INTERNET-DRAFT J. Ott/C. Perkins
Expires: January 2003 TZI/ISI
July 2002
SDPng Transition
draft-ietf-mmusic-sdpng-trans-01.txt
Status of this memo
This document is an Internet-Draft and is subject to all provisions
of Section 10 of RFC2026.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
other groups may also distribute working documents as Internet-
Drafts.
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".
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt
To view the list Internet-Draft Shadow Directories, see
http://www.ietf.org/shadow.html.
Distribution of this document is unlimited.
Abstract
The Session Description Protocol (SDP) is today widely used in the
Internet to announce as well as negotiate multimedia sessions and
exchange capabilities. Having originally been designed for session
announcements only, as opposed to announcements and capabilities
negotiation announcements, native SDP lacks numerous features to be
applicable in many session scenarios. Numerous extensions have been
developed to circumvent SDP's shortcomings -- but they have also
repeatedly shown its inherent limitations. A successor protocol --
termed "SDPng" for the time being -- is developed to address the
aforementioned needs of Internet applications in a more structured
manner. With the huge installed base of SDP-based applications, a
migration path needs to be developed to move from SDP to SDPng over
time. This document outlines how this migration can be achieved: in
general as well as for the various IETF control protocols that
potentially make use of SDP and SDPng.
This document is a product of the Multiparty Multimedia Session
Control (MMUSIC) working group of the Internet Engineering Task
Force. Comments are solicited and should be addressed to the working
Ott/Perkins [Page 1]
INTERNET-DRAFT SDPng Transition July 2002
group's mailing list at mmusic@ietf.org and/or the authors.
1. Introduction
SDP is now widely used within the Internet community to describe
media sessions and, in a limited fashion, system capabilities
relating to (multi)media sessions, for a variety of application
scenarios: session announcements, interactive session setup,
capability assessment and remote control of media streams. All but
the first of these are rather different from what SDP was originally
designed for -- but all of them share the idea of setting up and
configuring media streams. Over time, its wide range of uses has
revealed numerous shortcomings -- most of which stem from the fact
that SDP has been used for lack of a better alternative and its
semantics have been re-interpreted to make it fit the respective
scenarios' needs. In many cases, workrounds (typically called
"extensions") for those shortcomings could be found which are often
rather cumbersome. While this practice has extended SDP's lifetime
and provided at least a suitable basis for numerous applications, in
parallel, a successor protocol -- currently referred to as "SDPng"
-- has been developed.
It is worthwhile noting that the aforementioned applications'
needs are sufficiently similar for a single description protocol
to take care of them if it was designed for this purpose from
the beginning. As a lesson learned from SDP, any further
expansion in scope should be avoided where no clear fit can be
seen -- and specific (different) solutions should be developed
instead.
The design of SDPng takes into account the requirements arising from
the above application scenarios and puts particular emphasis on
protocol extensibility and modularization of extensions, at the same
time keeping the core description format simple. SDPng uses a
different (more expressive) syntax than SDP does and hence is not
backward compatible at the syntax level. Nevertheless, the concepts
of SDPng take into account the migration issues from SDP to SDPng by
providing straightforward mappings between the two formats where
possible and try to maximize compatibility from a semantics
perspective.
The current revisions of SDP and SDPng are documented in
[1] draft-ietf-mmusic-sdp-new-10.txt and
[2] draft-ietf-mmusic-sdpng-05.txt.
For SDP, numerous additional documents need to be taken into account:
[3] RFC 3108
[4] draft-ietf-mmusic-fid-06.txt
Ott/Perkins [Page 2]
INTERNET-DRAFT SDPng Transition July 2002
[5] draft-andreasen-mmusic-sdp-simcap-05.txt
[6] draft-ietf-mmusic-sdp-offer-answer-02.txt
[7] draft-fairlie-mmusic-sdp-sctp-00.txt
[8] draft-ietf-mmusic-sdp-comedia-03.txt
[9] draft-taylor-mmusic-sdp-tdm-01.txt
[10] draft-ietf-mmusic-sdp4nat-02.txt
[11] draft-ietf-mmusic-kmgmt-ext-05.txt
[12] draft-ietf-mmusic-sdp-ipv6-03.txt
This document outlines a migration path from SDP to SDPng, starting
from a short overview of the current application scenarios. In the
next step, we highlight which design decisions taken for SDPng should
simplify a smooth migration and describe how mappings between the two
description formats can be performed at an abstract level. We then
address procedural issues for integrating SDP and SDPng into the
various protocols relying on those media description formats.
