[Search] [txt|ps|pdf|bibtex] [Tracker] [WG] [Email] [Nits]

Versions: 00 01 02 03                                                   
Internet Engineering Task Force
Internet Draft                          Herpel-Thomson/Balabanian-Nortel
ietf-avt-rtp-mpeg4-00.txt               Basso-AT&T/Civanlar-AT&T/Hoffman-Sun
March 09, 1998                          Speer-Sun/Schulzrinne-Columbia U.
                                        Expires: September 09, 1998

        RTP payload format for MPEG-4 Elementary Streams


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 made obsolete by other documents at any time.  It is
inappropriate to use Internet-Drafts as reference material or to cite them
otherthan as ``work in progress''.

To learn the current status of any Internet-Draft, please check the
``1id- abstracts.txt'' listing contained in the Internet-Drafts Shadow
Directories on ftp.is.co.za (Africa), nic.nordu.net (Europe),
munnari.oz.au (Pacific Rim), ds.internic.net (US East Coast), or
ftp.isi.edu (US West Coast).

Distribution of this document is unlimited.


This memorandum describes a scheme to encapsulate MPEG-4 Elementary Streams
into RTP packets. Two approaches are described. The first allows to map one
MPEG-4 Elementary Stream to one RTP session, for maximum compatibility to the
design principles of RTP. The second responds to the observation that MPEG-4
applications may well consist of such a large number of streams that the
encapsulation of a bundle of Elementary Streams in one RTP session is required.

This specification is a product of the Audio/Video Transport working group
within the Internet Engineering Task Force. Comments are solicited and should
be addressed to the working group's mailing list at rem-conf@es.net and/or the

1 Introduction

MPEG-4 is a recent standard from  ISO/IEC for the coding of natural and
synthetic audio-visual data in the form of audiovisual objects that are
arranged into an audiovisual scene by means of a scene description
[1][2][3][4]. This memorandum specifies how  MPEG-4 encoded data streams are
mapped into the real-time transport protocol (RTP)[5].

ietf-avt-rtp-mpeg4-00.txt                                       [Page 1]

INTERNET-DRAFT                   - 2 -                  March 13th, 1998

This would allow the use of RTP tools to monitor MPEG-4 delivery performance,
the use of RTP mixers to combine MPEG-4 streams received from multiple
end-systems into a set of consolidated streams for multicasting and the use of
RTP translators.

1.1 Overview of MPEG-4 End-System Architecture

Figure 1 below shows the general architecture of MPEG-4 terminals. The
Compression Layer processes individual audio-visual media streams without
regard to delivery technologies. The compression schemes in MPEG-4 achieve
efficient encoding over a wide range from Kbps to multiple Mbps. The MPEG-4
compression schemes are defined in the ISO/IEC specifications 14496-2 and
14496-3 [2][3]. The media content at this layer is organized in Elementary

The MPEG-4 Systems specification, ISO/IEC 14496-1 [1], defines the concepts
needed to describe the relations between Elementary Streams in a way that
allows to create distributed, yet integrated, content presentations and to
synchronize the streams. This part of the specification is both media unaware
and delivery technology unaware.

The hierarchical relations, location and properties of Elementary Streams of a
presentation are described by a dynamic set of Object Descriptors (ODs). Each
Object Descriptor groups one or more Elementary Stream Descriptors that all
refer to a single content item (media object). Hence, multiple alternative or
hierarchical representations of each content item can be indicated.

Object Descriptors are themselves conveyed through one or more Elementary
Streams. A complete set of ODs can be seen as an MPEG-4 resource or session
description at a stream level. The resource description may itself be
hierarchical, accomplished by creating more than one Elementary Stream
conveying Object Descriptors.

The session description is accompanied by a dynamic scene description (Binary
Format for Scene, BIFS), again conveyed through one or more Elementary Streams. At this level, content is identified in terms of media objects. The
spatiotemporal location of each object is defined by BIFS. The media content of
the object, if synthetic and static, will also be described by BIFS, while
natural and animated synthetic objects may refer to an Object Descriptor that
points to one or more Elementary Streams that carry the coded representation
of the object or its animation data.

By conveying the session (or resource) description as well as the scene (or
content composition) description through their own Elementary Streams it is
made possible to change portions of the content composition and the number and
properties of media streams that carry the media content separately and
dynamically at well known instants in time.

