CLUE R. Hansen
Internet-Draft Cisco Systems
Intended status: Standards Track February 25, 2013
Expires: August 29, 2013
SDP and CLUE message interactions
draft-hansen-clue-sdp-interaction-01
Abstract
This document attempts to help resolve some of the complexities of
interaction between SDP and CLUE messages in call flows by providing
some strategies and some suggested syntax.
Status of this Memo
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 August 29, 2013.
Copyright Notice
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.
Hansen Expires August 29, 2013 [Page 1]
Internet-Draft SDP and CLUE message interaction February 2013
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Initial Assumptions . . . . . . . . . . . . . . . . . . . . . . 3
4. The CLUE framework: dividing the information between SDP
and CLUE messaging . . . . . . . . . . . . . . . . . . . . . . 4
4.1. CLUE information principally in CLUE channel . . . . . . . 4
4.2. Media encoding/decoding information in SDP, media
content information in CLUE messaging . . . . . . . . . . . 4
5. Interdependence of SDP and CLUE negotiation . . . . . . . . . . 6
6. Security Considerations . . . . . . . . . . . . . . . . . . . . 7
7. References . . . . . . . . . . . . . . . . . . . . . . . . . . 7
7.1. Normative References . . . . . . . . . . . . . . . . . . . 7
7.2. Informative References . . . . . . . . . . . . . . . . . . 7
Appendix A. Changes From Draft -00 . . . . . . . . . . . . . . . . 8
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 8
Hansen Expires August 29, 2013 [Page 2]
Internet-Draft SDP and CLUE message interaction February 2013
1. Introduction
One issue that has repeatedly come up in the development of CLUE is
the interconnected nature of many of the issues - making decisions in
any one area requires that decisions are made in other areas. One
particularly problematic area has been that of producing call flows:
many of the decisions that need to be made revolve around how offer/
answer exchanges and CLUE messages will interact, but without a good
understanding of what will be in SDP and what will be in CLUE these
decisions have been difficult to make.
In the hope of resolving some of these issues and allowing us to make
more progress on the subject of call flows and CLUE signalling
generally this draft addresses two issues that are hopefully not
dependent on decisions in other areas, both aspects of the
relationship between CLUE signalling and SDP. Hopefully this draft
will either provoke discussion, or document decisions that people
feel are obvious but aren't currently reflected in writing.
2. Terminology
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] and
indicate requirement levels for compliant implementations.
3. Initial Assumptions
This section enumerates a few assumptions based on previous
discussion which are, at this stage, hopefully uncontroversial.
CLUE information such as capture descriptions are unsuitable for SDP,
and as such there will be an alternate method for sending CLUE
messages end to end. In a call scenario where both sides wish to
send and receive this CLUE negotiation takes the form of two
independent, uni-directional exchanges; on each exchange one device
provides its send capabilities while the other side determines what
it wishes to receive.
This CLUE negotiation will never enable or require a call to exceed
boundaries negotiated in SDP. This most obviously applies to
bandwidth, both for the total call and for negotiated sessions, but
also means that codec-specific limitations such as the maximum number
of H.264 macroblocks a second a receiver can process MUST be
respected.
Hansen Expires August 29, 2013 [Page 3]
Internet-Draft SDP and CLUE message interaction February 2013
4. The CLUE framework: dividing the information between SDP and CLUE
messaging
The CLUE Framework [I-D.ietf-clue-framework] defines the information
that will be needed to successfully negotiate a CLUE call, but does
not define the mechanism by which this information is conveyed. This
section provides two options for dividing this information between
SDP and CLUE signalling, without proposing explicit signalling for
either channel (merely what information needs to be conveyed in
each).
4.1. CLUE information principally in CLUE channel
One approach that has been a major part of CLUE discussions has been
to make no significant additions to SDP, and continue to use it only
for the negotiation of RTP sessions. The sessions are then
potentially subdivided into multiple streams using CLUE signalling.
In this model standard SDP signalling provides the envelope within
which CLUE negotiates the number and content of multiple streams.
