INTERNET-DRAFT          RTCP attribute in SDP            May 12, 2003

INTERNET DRAFT                                             C. Huitema
<draft-ietf-mmusic-sdp4nat-04.txt>                          Microsoft
Expires November 12, 2003                                May 12, 2003

                        RTCP attribute in SDP

Status of this memo

This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC2026.

This document is an Internet-Draft. 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.

The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.

Abstract

The session description protocol (SDP) is used to describe the
parameters of media streams used in multimedia sessions. When a
session requires multiple ports, SDP assumes that these port have
consecutive numbers. However, when the session crosses a network
address translation device that also uses port mapping, the ordering
of ports can be destroyed by the translation. To handle this, we
propose an extension attribute to SDP.

1.      Introduction

The session invitation protocol (SIP, [RFC3261]) is often used to
establish multi-media sessions on the Internet. There are often
cases today in which one or both end of the connection are hidden
behind a network address translation device [RFC2766]. In this case,
the SDP text must document the IP addresses and UDP ports as they
appear on the 'public Internet' side of the NAT; in this memo, we
will suppose that the host located behind a NAT has a way to obtain
these numbers; a possible way to learn these numbers is briefly
outlined in section 3. However, just learning the numbers is not
enough.

The SIP messages use the encoding defined in SDP [RFC2327] to
describe the IP addresses and TCP or UDP ports used my the various
media. Audio and video are typically sent using RTP [RTP-NEW], which
requires two UDP ports, one for the media and one for the control
protocol (RTCP). SDP carries only one port number per media, and
states that 'other ports used by the media application (such as the
RTCP port) should be derived algorithmically from the base media
port.' RTCP port numbers were necessarily derived from the base
media port in older versions of RTP (such as [RFC1889]), but now
that this restriction has been lifted, there is a need to specify
RTCP ports explicitly in SDP. Note, however, that implementations of
RTP adhering to the earlier [RFC1889] specification may not be able
to make use of the SDP attributes specified in this document.

When the NAT device performs port mapping, there is no guarantee
that the mappings of two separate ports reflects the sequencing and
the parity of the original port numbers; in fact, when the NAT
manages a pool of IP addresses, it is even possible that the RTP and
the RTCP ports may be mapped to different addresses. In order to
successfully establish connections despite the misordering of the
port numbers and the possible parity switches caused by the NAT, we
propose to use a specific SDP attribute to document the RTCP port
and optionally the RTCP address, and we also propose to make the
behavior of RTP implementations more conforming to the robustness
principle.

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 [RFC2119].

2.      Description of the solution

The main part of our solution is the declaration of an SDP attribute
for documenting the port used by RTCP.

2.1.    The RTCP attribute

The RTCP attribute is used to document the RTCP port used for media
stream, when that port is not the next higher (odd) port number
following the RTP port described in the media line. The RTCP
attribute is a ôvalueö attribute, and follows the general syntax
specified page 18 of [RFC2327]: "a=<attribute>:<value>". For the
RTCP attribute:

* the name is the ascii string 'rtcp' (lower case),

* the value is the RTCP port number and optional address.

The formal description of the attribute is defined by the following
ABNF syntax:

rtcp-attribute =  ôa=rtcp:ö port  [nettype space addrtype space
                         connection-address] CRLF

In this description, the 'port', 'nettype', 'addrtype' and
ôconnection-addressö tokens are defined as specified in 'Appendix A:
SDP Grammar' of [RFC2327].

Example encodings could be:

    m=audio 49170 RTP/AVP 0
    a=rtcp:53020

    m=audio 49170 RTP/AVP 0
    a=rtcp:53020 IN IP4 126.16.64.4

    m=audio 49170 RTP/AVP 0
    a=rtcp:53020 IN IP6 2001:2345:6789:ABCD:EF01:2345:6789:ABCD

The RTCP attribute MAY be used as a media level attribute; it MUST
NOT be used as a session level attribute.