A homogeneous encapsulation of Elementary Streams carrying media or control

ietf-avt-rtp-mpeg4-00.txt                                       [Page 2]

INTERNET-DRAFT                   - 3 -                  March 13th, 1998

(ODs, BIFS) data is defined by the Access Unit (AU) Layer that primarily
serves to convey the synchronization between streams. The AU Layer organizes
the Elementary Streams in Access Units, the smallest elements that can be
attributed individual timestamps. Integer or fractional AUs are then
encapsulated in AU Layer PDUs (AL-PDU). All consecutive data of one stream at
this layer is called an AL-packetized Elementary Stream. The interface between
the compression layer and the AU Layer is called Elementary Stream Interface
(ESI). The ESI is informative.

The Delivery Layer in MPEG-4 consists of the Delivery Multimedia Integration
Framework defined in ISO/IEC 14496-6 [4]. This layer is media unaware but
delivery technology aware. It provides transparent access to and delivery of
content irrespective of the technologies used. The interface between the AU
Layer and DMIF is called  DMIF Application Interface (DAI). It offers content
location independent procedures for establishing MPEG-4 sessions and access to
transport channels.

media aware        +-----------------------------------------+
delivery unaware   |           COMPRESSION LAYER             |
14496-2 Visual     |streams from as low as Kbps to multi-Mbps|
14496-3 Audio      +-----------------------------------------+  Elementary
media and         |              SYSTEMS LAYER                |
delivery unaware  | manages elementary streams, their synch-  |
14496-1 Systems   | ronization and hierarchical relations     |
                  +-------------------------------------------+ DMIF
delivery aware    |               DELIVERY LAYER              |
media  unaware    |provides transparent access to and delivery|
14496-6 DMIF      | of content irrespective of delivery       |
                  |                technology                 |

                Figure 1: General MPEG-4 terminal architecture

1.2 MPEG-4 Elementary Stream Data Packetization

Figure 2 below shows the MPEG-4 System data plane. For ease of explanation the
encoding side is described next. The Elementary Streams from encoders are fed
into  the Access Unit Layer  with indications of AU boundaries, Random Access
Points, Desired Composition Time and the current time.

ietf-avt-rtp-mpeg4-00.txt                                       [Page 3]

INTERNET-DRAFT                   - 4 -                  March 13th, 1998

The Access Unit Layer fragments the elementary streams into AL-PDUs each
containing a header which encodes information conveyed through the ESI. If the
AU is larger than a AL-PDU fragment then a subsequent PDU is generated with a
smaller header until the complete AU is packetized into AL-PDU fragments.

The syntax of the Access Unit Layer is not fixed but can be adapted to the
needs of the stream to be transported. This includes the possibility to select
presence or absence of individual syntax elements as well as configuration of
their length in bits. The configuration for each individual stream is conveyed
in an ALConfigDescriptor which is an integral part of the Elementary Stream
Descriptor for this stream.

The AL-packetized streams having the same priority and transport QoS
requirements (as indicated through the DAI interface signaling) may then
optionally be multi-plexed in FlexMux Streams in order to fill out the
bandwidth on the network and reduce latency for each stream.  Finally, as
figure 2 shows, the AL-packetized stream or the FlexMux stream is passed on to
any transport layer, generically called TransMux.

              E N C O D E R S  and D E C O D E R S
                       (elementary streams)
   ^    ^    ^        ^    ^        ^         ^
   v    v    v        v    v        v         v                +---++---++---+    +---++---+    +---+     +---+ Access Unit  |ISO/IEC
 |AL ||AL ||AL |    |AL ||AL |    |AL |     |AL |   Layer      |14496-1
 +---++---++---+    +---++---+    +---+     +---+              |MPEG-4
   ^    ^    ^        ^    ^        ^         ^                |Systems
   |    |    |        |    |        |         |               /
   v    v    v        v    v        v         |               +---------------+   +---------+ +-------+     |_______________|_ AL-packetized
|   FlexMux     |   | FlexMux | |FlexMux|     |                |  Stream
+---------------+   +---------+ +-------+     |  FlexMux Layer |
        ^                ^          ^         |                |
        | __ FlexMux-    |          |         |                |
        |/    Stream     |          |         |                |
        |                |          |         |                |
  +---+   +-----+   +----+   +----+  +-----+   +---+           |DMIF
  |RTP|   |MPEG2|   |AAL5|   |AAL2|  |H.223|   |DAB|   TransMux|
  |UDP|   | TS  |   |ATM |   |ATM |  |PSTN |   |mux|etc., Layer|
  |IP |   |     |   |    |   |    |  |     |   |   |           |
  +---+   +-----+   +----+   +----+  +-----+   +---+           |
    ^        ^         ^        ^       ^        ^            /
    v        v         v        v       v        v
                     (TransMux streams)
                      N E T W O R K S

ietf-avt-rtp-mpeg4-00.txt                                       [Page 4]

INTERNET-DRAFT                   - 5 -                  March 13th, 1998

            Figure 2: The MPEG-4 System Data Plane

MPEG does not restrict the permissible transport protocol stacks. Just the
properties that such a protocol stack (the TransMux Layer) should
have are specified by MPEG.