This method has a number of advantages - there is no need for
additional SDP syntax, making interoperability with existing devices
simple and concentrating new signalling in a single location (the
CLUE negotiation). There is also clear separation of
responsibilities between SDP and CLUE: as normal SDP negotiates the
specifics of the RTP sessions: address and ports, supported codecs,
receive maxima and so on, while CLUE messaging then specifies how
many streams are to be multiplexed on a port, details for
demultiplexing, content of those streams, encoding limits and so on.
The only necessary addition to the SDP would be a label [RFC4574]
attribute per media line to allow CLUE messaging to identify them.
Unfortunately, there are some downsides to this approach. The
primary one is that all multiplexing of streams is entirely dependant
on the CLUE channel - as such this is not a method applicable to
other applications. Since other groups within the IETF have an
interest in such multiplexing for reasons other than enabling
telepresence scenarios they would have to invent other methods for
negotiating similar multiplexing - both inefficient, and likely
problematic when CLUE and some other solution involving
multistreaming are both used in the call scenario.
4.2. Media encoding/decoding information in SDP, media content
information in CLUE messaging
An alternative approach is to divide the information in the CLUE
Framework [I-D.ietf-clue-framework] into the information specific to
encoding and decoding RTP streams, and the content of those streams.
Hansen Expires August 29, 2013 [Page 4]
Internet-Draft SDP and CLUE message interaction February 2013
On the advertising side this split is fairly natural: most of the
information in the framework relates to the number, content, physical
dimensions and simultaneity of the media captures available,
information related to the contents of the media streams rather than
the streams themselves. In contrast, the encoder and encoder group
information gives the limits on the media streams the sender can
provide, with parameters such as bandwidth, max h.264 macroblocks per
second and other parameters relevant to SDP. These are defined as
sender limitations rather than receiver ones and so are not directly
analogous to existing SDP parameters, but are better suited to SDP
than CLUE.
When it comes to receiver selection the separation between parameters
that logically should be in CLUE and should be in SDP is no longer so
clean-cut, as the receiver must specify capture encodings, choosing
both which captures they wish to receive and the media limitations on
such streams. The latter limitations are obviously suited to SDP,
but information about captures is more relevant to the CLUE channel.
The CLUE-specific information, however, is limited to simply
selecting a capture for the stream.
The ability to describe the sender's encoder limitations for
multiplexed streams along with the receiver's selection of those
streams and the media limitations, SSRCs and other demultiplexing
information are all requirements that are not specific to CLUE;
having them in SDP means that a consistent mechanism can be used by
CLUE as well as by other call scenarios wishing to support additional
media streams in this fashion. Capture information, in contrast, is
CLUE-specific and as such is sensible to keep in the CLUE channel.
The CLUE channel will also reference the SDP, linking captures to
encoding capabilities and identifying which capture is desired for
each stream. This split of information means that any change in
capture information on the part of a sender does not necessitate an
offer/answer exchange of SDP if there is no corresponding change to
the encoding capabilities of that sender - only a new CLUE
advertisement is required.
This approach leads to a number of dependencies between the SDP and
CLUE messages - the sender must define which captures and capture
scenes are usable with which streams/encodings, while the receiver
must define what capture they wish to receive with a particular
encoding. These could take the form of references in SDP to CLUE,
references in CLUE to SDP or references in each to the other.
However, this draft proposes that all such references MUST be from
CLUE messages to SDP, not the other way around. By ensuring all
depenencies are unidirectional it reduces the complexity of
integrating the two signalling methods. There are multiple reasons
for having references be CLUE->SDP and not the other way around: one
Hansen Expires August 29, 2013 [Page 5]
Internet-Draft SDP and CLUE message interaction February 2013
is that logically CLUE is providing metadata about the contents of
streams that are negotiated in SDP, so it makes sense for CLUE to be
dependent on SDP and not visa-versa. Another is that middle boxes
wishing to monitor or alter SDP can then do so without necessarily
needing to involve themselves in the CLUE channel as SDP remains
self-contained.