Finally, we summarize work items on the agenda for SDPng.
2. Application Scenarios
The following session control protocols that make use of SDP have
been standardized in the IETF so far:
1. SDP was originally developed to announce (Mbone-based)
multimedia sessions via session directories using the Session
Announcement Protocol (SAP) -- but other mechanisms for
disseminating the session descriptions (such as HTTP, SMTP,
NNTP, etc.) are conceivable as well.
The major property of this application scenario is that the
creator of the session description defines a (set of) fixed
choice(s) for all media types in a conference and the conference
partipants have no way to influence these. If they support at
least one of the codecs for a particular media type they can
participate in this media session, otherwise they cannot. There
is no interaction between sender(s) and receiver(s) to negotiate
the media stream codecs and parameters.
This scenario is referred to as "announcement".
2. Another use of SDP is in conjunction with the Real-Time
Streaming Protocol (RTSP). In RTSP, SDP is used to convey
descriptions of a media stream interactively requested to be
played from a server (or recorded by a server). SDP itself is
not used for capability negotiation, not even for the addresses
to be used; those are negotiated within RTSP and may override
Ott/Perkins [Page 3]
INTERNET-DRAFT SDPng Transition July 2002
the addresses specified as part of SDP.
This scenario is referred to as "retrieval".
3. With SIP, SDP is used to propose media stream configurations and
choose out of these (i.e. enable a subset of these). By
proposing and accepting media stream configurations, endpoints
use SDP to implicitly describe their capabilities and carry out
a negotiation procedure on the media streams to use.
In the context of SIP, specific meanings (including required
extensions) have been defined for use of SDP with unicast
addresses, for connection-oriented transports, and for certain
media level attributes (such as the direction attribute send-
only, receive-only, and inactive).
Numerous extensions have been proposed to extend SDP to better
suit SIP's needs. Besides a description of the offer/answer
model, these extensions particularly include the ability to
describe simultaneous capabilites and to group media stream
semantically.
This scenario is referred to as "offer/answer".
4. SDP is used to convey the capability descriptions of a MEGACO
media gateway (MG) to its media gateway controller (MGC) as well
as for the MGC to instruct the MG where to send media streams to
and from where to receive media streams, including codec and
parameter choice.
For this purpose, SDP has been modified/extended to some degree
to fit the MEGACO needs.
This scenario is referred to as "gateway control".
It should noted that the original SDP concept already provided an
extension mechanism to cover other network types than IPv4 and IPv6;
however, specific extensions have only been defined recently for ATM
and are now under discussion for TDM. Extensions to other transport
(including radio interfaces or next generation wireless networks) as
well as to new types of session descriptions (e.g. electronic program
guides) are conceivable.
3. Mapping SDP to SDPng
On a transition path from SDP to SDPng, allowing for a somewhat
straightforward mapping of (parts of) one description format onto the
other is of crucial importance. SDPng has been designed in way that
allows many of the session description features of SDP to be easily
mapped onto the SDPng format and vice versa -- except that SDPng is
more expressive than SDP and hence information loss is not unlikely
to occur when doing the reverse mapping. The final mapping rules
between SDP and SDPng to be drawn up shall ensure that when mapping
Ott/Perkins [Page 4]
INTERNET-DRAFT SDPng Transition July 2002
SDP to SDPng and then back to SDP will produce an SDP that is
functionally identical to the one originally fed into the mapping
process. Note that the use of a number of SDP extensions (FID,
SIMCAP) may be implied in this mapping process, depending on the use
of SDP. The mapping rules will ensure that no information loss will
occur when translating from SDP to SDPng.