2 Encapsulation of packetized MPEG-4 data in RTP packets

2.1 Analysis of requirements

The RTP specification recommends using a separate session for each media stream.
So, a straightforward use of RTP for MPEG-4 payloads may require one RTP
session for each Elementary Stream, i.e., one or more streams per audiovisual
object. Since a typical MPEG-4 session may involve a large number of objects,
that may be as high as a few hundred, this approach is not always practical.
Allocating and controlling hundreds of destination addresses for each MPEG-4
session will pose insurmountable session administration problems. The
input/output processing overhead at the end-points will be extremely high also.
On the other hand, creating a single payload for the entire object collection
defies the highly valued object based scalability property of MPEG-4 for
Internet applications.

Therefore, one RTP payload type for MPEG-4 data should allow selective bundling
of several objects or Elementary Streams, respectively. The bundling technique
needs to address the following requirements:

a) The determination and announcement of the object bundles should be dynamic
and may be done by the sender or the receiver(s) of an MPEG-4 session through
non-RTP means. The required descriptors will be defined by DMIF and may either
be encapsulated in a DMIF-defined protocol or may be part of an SDP [7]
session description. They might be conveyed through native signaling,
using e.g. SIP [8] INVITE or RTSP [9] DESCRIBE methods.

b) Once an object bundle is defined, it is considered to be a single payload
type MPEG-4 FlexMux Stream for RTP purposes and thus identified as a single
RTP session.

Given that there are also applications with fewer objects and, hence,
Elementary Streams, a second RTP payload type for MPEG-4 data should allow
efficient encapsulation of a single Elementary Stream:

c) There need to be a payload type for MPEG-4 Single Stream.

d) In general, an AL-PDU does not contain information about its size. Therefore
an RTP packet in æMPEG-4 Single StreamÆ mode should contain exactly
one AL-PDU since, in that case, its length can be derived from the packet

ietf-avt-rtp-mpeg4-00.txt                                       [Page 5]

INTERNET-DRAFT                   - 6 -                  March 13th, 1998

length indication of the underlying protocol, usually UDP/IP.

e) Different from other RTP payload formats both payload formats do not expose
whether the payload is video or audio. This is no disadvantage since it is
assumed that subsets of MPEG-4 streams of a presentation are identified by
means of descriptive information, like stream priority and object dependencies,
conveyed in Object Descriptors and DMIF descriptors, and not by the payload

2.2 MPEG-4 Single Stream Payload Type

For this payload type a single MPEG-4 AL-PDU is mapped into one RTP packet.
For an optimized encapsulation, the redundancies between the AL-PDU header and
the RTP header are removed as far as possible.

The AL syntax is flexible and may vary from stream to stream, therefore the RTP
packer and unpacker processes need to have access to the ALConfigDescriptor for
this stream that determines the variable syntax elements.

Furthermore it is assumed that the MPEG-4 Access Unit Layer packer is aware of
the path-MTU (Maximum Transfer Unit) of the network. Hence, the AL is
responsible for splitting Access Units in appropriate AL-PDUs that do not lead
to transport packets that are larger than the path MTU size. This includes that
such an AL entity is responsible for re-packetizing Access Units when an
AL-packetized stream is read from a file that has assumed a different path MTU
size upon creation.

An MPEG-4 AL-packetized Elementary Stream that does not respect the path-MTU
of the underlying transport network and that cannot be re-packetized in an
AL-aware manner for any reason has to use the FlexMux Stream payload type that
does define means for fragmentation on the RTP layer.

Note: An AU corresponds to Application Data Unit in ALF. An AL-PDU which
carries less than a complete AU may be a stupid fragment made for the sole
purpose to adapt packet size to some network requirements. Therefore AL-PDUs
never have to be fragmented, since the AL packer (that is closely linked to the RTP packer for the purpose of our work here) is path MTU aware. The fallback
solution is expressed in the previous paragraph.

