |
Interworking between the Session Initiation Protocol (SIP) and the Extensible Messaging and Presence Protocol (XMPP): Media Sessions.
draft-ietf-stox-media-02
This document defines a bi-directional protocol mapping for use by gateways that enable the exchange of media signalling messages between systems that implement the Jingle extensions to the Extensible Messaging and Presence Protocol (XMPP) and those that implement the Session Initiation Protocol (SIP).
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 June 15, 2014.
Copyright (c) 2013 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.
The Session Initiation Protocol [RFC3261] is a widely-deployed technology for the management of media sessions (such as voice and video calls) over the Internet. SIP itself provides a signalling channel (sometimes via the User Datagram Protocol [RFC0768]), over which two or more parties can exchange messages for the purpose of negotiating a media session that uses a dedicated media channel such as the Real-time Transport Protocol [RFC3550].
The Extensible Messaging and Presence Protocol (XMPP) [RFC6120] also provides a signalling channel, typically via the Transmission Control Protocol [RFC0793]. Given the significant differences between XMPP and SIP, it is difficult to combine the two technologies in a single user agent. Therefore, developers wishing to add media session capabilities to XMPP clients have defined an XMPP-specific negotiation protocol called Jingle [XEP-0166].
However, Jingle was designed to easily map to SIP for communication through gateways or other transformation mechanisms. Therefore, consistent with existing specifications for mapping between SIP and XMPP (see [I-D.ietf-stox-core] and other related specifications), this document describes a bidirectional protocol mapping for use by gateways that enable the exchange of media signalling messages between systems that implement SIP and those that implement the XMPP Jingle extensions.
It is important to note that SIP and Jingle sessions can be gateway-ed in a rather simple fashion if all media was always routed and potentially even transcoded in through a gateway. This specification aims to define a mapping that goes beyond the above and allows gateways to (wherever possible) only intervene at the signalling level, letting user agents exchange media in an end-to-end manner. Such gateways would likely focus on handling handling RTP session establishment and control within the context of what users would perceive as "calls". This document is hence primarily dealing with calling scenarios as opposed to generic media sessions with SIP.
The discussion venue for this document is the mailing list of the STOX WG; visit https://www.ietf.org/mailman/listinfo/stox for subscription information and discussion archives.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
A number of technical terms used here are defined in [RFC3261], [RFC6120], [XEP-0166], and [XEP-0167]. The term "JID" is short for "Jabber Identifier".
Even if Jingle semantics have many similarities with those used in SIP, there are some use cases that cannot be handled in exactly the same way due to the Offer/Answer model used in SIP in conjunction with SDP.
More specifically, mapping SIP and SDP Offer/Answer to XMPP is often complicated due to the difference in how each handles backward compatibility. Jingle, as most other XMPP extensions, relies heavily on the protocol's advanced service discovery [XEP-0030] mechanisms. In other words, XMPP entities are able to verify the capabilities of their intended peer before actually attempting to establish a session with it.
SDP Offer/Answer on the other hand uses a least common denominator approach where every SDP offer has to be understandable by legacy endpoints. Newer, unsupported aspects in this offer can therefore only appear as optional or their use be limited to subsequent Offer/Answer exchanges, once their support has been confirmed.
Use of "trickle ICE" (see [I-D.ietf-mmusic-trickle-ice] and [I-D.ivov-mmusic-trickle-ice-sip]) is one example where the issue occurs. SIP endpoints need to always behave as vanilla ICE agents when sending their first offer and make sure they gather all candidates before sending a SIP INVITE. This is necessary because otherwise ICE agents with no support for trickle can prematurely declare failure. Jingle endpoints, on the other hand can verify support for trickle ICE prior to engaging in a session and adapt their behaviour accordingly.
In order to work around such issues, [XEP-0176] defines an Offer/Answer support mode through the "urn:ietf:rfc:3264" feature tag. It indicates that a specific XMPP entity can only be contacted through the use of Offer/Answer semantics. Implementations conforming to this specification MUST support Offer/Answer model with Jingle. Note that such endpoints are not required to actually declare support for this tag because this would mean that they too would only be reachable through Offer/Answer semantics.
As mentioned, Jingle was designed in part to enable straightforward protocol mapping between XMPP and SIP. However, given the significantly different technology assumptions underlying XMPP and SIP, Jingle is naturally different from SIP in several important respects:
Jingle is designed in a modular fashion, so that session description data is generally carried in a payload within the generic Jingle elements, i.e., the <jingle/> element and its <content/> child. The following example illustrates this structure, where the XMPP stanza is a request to initiate an audio session using RTP over a raw UDP transport.