3.      Discussion of the solution

The implementation of the solution is fairly straightforward. The
three questions that have been most often asked regarding this
solution are whether this is useful, i.e. whether a host can
actually discover port numbers in an unmodified NAT, whether it is
sufficient, i.e. whether or not there is a need to document more
than one ancillary port per media type, and whether relaxing the RTP
requirements is legitimate.

3.1.    How do we discover port numbers?

The proposed solution is only useful if the host can discover the
ôtranslated port numbersö, i.e. the value of the ports as they
appear on the 'external side' of the NAT. One possibility is to ask
the cooperation of a well connected third party that will act as a
server according to STUN [RFC3489]. We thus obtain a four step
process:

1- The host allocate two UDP ports numbers for an RTP/RTCP pair,

2- The host sends a UDP message from each port to the STUN server,

3- The STUN server reads the source address and port of the packet,
and copies them in the text of a reply,

4- The host parses the reply according to the STUN protocol and
learns the external address and port corresponding to each of the
two UDP port.

This algorithm supposes that the NAT will use the same translation
for packets sent to the third party and to the ôSDP peerö with which
the host wants to establish a connection. There is no guarantee that
all NAT boxes deployed on the Internet have this characteristic.
Implementers are referred to the STUN specification [RFC3489] for an
extensive discussion of the various types of NAT.

3.2.    Do we need to support multiple ports?

Most media streams are transmitted using a single pair of RTP and
RTCP ports. It is possible, however, to transmit a single media over
several RTP flows, for example using hierarchical encoding. In this
case, SDP will encode the port number used by RTP on the first flow,
and the number of flows, as in:

       m=video 49170/2 RTP/AVP 31

In this example, the media is sent over 2 consecutive pairs of
ports, corresponding respectively to RTP for the first flow (even
number, 49170), RTCP for the first flow (odd number, 49171), RTP for
the second flow (even number, 49172), and RTCP for the second flow
(odd number, 49173).

In theory, it would be possible to modify SDP and document the many
ports corresponding to the separate encoding layers. However,
layered encoding is not much used in practice, and when used is
mostly used in conjunction with multicast transmission. The
translation issues documented in this memo apply uniquely to unicast
transmission, and thus there is no short term need for the support
of multiple port descriptions. It is more convenient and more robust
to focus on the simple case in which a media is sent over exactly
one RTP/RTCP stream.

3.3.    Why not expand the media definition?

The RTP ports are documented in the media description line, and it
would seem convenient to document the RTCP port at the same place,
rather than create an RTCP attribute. We considered this design
alternative and rejected it for two reasons: adding an extra port
number and an option address in the media description would be
awkward, and more importantly it would create problems with existing
applications, which would have to reject the entire media
description if they did not understand the extension. On the
contrary, adding an attribute has a well defined failure mode:
implementations that donÆt understand the 'a=rtcp' attribute will
simply ignore it; they will fail to send RTCP packets to the
specified address, but they will at least be able to receive the
media in the RTP packets.

4.      UNSAF considerations

The RTCP attribute in SDP is used to enable establishment of
RTP/RTCP flows through NAT. This mechanism can be used in
conjunction with an address discovery mechanism such as STUN
[RFC3489]. STUN is a short term fix to the NAT traversal problem,
which requires thus consideration of the general issues linked to
'Unilateral self-address fixing' [RFC3424].

The RTCP attribute addresses a very specific problem, the
documentation of port numbers as they appear after address
translation by a port-mapping NAT. The RTCP attribute SHOULD NOT be
used for other applications.

We expect that, with time, one of two exit strategies can be
developed. The IETF may develop an explicit 'middlebox control'
protocol that will enable applications to obtain a pair of port
numbers appropriate for RTP and RTCP. Another possibility is the
deployment of IPv6, which will enable use of 'end to end' addressing
and guarantee that the two hosts will be able to use appropriate
ports. In both cases, there will be no need for documenting a ônon
standardö RTCP port with the RTCP attribute.