The Single Stream Payload type eliminates redundant transmission of sequence
numbers and timestamps between RTP header and AL-PDU header. In conjunction
with RTP header compression [6] it therefore allows relatively low overhead

2.2.1 RTP Header usage

The payload type (PT) shall be set to the value identified for MPEG-4 Single

ietf-avt-rtp-mpeg4-00.txt                                       [Page 6]

INTERNET-DRAFT                   - 7 -                  March 13th, 1998

The marker (M) bit shall be set to one for each AL-PDU that is the last AL-PDU
of a fragmented Access Unit. Such an AL-PDU is identified either by:

        accessUnitEndFlag = 1, if this flag is present in the header, or
        looking ahead that the subsequent AL-PDU has accessUnitStartFlag=1 or
        knowing from ALConfigDescriptor that each AL-PDU contains a complete
        Access Unit.

Note: M is used as indicator for fragmentation that occurs on the AU Layer.
This may be seen as not being strictly the same as the original RTP intention
to use it to signal the end of a presentation unit (e.g. a video frame).
However, for the purpose of this document an AU should be seen as presentation

The extension (X) bit shall be set to zero.

The sequence number shall be derived  from the value encountered in the
sequenceNumber field of the AL-PDU, if present, incremented by a constant
random offset. If sequenceNumber has less than 16 bit length, the MSBs shall
initially be filled with a random value that is incremented by one each time
the sequenceNumber value of the AL-PDU returns to zero. As an exception, the
MSBs shall not be incremented if the value sequenceNumber=0 is encountered in
multiple consecutive AL-PDUs since this indicates a deliberate duplication of
the AL-PDU. The sequenceNumber shall then be removed from the AL-PDU header by
bit-shifting the subsequent header elements towards the beginning of the
AL-PDU header. When unpacking the RTP packet this process can be reversed with
the knowledge of the ALConfigDescriptor.  If no sequenceNumber field is
configured for this stream, acc. to the ALConfigDescriptor, then the RTP
packer shall generate its own sequence numbers, starting at a random value.

The timestamp shall be set to the value encountered in the compositionTimeStamp
field of the AL-PDU, if present. If compositionTimeStamp has less than 32 bits
length, the MSBs of timestamp shall be set to zero. If compositionTimeStamp has
more than 32 bits length, MPEG-4 Single Stream payload type can not be used
and MPEG-4 FlexMux Stream must be used instead for this stream.  In case
compositionTimeStamp is not present in this AL-PDU, but has been present
in a previous AL-PDU, this same value shall be taken again as timestamp.
The compositionTimeStamp, if present, shall then be removed from the AL-PDU
header by bit-shifting the subsequent header elements towards the beginning of
the AL-PDU header. When unpacking the RTP packet this process can be reversed
with the knowledge of the ALConfigDescriptor and by evaluating the
compositionTimeStampFlag.  The AL-PDU header also allows a decodingTimeStamp
to be present, if the decoding time is different from the composition time. For
those AL-PDUs in which only a decodingTimeStamp or both, composition and
decodingTimeStamp are present in the AL-PDU header the above procedure shall
be applied to the decodingTimeStamp, leaving the compositionTimeStamp
unaffected.  If compositionTimeStamp is never present in AL-PDUs for this
stream, acc. to the ALConfigDescriptor, the RTP packer shall convey a

ietf-avt-rtp-mpeg4-00.txt                                       [Page 7]

INTERNET-DRAFT                   - 8 -                  March 13th, 1998

reading of a local clock at the time the RTP packet is sent.

Note: Timestamps should start at a random value for security reasons. Since
they are copied from the AL, this would involve changing the remaining
timestamps in the AL-PDU Header as well. This includes changing OCR timestamps
that may have a different accuracy. Therefore it seems to be difficult to
satisfy the random start value constraint.

An SSRC value for each RTP session shall be selected randomly, as described in
RFC1889 [5].

2.2.2 RTP Payload

The RTP Payload consists of one single AL-PDU, including the AL-PDU header,
omitting the sequenceNumber and decodingTimeStamp fields, as specified in the
previous section. If the resulting, smaller, AL-PDU header consumes a
non-integer number of bytes, zero padding bits shall be inserted to byte-align
the AL-PDU payload.

Note: 32 bit alignment of AL-PDU payload would be desirable for processing
efficiency with high bitrate streams, while for low bitrate streams
unfortunately it just increases the overhead.