<iq from='romeo@example.net/v3rsch1kk3l1jk'
id='ne91v36s'
to='juliet@example.com/t3hr0zny'
type='set'>
<jingle xmlns='urn:xmpp:jingle:1'
action='session-initiate'
initiator='romeo@example.net/v3rsch1kk3l1jk'
sid='a73sjjvkla37jfea'>
<content creator='initiator'
media='audio'
name='this-is-the-audio-content'
senders='both'>
<description xmlns='urn:xmpp:jingle:app:rtp:1'>
<payload-type id='96' name='speex' clockrate='16000'/>
<payload-type id='97' name='speex' clockrate='8000'/>
<payload-type id='18' name='G729'/>
<payload-type channels='2'
clockrate='16000'
id='103'
name='L16'/>
<payload-type id='98' name='x-ISAC' clockrate='8000'/>
</description>
<transport xmlns='urn:xmpp:jingle:transport:raw-udp'>
<candidate ip='10.1.1.104' port='13540' generation='0'/>
</transport>
</content>
</jingle>
</iq>
In the foregoing example, the syntax and semantics of the <jingle/> and <content/> elements are defined in [XEP-0166], the syntax and semantics of the <description/> element are defined in [XEP-0167], and the syntax and semantics of the <transport/> element are defined in [XEP-0177]. Other <description/> elements are defined in specifications for the appropriate application types (see for example [XEP-0167]) and other <transport/> elements are defined in the specifications for appropriate transport methods (see for example [XEP-0176], which defines an XMPP profile of [RFC5245]).
At the core Jingle layer, the following mappings are defined.
+--------------------------------+--------------------------------+
| Jingle | SIP |
+--------------------------------+--------------------------------+
| <jingle/> 'action' | [ see next table ] |
+--------------------------------+--------------------------------+
| <jingle/> 'initiator' | [ no mapping ] |
+--------------------------------+--------------------------------+
| <jingle/> 'responder' | [ no mapping ] |
+--------------------------------+--------------------------------+
| <jingle/> 'sid' | local-part of Dialog ID |
+--------------------------------+--------------------------------+
| local-part of 'initiator' | <username> in SDP o= line |
+--------------------------------+--------------------------------+
| <content/> 'creator' | [ no mapping ] |
+--------------------------------+--------------------------------+
| <content/> 'name' | [ no mapping ] |
+--------------------------------+--------------------------------+
| <content/> 'profile' | <proto> in SDP m= line |
+--------------------------------+--------------------------------+
| <content/> 'senders' value of | a= line of sendrecv, recvonly, |
| both, initiator, or responder | or sendonly |
+--------------------------------+--------------------------------+
The 'senders' attribute is optional in Jingle, thus in case it's absent it's RECOMMENDED that the direction value is considered as 'sendrecv'.
The 'action' attribute of the <jingle/> element has nine allowable values. In general they should be mapped as shown in the following table, with some exceptions as described herein.
+-------------------+-----------------+
| Jingle Action | SIP Method |
+-------------------+-----------------+
| content-accept | INVITE response |
| | (1xx or 2xx) |
+-------------------+-----------------+
| content-add | INVITE request |
+-------------------+-----------------+
| content-modify | INVITE request |
+-------------------+-----------------+
| content-remove | INVITE request |
+-------------------+-----------------+
| session-accept | INVITE response |
| | (1xx or 2xx) |
+-------------------+-----------------+
| session-info | [varies] |
+-------------------+-----------------+
| session-initiate | INVITE request |
+-------------------+-----------------+
| session-terminate | BYE |
+-------------------+-----------------+
| transport-info | unnused |
+-------------------+-----------------+
Jingle application formats for audio and video exchange via RTP are specified in [XEP-0167]. These application formats effectively maps to the "RTP/AVP" profile specified in [RFC3551] and the "RTP/SAVP" profile specified in [RFC3711], where the media types are "audio" and "video", and the specific mappings to SDP syntax are provided in [XEP-0167].
As stated in [XEP-0167] future versions of this specification might define how to use other RTP profiles such as "RTP/AVPF" and "RTP/SAVPF" as defined in RFC4585 and RFC5124 respectively.