The SDPng design uses a structure of four sections: definitions,
potential or actual configurations, constraints and session
attributes. Of these, the "Configurations" and "Session
Attributes" sections map well onto the current SDP. The
"Definitions" and "Constraints" sections provide additional
structure which is not directly expressible in SDP.
o At the media description level, the Potential and Actual
Configurations specified in the "Configurations" section maps
well to media descriptions ("m=", possibly "c=", and
associated attributes ("a=") lines).
o At the session description level, the SDP session parameters are
largely reflected in the "Session Attributes" section of
SDPng. The attributes proven suitable for session announcements
have been used as the basis when defining SDPng.
In SDPng, media descriptions are explicitly tagged with identifiers
and thus are easily referenced for semantically grouping media
streams (e.g. to describe alternative audio in different languages,
media streams to be synchronized, or media streams to carry the same
information simultaneously but with different encodings) -- as has
been defined for SDP in a limited fashion by the "fid" attribute
set. SDPng allows even to more formally describe the syntax of
individual or compound media streams in the "Session Attributes"
section. Furthermore, SDPng supports a superset of additional
constraints that may be realized by the "simcap" extensions for SDP
in the "Constraints" section.
Additional address families such as ATM or TDM bearers, next
generation wireless network bearers, DVB channels, etc. can be
incorporated into SDPng by defining the appropriate extensions for
the SDPng transports.
Similarly, new codecs can be added by just defining new codec
specifications or defining entire new classes of applications to be
described as new content types ("codec") to be carried in a media
session (including e.g. text, fax, slide presentations, shared
editors, etc). If necessary, sophisticated parameter structures can
be supported (even though the authors believe that simplicity is key
to interoperability here). This is similar to, but more structured
than, the definition of new codecs attributes/MIME registrations in
SDP.
By means of its conceptual differentiation into Potential and Actual
Configurations, SDPng supports both indicating a system's
Ott/Perkins [Page 5]
INTERNET-DRAFT SDPng Transition July 2002
capabilities (without specifying transport addresses) separately from
the instantiation of a particular media stream as well as conveying
capability descriptions and instantiation proposals at the same time
-- thereby providing a good fit for all the above session control
scenarios: the "announcement" and "retrieval" scenarios will just
use rather fixed Actual Configurations. The "offer/answer" model
will use use Actual Configuration but use them to negotiate media
strams in a two-way handshake but may in addition use Potential
Configurations to indicate capabilities that shall not be used
immediately. The "gateway control" scenario will use both:
Potential Configurations to describe an MG's capabilities and Actual
Configurations for setting up media sessions at MGs as well as
retrieving information about currently active media sessions. This
differentiation is not directly expressible in SDP, although various
extensions can be used to overload SDP semantics to achieve at least
part of this effect.
Finally, SDPng is also intended to allow for content-independent
negotiation of session parameters by defining collapsing/intersection
rules. In particular, SDPng tries to take the need for multicast-
based distributed calculation of joint capabilities into account for
those rules (but note that it is *not* intended as a generic format
for describing conference state information). Such functionality is
not covered by current SDP.
4. Integration with Session Control Protocols
This section outlines for each of the session control protocols
described above how SDP and SDPng can be used in parallel and
indicates how a suitable transition could be achieved.
4.1. Session Announcement Protocol (SAP)
There are two revision of SAP specified, version 0 which is
implemented in a number of experimental tools, and version 1 which is
defined in RFC 2972.
SAPv0:SAPv0 does not support a mechanism to identify the content type
of a session announcement but implicitly assumes SDP. Proper
parsers will note that the contents of the SAPv0 message does
not begin with a "v=" line and hence will ignore the entire
announcement. SDPng contents MAY be identified by the character
sequence "<sdpng" in the beginning of the announcement body --
but such content is not strictly legal in SAPv0.
SAPv1:In SAPv1, an explicit payload type field (containing a MIME
type) is available and SHOULD used to differentiate between SDP
anf SDPng contents. two approaches are conceivable: Either
multipart MIME message is used with two parts containing the
same session descriptions -- one expressing it in SDP and the
other in SDPng. Alternatively, two alternate session
announcements may be used (being properly distinguished by the
SDP "o=" field and the SDPng equivalent).