    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
   |V=2|P|X|  CC   |M|     PT      |       sequence number         |  RTP
   |                           timestamp                           |  Header
   |           synchronization source (SSRC) identifier            |
   :            contributing source (CSRC) identifiers             :
   :                             ....                              :
   |AL-PDU Header (variable # of bytes)            |               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+               |  RTP
   |                                                               |
   |       AL-PDU Payload (byte aligned)                           |  Payload
   |                                                               |
   |                                                               |
   |                               +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                               :...........RTP padding (opt.)  |

The operation of the RTP unpacker for this payload type can be visualized by
the figure below. Note that an implementation may as well integrate RTP
parsing and AL-PDU regeneration in one unit.

ietf-avt-rtp-mpeg4-00.txt                                       [Page 8]

INTERNET-DRAFT                   - 9 -                  March 13th, 1998

  Natural/Synthetic Audio & Visual data,
     Scene Description & OD streams

  ^     ^     ^                     ^
  |     |     |                     |
  |     |     |                     |
+---+ +---+ +---+                 +---+
|AL | |AL | |AL |.  .  .  .  .  . |AL |
+---+ +---+ +---+                 +---+
  ^    ^      ^                     ^
  |    |      |                     |
  |    |      |                     |
|        Regenerate AL-PDUs            |
 MPEG-4  ^   Time  ^ Sequence^
Payloads |  Stamps |  Numbers|
         |         |         |
       |      RTP Parser       |

2.3 MPEG-4 FlexMux Stream Payload Type

This payload type is intended for use if the number of Elementary Streams in
an MPEG-4 application are too high for the Single Stream payload type to be
efficient, i.e., the number of RTP packets per second as well as the associated
processing for the high number of RTP sessions becomes unmanageable. In this
case multiple MPEG-4 Elementary Streams have to be mapped into one RTP session.

Therefore this payload type defines how data from multiple streams can be
aggregated in one RTP packet, encapsulated in MPEG-4 FlexMux-PDUs as
specified in [1]. Notably this enables the use of FlexMux MuxCode mode that
allows to merge AL-PDUs from different streams in a pre-agreed sequence and,
hence, does not need any further overhead to label these AL-PDUs.

This payload type is intended to carry an integer number of FlexMux-PDUs per
RTP packet, however, it also defines a means to support fragmentation of large
FlexMux-PDUs in case the MPEG-4 Access Unit Layer packer is not aware of the
path-MTU of the network.

2.3.1 RTP Header usage

ietf-avt-rtp-mpeg4-00.txt                                       [Page 9]

INTERNET-DRAFT                   - 10 -                 March 13th, 1998

The payload type (PT) shall be set to the value identified for MPEG-4 FlexMux

The marker (M) bit shall be zero if the payload starts with the beginning of a
FlexMux-PDU. It shall be one if the payload contains the continuation of a
FlexMux-PDU. The FlexMux-PDU length indication allows to know whether such an
RTP packet contains the last fragment of a fragmented FlexMux-PDU.

The extension (X) bit shall be set to zero.

The RTP packer shall generate sequence numbers, incrementing monotonically by
one, for each RTP packet generated in order. For the initial RTP packet of a
session a random value shall be chosen for sequence number.  If two packets
with the same sequence number are received immediately after each other, a
duplication of the packet shall be assumed and the later packet may be
discarded.  If the sequence numbers are discontinuous, packet loss shall be
assumed as soon as the timestamp of the packet after the discontinuity
requires delivery of that packet. The MPEG-4 Elementary Streams encapsulated
in the FlexMux Stream that forms the RTP payload need to implement their
own AL-PDU loss detection using the sequenceNumber syntax element or need to
rely on the error resiliency of the compression scheme used.

The timestamp shall convey a reading of a local clock at the time the RTP
packet is sent.

An SSRC value for each RTP session shall be selected randomly, as described in
RFC1889 [5].

2.3.2 RTP Payload

The RTP payload consists of one or more complete FlexMux-PDUs as visualized in
the figure below. Each FlexMux-PDU consists of an index element that
identifies the content of the FlexMux-PDU, the length of the payload, a
version number (only for MuxCode mode) and the payload itself as specified in
[1]. FlexMux-PDUs are byte-aligned in the payload of the RTP packet.

Note: It may be computationally beneficial to word-align the FlexMux-PDUs.
However, on average this means a 2 byte per FlexMux-PDU overhead for alignment.
That seems to be rather high.