A basic Jingle transport method for exchanging media over UDP is specified in [XEP-0177]. This transport method involves the negotiation of an IP address and port only. It does not provide NAT traversal, effectively leaving the task to intermediary entries. The Jingle 'ip' attribute maps to the connection-address parameter of the SDP c= line and the 'port' attribute maps to the port parameter of the SDP m= line. Use of SIP without ICE would generally map to use of Raw UDP on the XMPP side of a session.
A more advanced Jingle transport method for exchanging media over UDP is specified in [XEP-0176]. Under ideal conditions this transport method provides NAT traversal by following the Interactive Connectivity Exchange methodology specified in [RFC5245].
The relevant SDP mappings are provided in [XEP-0176], however there are a few syntax incompatibilities which need to be addressed by gateways conforming to this specification:
[RFC3264] stipulates that streams are placed on hold by setting their direction to "sendonly". A session is placed on hold by doing this for all the streams it contains. The same semantics are also supported by Jingle through the "senders" element and its "initiator" and "responder" values.
[example to follow]
In addition to these semantics however Jingle also defines a more concise way for achieving the same, which consists in sending a "hold" command within a "session-info" action:
<iq from='juliet@capulet.lit/balcony'
id='xv39z423'
to='romeo@montague.lit/orchard'
type='set'>
<jingle xmlns='urn:xmpp:jingle:1'
action='session-info'
initiator='romeo@montague.lit/orchard'
sid='a73sjjvkla37jfea'>
<hold xmlns='urn:xmpp:jingle:apps:rtp:info:1'/>
</jingle>
</iq>
Gateways that receive a "hold" command from their Jingle side MUST generate a new offer on their SIP side, placing all streams in a "sendonly" state.
When relaying offers from SIP to XMPP however, gateways are not required to translate "sendonly" attributes into a "hold" command as this would not always be possible (e.g. when not all streams have the same direction). Additionally such conversions might introduce complications in case further offers placing a session of hold also contain other session modifications.
[[OPEN ISSUE: do we need to mention double hold here? That is, when you put me on hold after I did it first. Direction would then be "inactive".]]
[RFC3959] and [RFC3960] describe a number of scenarios relying on "early media". While similar attempts have also been made for XMPP [XEP-0269] support for early media is not currently widely supported in Jingle implementations. Therefore, gateways SHOULD NOT forward SDP answers from SIP to Jingle until a final response has been received, except in cases where the gateway is in a position to confirm specific support for early media by the endpoint (one approach to such support can be found in [XEP-0269] but it has not yet been standardized).
Gateways MUST however store early media SDP answers when they are sent inside a reliable provisional response. In such cases, a subsequent final response may follow without an actual answer and the one from the provisional response will need to be forwarded to the Jingle endpoint.
[RFC3261] defines a "Max-Forwards" header that allows intermediate entities such as SIP proxies to detect and prevent loops from occurring. The specifics of XMPP make such a prevention mechanism unnecessary for XMPP-only environments. With the introduction of SIP-to-XMPP gatewaying however, it would be possible for loops to occur where messages are being repeatedly forwarded from XMPP to SIP to XMPP to SIP, etc.
To compensate for the lack of a "Max-Forwards" header in SIP, gateways MUST therefore keep track of all SIP transactions and Jingle sessions that they are currently serving and they MUST block re-entrant messages.
[[OPEN ISSUE: In order for this to work, we need a consistent way of translating dialog IDs into Jingle sessions, and vice versa, so that the following can be verified: jingleSessID == toJingleSessID(toSipCallID( jingleSessID )). We need to mention mention spirals here as well. Alice could call Bob, but Bob forwards his call to Romeo. A spiral on the SIP side could end up becoming a loop if the gateway is in between.]]
[RFC4566] defines "a=fmtp" attributes for the transmission of format specific parameters as a single transparent string. Such strings can be used to convey either a single value or a sequence of parameters, separated by semi-colons, comas or whatever delimiters are chosen by a particular payload type specification.
<format name="paramName" value="paramValue"/>
[XEP-0177] on the other hand defines a "<format/>" element as follows:
These differences make it impossible to devise a generic mechanism that accurately translates format parameters from Jingle to SDP without the specifics of the payload being known to the gateway. This specification therefore makes the following recommendations for a best-effort attempt at translation:
[[OPEN ISSUE: we need to add examples for these transformations.]]
[RFC3261] defines semantics for dialog forking. Such semantics have not been defined for Jingle and need to be hidden from XMPP endpoints.