Ott/Perkins [Page 6]
INTERNET-DRAFT SDPng Transition July 2002
It is RECOMMENDED that implementations recognize the MIME
multipart/alternative type in SAPv1 announcements, allowing for
a simple transition to SDPng.
It should be noted that current session directory implementations
only support SDP. Nevertheless, using the SAP Message Identifier
Hash and the source address, they should be able to perform session
deletions and modifications properly -- even without understanding
the format contained in the SAP message body.
For the introduction of SDPng, session announcements SHOULD be made
"bi-lingual", i.e. in SDP and SDPng. If a SAP announcer for some
reason knows that all its potential audience will support SDPng, the
SDP announcement SHOULD be omitted.
It should be noted that, for IPv4-based multicast sessions, session
directories still may rely on parsing the session specifications to
avoid clashes in the multicast address space. Introducing a new
session description language will prevent older implementations from
continuing this practice successfully -- assuming that only SDPng
announcements are used and/or that old implementations do not support
MIME multipart/alternative message bodies. This use of SAP is
deprecated, of course.
4.2. Real-Time Streaming Protocol (RTSP)
RTSP uses SDP to provide presentation descriptions (with a
presentation comprising one or more media sessions), typically
communicated from the server to the client (for playing) and in the
opposite direction for recording. The presentation description may
also include initialization data for the various media streams and
URLs to be used for controlling the entire presentation as well as
the individual media sessions. Transport parameters -- such as IP
addresses, port numbers, etc. -- are conveyed as part of RTSP header
fields.
RTSP uses the Content-Type: header field to indicate the format of
the enclosed entity. This provides a straightforward means for
distinguishing SDP and SDPng-based presentation descriptions. In
addition, the Accept: header SHOULD be used by the client, to
indicate which content types it supports. If the client specifies
both SDP and SDPng as acceptable, the server SHOULD provide only the
SDPng-based presentation description.
If the client does not indicate a particular Content-Type: the server
can, theoretically, use MIME multipart bodies to convey both
description types simultaneously.
[Editors note: can implementors comment on their ability to
parse such content?]
In general, it would be preferrable to have the servers migrate to
always supporting both description formats, thus enabling the clients
Ott/Perkins [Page 7]
INTERNET-DRAFT SDPng Transition July 2002
to choose.
Finally, RTSP makes special provision to interworking with firewalls
by including the crucial transport parameters in a separate RTSP
header field _in_addition_ to the presentation description. This
practice in principle allows to change the presentation description
format without having to worry about the operation of firewalls and
similar devices.
4.3. Session Initiation Protocol (SIP)
The use of SDP with SIP follows the offer/anser model and is
described in [6]. It is key to the (efficiency of the) offer/answer
model that a complete capability exchange and media stream
instantiation be carried out in one round-trip -- which is supported
by SDP. While SDPng allows to separate capability exchange from
media sesssion instantiation, those two pieces are also easily
integrated in a single step.
SIP also uses a Content-Type: header to indicate the nature of data
carried in its message body; and SIP explicitly calls for supporting
MIME multipart message bodies. While, again, the use of MIME
multipart/alternative would in principle be possible (from a
theoretical perspective), issues regarding the actual implementation
of multipart/alternative in SIP entities have been raised. As
backward compatibility has to be achieved, a different approach is
suggested:
A SIP UAC MAY use an SDPng message body in a SIP INVITE (or other)
message. If the SIP UAS does not support SDPng, it will return a
"415 Unsupported Media Type" response to the UAC and indicate
acceptable content types in the Accept: header (probably including
"application/sdp"). The SIP UAC MUST then retry INVITE (or other)
message using the indicated session description language. The SIP
UAC SHOULD cache knowledge about which peers did not understand SDPng
as session description formats for a limited amount of time (e.g.
several days) so that extra round-trips for session setup are only
incurred infrequently. Whenever a peer has sent an SDPng description
(or it is known from other means that the peer supports SDPng), this
information SHOULD also be cached.
The SIP Accept: header can be exploited to determine the capability
of a peer to understand SDPng in addition (or instead) of plain SDP.