5.      Security Considerations

This SDP extension is not believed to introduce any significant
security risk to multi-media applications. One could conceive that a
malevolent third party would use the extension to redirect the RTCP
fraction of an RTP exchange, but this require intercepting and
rewriting the signaling packet carrying the SDP text; if an
interceptor can do that, many more attacks are available, including
a wholesale change of the addresses and port numbers at which the
media will be sent.

In order to avoid attacks of this sort, when SDP is used in a
signaling packet where it is of the form application/sdp, end-to-end
integrity using S/MIME [RFC3369] is the technical method to be
implemented and applied.  This is compatible with SIP [RFC3261].

6.      IANA Considerations

This document defines a new SDP parameter, the attribute field
'rtcp', which per [RFC2327] should be registered by IANA. This
attribute field is designed for use at media level only.

7.      Copyright

The following copyright notice is copied from RFC 2026 [Bradner,
1996], Section 10.4, and describes the applicable copyright for this
document.

Copyright (C) The Internet Society March 21, 2001. 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 implementation 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 assignees.

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.

8.      Intellectual Property

The following notice is copied from RFC 2026 [Bradner, 1996],
Section 10.4, and describes the position of the IETF concerning
intellectual property claims made against this document.

The IETF takes no position regarding the validity or scope of any
intellectual property or other rights that might be claimed to
pertain to the implementation or use other technology described in
this document or the extent to which any license under such rights
might or might not be available; neither does it represent that it
has made any effort to identify any such rights.  Information on the
IETF's procedures with respect to rights in standards-track and
standards-related documentation can be found in BCP-11.  Copies of
claims of rights made available for publication and any assurances
of licenses to be made available, or the result of an attempt made
to obtain a general license or permission for the use of such
proprietary rights by implementers or users of this specification
can be obtained from the IETF Secretariat.

The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights which may cover technology that may be required to practice
this standard.  Please address the information to the IETF Executive
Director.

9.      Acknowledgements

The original idea for using the 'rtcp' attribute was developed by
Ann Demirtjis. The draft was reviewed by the MMUSIC and AVT working
groups of the IETF.

10.     References

Normative references

[RFC2327] M. Handley, V. Jacobson, 'SDP: Session Description
Protocol', RFC 2327, April 1998.

[RTP-NEW] H. Schulzrinne, S. Casner, R. Frederick, V. Jacobson. "RTP:
A Transport Protocol for Real-Time Applications", Work in progress,
March 2003.

[RFC1889] H. Schulzrinne, S. Casner, R. Frederick, V. Jacobson. "RTP:
A Transport Protocol for Real-Time Applications", RFC 1889, January
1996.

[RFC2119] S. Bradner, 'Key words for use in RFCs to Indicate
Requirement Levels', RFC 2119, March 1997.

[RFC2234] D. Crocker, P. Overell, "Augmented BNF for Syntax
Specifications:  ABNF", RFC 2234, November 1997.

Informative references

[RFC2766] G. Tsirtsis, P. Srisuresh. 'Network Address Translation -
Protocol Translation (NAT-PT)', RFC 2766, February 2000.

[RFC3261] J. Rosenberg, H. Schulzrinne, G. Camarillo, A. Johnston,
J. Peterson, R. Sparks, M. Handley, E. Schooler. SIP: Session
Initiation Protocol. RFC 3261, June 2002.

[RFC3369] R. Housley. Cryptographic Message Syntax (CMS). RFC 3369,
August 2002.

[RFC3424] L. Daigle, "IAB considerations for UNilateral self-address
fixing (UNSAF) across network address translation," RFC 3424,
November 2002.

[RFC3489] J. Rosenberg, J. Weinberger, C. Huitema, R. Mahy. ôSTUN -
Simple Traversal of User Datagram Protocol (UDP) Through Network
Address Translators (NATs)ö. RFC 3489, March 2003

11.     Author's Address

Christian Huitema
Microsoft Corporation
One Microsoft Way
Redmond, WA 98052-6399

Email: huitema@microsoft.com