To achieve this SIP-to-XMPP MUST NOT forward more than one provisional response on their Jingle side. Typically they would do so only for the first provisional response they receive and ignore the rest. This provisional response SHOULD be forwarded as originating from a bare Jabber ID (JID) corresponding to the AOR URI found in the "From" header of the SIP provisional response. The gateway MUST NOT attempt to translate GRUUs into full JIDs because it cannot know at this stage, which of the dialogs established by these provisional responses will be used for the actual session.
Likewise, gateways conforming to this specification MUST NOT forward more than a single final response received through SIP to the Jingle side. The gateway SHOULD terminate the SIP sessions whose received final response wasn't forwarded to the Jingle side.
The following sections provide sample scenarios (or "call flows") that illustrate the principles of interworking from Jingle to SIP. These scenarios are not exhaustive.
The protocol flow for a basic voice chat for which an XMPP user (juliet@example.com) is the initiator and a SIP user (romeo@example.net) is the responder. The voice chat is consummated through a gateway. To simplify the example, the transport method negotiated is "raw user datagram protocol" as specified in [XEP-0177].
INITIATOR ...XMPP... GATEWAY ...SIP... RESPONDER
| | |
| session-initiate | |
|----------------------->| |
| IQ-result (ack) | |
|<-----------------------| |
| | INVITE |
| |---------------------->|
| | 180 Ringing |
| |<----------------------|
| session-info (ringing) | |
|<-----------------------| |
| IQ-result (ack) | |
|----------------------->| |
| | 200 OK |
| |<----------------------|
| session-accept | |
|<-----------------------| |
| IQ-result (ack) | |
|----------------------->| |
| | ACK |
| |---------------------->|
| MEDIA SESSION |
|<==============================================>|
| | BYE |
| |<----------------------|
| session-terminate | |
|<-----------------------| |
| IQ-result (ack) | |
|----------------------->| |
| | 200 OK |
| |---------------------->|
| | |
The packet flow is as follows.
First the XMPP user sends a Jingle session-initiation request to the SIP user.
<iq from='juliet@example.com/t3hr0zny'
id='hu2s61f4'
from='romeo@example.net/v3rsch1kk3l1jk'
type='set'>
<jingle xmlns='urn:xmpp:jingle:1'
action='session-initiate'
initiator='juliet@example.com/t3hr0zny'
sid='a73sjjvkla37jfea'>
<content creator='initiator'
media='audio'
name='this-is-the-audio-content'>
<description xmlns='urn:xmpp:jingle:app:rtp:1'>
<payload-type id='96' name='speex' clockrate='16000'/>
<payload-type id='97' name='speex' clockrate='8000'/>
<payload-type id='18' name='G729'/>
</description>
<transport xmlns='urn:xmpp:jingle:transport:raw-udp'>
<candidate ip='192.0.2.101' port='49172' generation='0'/>
</transport>
</content>
</jingle>
</iq>
The gateway returns an XMPP IQ-result to the initiator on behalf of the responder.
<iq from='juliet@example.com/t3hr0zny'
id='hu2s61f4'
to='romeo@example.net/v3rsch1kk3l1jk'
type='result'/>
The gateway transforms the Jingle session-initiate action into a SIP INVITE.
INVITE sip:romeo@example.net SIP/2.0
Via: SIP/2.0/TCP client.example.com:5060;branch=z9hG4bK74bf9
Max-Forwards: 70
From: Juliet Capulet <sip:juliet@example.com>;tag=t3hr0zny
To: Romeo Montague <sip:romeo@example.net>
Call-ID: 3848276298220188511@example.com
CSeq: 1 INVITE
Contact: <sip:juliet@client.example.com;transport=tcp>
Content-Type: application/sdp
Content-Length: 184
v=0
o=alice 2890844526 2890844526 IN IP4 client.example.com
s=-
c=IN IP4 192.0.2.101
t=0 0
m=audio 49172 RTP/AVP 18 96 97
a=rtpmap:96 sppex/16000
a=rtpmap:97 speex/8000
a=rtpmap:18 G729
The responder returns a SIP 180 Ringing message.
SIP/2.0 180 Ringing
Via: SIP/2.0/TCP client.example.com:5060;branch=z9hG4bK74bf9;\
received=192.0.2.101
From: Juliet Capulet <sip:juliet@example.com>;tag=t3hr0zny
To: Romeo Montague <sip:romeo@example.net>;tag=v3rsch1kk3l1jk
Call-ID: 3848276298220188511@example.com
CSeq: 1 INVITE
Contact: <sip:romeo@client.example.net;transport=tcp>
Content-Length: 0
The gateway transforms the ringing message into XMPP syntax.