Methods such as OPTIONS MAY be used to determine a peer's support for
SDPng. However, a peer's capabilities may not be known when the
first message is sent which may introduce an extra round-trip if
including SDP and SDPng in the initial INVITE message is not an
option. Further approaches to make a UA's support for SDPng known
ahead of time should be explored.
A number of SDP extensions have been motivated by SIP-based
applications and these need to be accommodated in SDPng as well.
Features such as "simcap" and "FID" are inherently supported by
Ott/Perkins [Page 8]
INTERNET-DRAFT SDPng Transition July 2002
SDPng; proper definitions for connection-oriented media need to be
fully understood and then incorporated. Key management attributes as
defined in [11] need to be included (not just for SIP) and so may
need to be general mechanisms to signal security capabilities [11]
[13] and indicate their optional or mandatory use. The same applies
to quality of service parameters [13] (which are largely also
motivated by SIP but are also useful with control protocols).
In the context of SIP, a number of special rules to deal with certain
SDP fields are set up (e.g. to work with NATs). The SDPng
development needs to make sure that similar definitions are provided
(as need be the handling of those in SIP).
[Editor's note: This section needs more work on details.]
4.4. Media Gateway Control Protocol (MEGACOP)
The MEGACO specification already supports two different encodings for
capability and media stream descriptions: a text-based variant based
upon (a modified) SDP and a binary representation of the same
information set. MGCs are required to implement both encodings while
MGs have the choice to pick either or both. Differentiation between
the protocol encoding variants is done using different port numbers:
2944 for the text-based and 2945 for the binary encoding.
Unfortunately, within the text-based encoding, there is no means to
differentiate several description formats. SDP messages are carried
as an "octet string" without any type identifier. Defining a third
port number for this further differentiation does not seem to be
appropriate, particularly since the message encoding is still a text
format.
The remaining means for distinction is that an SDP specification
would start with a "v=0" line while an SDPng document would begin
with an "<sdpng" part.
[Editor's Note: MEGACOP also supports a binary encoding for SDP
messages; we can assume that not all of the SDPng messages will
be expressible this binary encoding. How shall we handle
these?]
5. SDPng and Middleboxes
TBD.
6. Directing the Evolution of SDP
With the transition from SDP to SDPng, there is the question of the
evolution of SDP, and legacy systems which use it.
The SDP specification (draft-ietf-mmusic-sdp-new-10.txt) is stable,
and mostly corrects errors in the original specification, with the
addition of very few new features. This is expected to be published
Ott/Perkins [Page 9]
INTERNET-DRAFT SDPng Transition July 2002
as a draft standard RFC shortly.
A number of extensions to SDP for use in offer/answer scenarios are
also close to completion. These include grouping (draft-ietf-mmusic-
fid-05.txt) and capability negotiation (draft-andreasen-mmusic-sdp-
simcap-04.txt). We expect these to be completed, and published as
proposed standard RFCs, to bring minimal capability/alternative
descriptions to SDP.
Related is the SDP offer/answer model for SDP, currently under
development as draft-rosenberg-mmusic-sdp-offer-answer-01.txt). This
defines the model used to complete the steps of a negotiation using
SDP.
All these are subsumed into SDPng, so there should be no further need
for development in these areas; applications with requirements that
are not met by these specifications should use SDPng.
There have recently been proposals to add quality of service
negotiation for SDP and, similarly, we expect other extensions to be
proposed over time. Due to the well-known limitations of SDP, we do
not believe it appropriate to continue development of more elaborate
extensions: for negotiation, for QoS, for security, and for other
general-purpose or application-specific needs.
Instead, such new work should be done in the framework of SDPng where
applications and their requirements for (new) expressiveness in end-
to-end exchanges to negotiate and configure media sessions will
hopefully act as a driver for that process.