5. Interdependence of SDP and CLUE negotiation
With separate negotiation of SDP and CLUE there is the question of
how to deal with dependencies between these two channels. The number
of dependancies depends on how the information defined in the CLUE
Framework [I-D.ietf-clue-framework] is split between SDP and CLUE, as
discussed in the previous section, but even in the case where all new
information is in CLUE there will still be some dependencies as it
will be necessary to determine which m-lines the CLUE signalling is
referring to. However, because we have two signalling methods
changes that require alterations in both CLUE and SDP are no longer
atomic: one message will be processed before the other. There has
been debate within the working group about how this will be dealt
with, as such a decision has significant effect upon call flows.
This draft proposes that CLUE messages and SDP messages should be
independent: parameters in CLUE messages MAY exceed values negotiated
in SDP, or may make reference to SDP contents not present in the most
recent offer/answer exchange. Without this provision, SDP and CLUE
messages become part of a single negotiation, and a change on either
by either side may necessitate an exchange of the other message type.
For instance, removing stream information from SDP might first
necessitate sending a new CLUE message removing the references to
this stream. The state machine required to ensure validity of
negotiation will be complicated, and there will be a number of
invalid states which must be avoided. This is further complicated by
the fact that, even if both ends of a call obey the constraints to
ensure validity, a middle box may choose to rewrite an SDP such that
an invalid state is reached.
Making the two message types independent significantly reduces the
complexity of the state machines required. And with the message
flows independent there is no way for an invalid state to occur when
the two negotiations contain contradictory information. A cost of
this is that endpoints will now need to deal with the fact that CLUE
messages may contain parameters exceeding those negotiated in SDP, or
referencing SDP content that does not exist. However, this is
analogous to an issue endpoints already deal with in SDP. For
instance, the sum of bandwidth parameters for various m-lines can
exceed the overall session bandwidth. Not only is this not invalid,
Hansen Expires August 29, 2013 [Page 6]
Internet-Draft SDP and CLUE message interaction February 2013
but it can be desirable, as it allows the sender to prioritise
streams. What can be sent for any device is simply the intersection
of what is permitted by the most recent SDP offer/answer, and the
outcome of the CLUE negotiation; implementations should ignore
references to entities in the other negotiation that do no exist.
This does not mean that there will be no interaction between SDP and
CLUE messaging - a device wishing to add a new stream may well need
to update both their SDP and their CLUE negotiations. However, there
is no fixed order in which this must be done and no requirement for
them to be updated in a particular order or fashion; it is left to
the implementation to renegotiate the channels as it sees fit. If
updates to both negotiations are required for a new stream to be
added, then the new stream will not be available until both
renegotiations are complete - the completion of the first
renegotiation will have no effect.
6. Security Considerations
This draft only addresses how best to split information between SDP
and CLUE signalling and the interdependencies between these two
methods of signalling, it does not define the signalling or
information itself. As such this draft should require no additional
security considerations.
7. References
7.1. Normative References
[I-D.ietf-clue-framework]
Duckworth, M., Pepperell, A., and S. Wenger, "Framework
for Telepresence Multi-Streams",
draft-ietf-clue-framework-09 (work in progress),
February 2013.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
7.2. Informative References
[RFC4574] Levin, O. and G. Camarillo, "The Session Description
Protocol (SDP) Label Attribute", RFC 4574, August 2006.
Hansen Expires August 29, 2013 [Page 7]
Internet-Draft SDP and CLUE message interaction February 2013
Appendix A. Changes From Draft -00
o Reordered main sections, as in the discussion about
interdependence of SDP and CLUE is is useful to reference the
split between CLUE and SDP.
o Added more detail to the argument of why dependencies should be
CLUE->SDP and not the other way around or in both directions.
o Fixed spelling issues and did some minor rewording.
Author's Address
Robert Hansen
Cisco Systems
San Jose, CA 95134
USA
Email: rohanse2@cisco.com
Hansen Expires August 29, 2013 [Page 8]