    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
   |V=2|P|X|  CC   |M|     PT      |       sequence number         |  RTP
   |                           timestamp                           |  Header

ietf-avt-rtp-mpeg4-00.txt                                      [Page 10]

INTERNET-DRAFT                   - 11 -                 March 13th, 1998

   |           synchronization source (SSRC) identifier            |
   :            contributing source (CSRC) identifiers             :
   :                             ....                              :
   |    index      |   length      | version       :FM-PDU Payload |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+...............+               |  RTP
   |     ......           .......          ........      .......   |
   |               +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+  Payload
   |     ......    |     index     |   length      |     .......   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+               |
   :     ......       FM-PDU Payload    ........      .........    :
   :                                                               :
   |                               +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |     ......      ........      :...........RTP padding (opt.)  |

The operation of the RTP unpacker for this payload type can be visualized by
the figure below. Note that an implementation may as well integrate RTP
parsing and FlexDemulitplexing in one unit.

  Natural/Synthetic Audio & Visual data,
     Scene Description & OD streams

  ^     ^     ^                     ^
  |     |     |                     |
  |     |     |                     |
+---+ +---+ +---+                 +---+
|AL | |AL | |AL |.  .  .  .  .  . |AL |
+---+ +---+ +---+                 +---+
  ^    ^      ^                     ^
  |    |      |                     |
  |    |      |                     |
|     FlexMux demultiplexer            |
        RTP packet ^
        payload    | includes signaling
                   | of packet boundaries
       |      RTP Parser       |

ietf-avt-rtp-mpeg4-00.txt                                      [Page 11]

INTERNET-DRAFT                   - 12 -                 March 13th, 1998


A Authors' Addresses

  Carsten Herpel                             Vahe Balabanian
  THOMSON multimedia                         Nortel
  Goettinger Chaussee 76                     P.O.Box 3511, St. C
  30453 Hannover                             Ottawa, Ontario
  Germany                                    Canada K1Y 4H7
  Email: herpelc@thmulti.com                 Email: balabani@nortel.ca

  Andrea Basso                               M. Reha Civanlar
  AT&T Labs - Research                       AT&T Labs - Research
  100 Schultz Drive,                         100 Schultz Drive,
  Red Bank, NJ 07701                         Red Bank, NJ 07701
  USA                                        USA
  Email: basso@research.att.com              Email: civanlar@research.att.com

  Don Hoffman                                Michael F. Speer
  Sun Microsystems, Inc                      Sun Microsystems, Inc
  901 San Antonio Road                       901 San Antonio Road
  Palo Alto, CA 94303-4900                   Palo Alto, CA 94303-4900
  USA                                        USA
  Email: don.hoffman@eng.sun.com             Email: michael.speer@eng.sun.com

  Henning Schulzrinne
  Columbia University
  Dept. of Computer Science
  1214 Amsterdam Avenue
  New York, NY 10027
  Email: schulzrinne@cs.columbia.edu

B Bibliography

[1] ISO/IEC 14496-1 CD MPEG-4 Systems Oct. 1997

[2] ISO/IEC 14496-2 CD MPEG-4 Visual Oct. 1997

[3] ISO/IEC 14496-3 CD MPEG-4 Audio Oct. 1997

[4] ISO/IEC 14496-6 CD Delivery Multimedia Integration Framework, Oct. 1997

ietf-avt-rtp-mpeg4-00.txt                                      [Page 12]

INTERNET-DRAFT                   - 13 -                 March 13th, 1998

[5] Schulzrinne, Casner, Frederick, Jacobson RTP: A Transport Protocol
        for Real Time Applications  RFC 1889, Internet Engineering Task
        Force, Jan. 1996.

[6] S. Casner, V. Jacobsen Compressing IP/UDP/RTP Headers for Low-Speed
        Serial Links, draft-ietf-avt-crtp-04, Internet Engineering Task Force,
        Nov. 1997

[7] M. Handley, V. Jacobson SDP: Session Description Protocol, Internet
        Engineering Task Force, Nov. 1997

[8] Handley, Schulzrinne, Schooler SIP: Session Initiation Protocol, Internet
        Engineering Task Force, Nov. 1997

[9] H. Schulzrinne, A. Rao, R. Lamphier Real Time Streaming Protocol
        (RTSP), Internet Engineering Task Force, Nov. 1997

ietf-avt-rtp-mpeg4-00.txt                                      [Page 13]