7. Author's Addresses
Joerg Ott <jo@tzi.org>
Universitaet Bremen
MZH 5180
Bibliothekstr. 1
D-28359 Bremen
Germany
tel:+49-421-201-7028
sip:jo@tzi.org
Colin Perkins <csp@isi.edu>
USC Information Sciences Institute
3811 North Fairfax Drive, Suite 200
Arlington, VA 22203
USA
tel:+1-703-812-3705
Ott/Perkins [Page 10]
INTERNET-DRAFT SDPng Transition July 2002
8. References
[1] Mark Handley, Van Jacobson, Colin Perkins, "SDP: Session
Description Protocol," Internet Draft draft-ietf-mmusic-sdp-
new-10.txt, Work in Progress, May 2002.
[2] Dirk Kutscher, Joerg Ott, Carsten Bormann, "Session Description
and Capability Negotiation", Internet Draft draft-ietf-mmusic-
sdpng-05.txt, Work in Progress, July 2002.
For SDP, numerous additional documents need to be taken into account:
[3] 3108 R. Kumar, M. Mostafa, "Conventions for the use of the
Session Description Protocol (SDP) for ATM Bearer Connections,"
RFC 3108, May 2001.
[4] Gonzalo Camarillo, Jan Holler, Goran AP Eriksson, Henning
Schulzrinne, "Grouping of media lines in SDP," Internet Draft
draft-ietf-mmusic-fid-06.txt, Work in Progress, February 2002.
[5] F. Andreasen, "SDP Simple Capability Negotiation," Internet
Draft draft-andreasen-mmusic-sdp-simcap-05.txt, Work in
Progress, February 2002.
[6] Jonathan Rosenberg, Henning Schulzrinne, "An Offer/Answer Model
with SDP," Internet Draft draft-ietf-mmusic-sdp-offer-
answer-02.txt, Work in Progress, February 2002.
[7] Robert Fairlie-Cuninghame, "Guidelines for specifying SCTP-
based media transport using SDP," Internet Draft draft-fairlie-
mmusic-sdp-sctp-00.txt, Work in Progress, May 2001.
[8] David Yon, "Connection-Oriented Media Transport in SDP,"
Internet Draft draft-ietf-mmusic-sdp-comedia-03.txt, Work in
Progress, June 2002.
[9] Tom Taylor, "Conventions for the use of the Session Description
Protocol (SDP) for Digital Circuit Connections," Internet Draft
draft-taylor-mmusic-sdp-tdm-01.txt, Work in Progress, April
2002.
[10] Christian Huitema, "RTCP attribute in SDP," Internet Draft
draft-ietf-mmusic-sdp4nat-02.txt, Work in Progress, February
2002.
[11] Jari Arkko, Elisabette Carrarra, Fredrik Lindholm, Mats Naslund,
Karl Norrman, "Key Management Extensions for SDP and RTSP,"
Internet Draft draft-ietf-mmusic-kmgmt-ext-05.txt, Work in
Progress, June 2002.
[12] Sean Olson, Gonzalo Camarillo, Adam Roach, "Support for IPv6 in
SDP," Internet Draft draft-ietf-mmusic-sdp-ipv6-03.txt, Work in
Ott/Perkins [Page 11]
INTERNET-DRAFT SDPng Transition July 2002
Progress, February 2002.
[13] G. Camarillo (ed), W. Marshall (ed), J. Rosenberg, "Integration
of Resource Management and SIP", Internet Draft draft-ietf-sip-
manyfolks-resource-07.txt, Work in Progress, April 2002.
9. Full Copyright Statement
Copyright (C) The Internet Society (1997). All Rights Reserved.
This document and translations of it may be copied and furnished to
others, and derivative works that comment on or otherwise explain it
or assist in its implmentation may be prepared, copied, published and
distributed, in whole or in part, without restriction of any kind,
provided that the above copyright notice and this paragraph are
included on all such copies and derivative works. However, this
document itself may not be modified in any way, such as by removing
the copyright notice or references to the Internet Society or other
Internet organizations, except as needed for the purpose of
developing Internet standards in which case the procedures for
copyrights defined in the Internet Standards process must be
followed, or as required to translate it into languages other than
English.
The limited permissions granted above are perpetual and will not be
revoked by the Internet Society or its successors or assigns.
This document and the information contained herein is provided on an
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Ott/Perkins [Page 12]