<iq from='romeo@montague.net/v3rsch1kk3l1jk'
id='ol3ba71g'
to='juliet@example.com/t3hr0zny'
type='set'>
<jingle xmlns='urn:xmpp:jingle:1'
action='session-info'
initiator='juliet@example.com/t3hr0zny'
sid='a73sjjvkla37jfea'>
<ringing xmlns='urn:xmpp:jingle:apps:rtp:info:1'/>
</jingle>
</iq>
The initiator returns an IQ-result acknowledging receipt of the ringing message, which is used only by the gateway and not transformed into SIP syntax.
<iq from='juliet@example.com/t3hr0zny'
id='ol3ba71g'
to='romeo@example.net/v3rsch1kk3l1jk'
type='result'/>
The responder sends a SIP 200 OK to the initiator.
SIP/2.0 200 OK
Via: SIP/2.0/TCP client.example.com:5060;branch=z9hG4bK74bf9;\
received=192.0.2.101
From: Juliet Capulet <sip:juliet@example.com>;tag=t3hr0zny
To: Romeo Montague <sip:romeo@example.net>;tag=v3rsch1kk3l1jk
Call-ID: 3848276298220188511@example.com
CSeq: 1 INVITE
Contact: <sip:romeo@client.example.net;transport=tcp>
Content-Type: application/sdp
Content-Length: 147
v=0
o=romeo 2890844527 2890844527 IN IP4 client.example.net
s=-
c=IN IP4 192.0.2.201
t=0 0
m=audio 3456 RTP/AVP 97
a=rtpmap:97 speex/8000
The gateway transforms the 200 OK into a Jingle session-accept action.
<iq from='romeo@example.net/v3rsch1kk3l1jk'
id='pd1bf839'
to='juliet@example.com/t3hr0zny'
type='set'>
<jingle xmlns='urn:xmpp:jingle:1'
action='session-accept'
initiator='juliet@example.com/t3hr0zny'
responder='romeo@example.net/v3rsch1kk3l1jk'
sid='a73sjjvkla37jfea'>
<content creator='initiator'
media='audio'
name='this-is-the-audio-content'>
<description xmlns='urn:xmpp:jingle:app:rtp:1'>
<payload-type id='97' name='speex' clockrate='8000'/>
</description>
<transport xmlns='urn:xmpp:jingle:transport:raw-udp'>
<candidate ip='192.0.2.101' port='49172' generation='0'/>
</transport>
</content>
</jingle>
</iq>
If the payload types and transport candidate can be successfully used by both parties, then the initiator acknowledges the session-accept action.
<iq from='romeo@example.net/v3rsch1kk3l1jk'
id='pd1bf839'
to='juliet@example.com/t3hr0zny'
type='result'/>
The parties now begin to exchange media. In this case they would exchange audio using the Speex codec at a clockrate of 8000 since that is the highest-priority codec for the responder (as determined by the XML order of the <payloadtype/> children).
The parties may continue the session as long as desired.
Eventually, one of the parties (in this case the responder) terminates the session.
BYE sip:juliet@client.example.com SIP/2.0
Via: SIP/2.0/TCP client.example.net:5060;branch=z9hG4bKnashds7
Max-Forwards: 70
From: Romeo Montague <sip:romeo@example.net>;tag=8321234356
To: Juliet Capulet <sip:juliet@example.com>;tag=9fxced76sl
Call-ID: 3848276298220188511@example.com
CSeq: 1 BYE
Content-Length: 0
The gateway transforms the SIP BYE into XMPP syntax.
<iq from='romeo@example.net/v3rsch1kk3l1jk'
id='rv301b47'
to='juliet@example.com/t3hr0zny'
type='set'>
<jingle xmlns='urn:xmpp:jingle:1'
action='session-terminate'
initiator='juliet@example.com/t3hr0zny'
reasoncode='no-error'
sid='a73sjjvkla37jfea'/>
</iq>
The initiator returns an IQ-result acknowledging receipt of the session termination, which is used only by the gateway and not transformed into SIP syntax.
<iq from='romeo@example.net/v3rsch1kk3l1jk'
id='rv301b47'
to='juliet@example.com/t3hr0zny'
type='result'/>
This document has no actions for the IANA.
Detailed security considerations for session management are given for SIP in [RFC3261] and for XMPP in [XEP-0166] (see also [RFC6120]). The security considerations provided in [I-D.ietf-stox-core] also apply.