Network Working Group K. Svanbro
Request for Comments: 3409 Ericsson
Category: Informational December 2002
Lower Layer Guidelines for Robust RTP/UDP/IP Header Compression
Status of this Memo
This memo provides information for the Internet community. It does
not specify an Internet standard of any kind. Distribution of this
memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2002). All Rights Reserved.
Abstract
This document describes lower layer guidelines for robust header
compression (ROHC) and the requirements ROHC puts on lower layers.
The purpose of this document is to support the incorporation of
robust header compression algorithms, as specified in the ROHC
working group, into different systems such as those specified by
Third Generation Partnership Project (3GPP), 3GPP Project 2 (3GPP2),
European Technical Standards Institute (ETSI), etc. This document
covers only lower layer guidelines for compression of RTP/UDP/IP and
UDP/IP headers as specified in [RFC3095]. Both general guidelines
and guidelines specific for cellular systems are discussed in this
document.
Table of Contents
1. Introduction.................................................. 22. General guidelines............................................ 22.1. Error detection....................................... 22.2. Inferred header field information..................... 32.3. Handling of header size variation..................... 32.4. Negotiation of header compression parameters.......... 52.5. Demultiplexing of flows onto logical channels......... 52.6. Packet type identification............................ 52.7. Packet duplication.................................... 62.8. Packet reordering..................................... 62.9. Feedback packets...................................... 63. Cellular system specific guidelines........................... 73.1. Handover procedures................................... 73.2. Unequal error detection (UED)......................... 83.3. Unequal error protection (UEP)........................ 9
Svanbro Informational [Page 1]
RFC 3409 Lower Layer Guidelines for Robust HC December 2002
4. IANA Considerations........................................... 95. Security Considerations....................................... 96. References.................................................... 97. Author's Address..............................................108. Full Copyright Statement......................................11
Almost all header compression algorithms [RFC1144, RFC2507, RFC2508]
rely on some functionality from the underlying link layer. Headers
(compressed or not) are expected to be delivered without any residual
bit errors. IP length fields are inferred from link layer length
fields. Packet type identification may be separated from the header
compression scheme and performed at the underlying link layer.
[RFC2509], for example, elaborates on how to incorporate IP header
compression [RFC2507] in PPP [RFC1661].
It is important to be aware of such assumptions on required
functionality from underlying layers when incorporating a header
compression scheme into a system. The functionality required by a
specific header compression scheme from lower layers may also be
needed if incorporation of a header compression scheme is to be
prepared without knowing the exact details of the final scheme.
This document describes lower layer guidelines for robust RTP/UDP/IP
header compression [RFC3095] as specified by the ROHC working group.
[RFC3095] will from this point be referenced to as ROHC. These
guidelines should simplify incorporation of the robust header
compression algorithms into cellular systems like those standardized
by 3GPP, 3GPP2, ETSI, etc, and also into specific link layer
protocols such as PPP. The document should also enable preparation
of this incorporation without requiring detailed knowledge about the
final header compression scheme. Relevant standardization groups
standardizing link layers should, aided by this document, include
required functionality in "their" link layers to support robust
header compression.
Hence, this document clarifies the requirements ROHC put on lower
layers, while the requirements on ROHC may be found in [RFC3096].
All current header compression schemes [RFC1144, RFC2507, RFC2508]
rely on lower layers to detect errors in (compressed) headers. This
is usually done with link layer checksums covering at least the
Svanbro Informational [Page 2]
RFC 3409 Lower Layer Guidelines for Robust HC December 2002
compressed header. However, any error detecting mechanism may fail
to detect some bit errors, which are usually called residual bit
errors.
As for non-compressed IP packets, lower layers must provide similar
error detection, at least for ROHC headers. ROHC has been designed
not to increase the residual bit error rate (for reasonable residual
error rates) compared to the case when no header compression is used.
Headers passed up to the header decompressor should, however, have a
residual bit error probability close to zero.
A ROHC decompressor might make use of packets with erroneous headers,
even if they must be discarded. It is therefore recommended that
such invalid packets are passed up to the decompressor instead of
being discarded by lower layers, but the packet must then be
accompanied with an error indication.
Some fields of the RTP/UDP/IP headers may be classified as inferred,
that is their values are to be inferred from other values or from an
underlying link layer. A ROHC decompressor requires that at least
the following information can be inferred from any underlying link
layer:
Packet Length (IPv4) / Payload Length (IPv6)
The received packet (with compressed header) length.
Length (UDP)
This field is redundant with the Packet Length (IPv4) or the
Payload Length (IPv6) field.
In summary, all these fields relate to the length of the packet the
compressed header is included in. These fields may thus be inferred
by the decompressor if one packet length value is signaled from the
link layer to the decompressor on a per packet basis. This packet
length value should be the length of the received packet including
the (compressed) header.
It is desirable for many cellular link layer technologies that bit
rate variations and thus packet size variations are minimized.
However, there will always be some variation in compressed header
sizes since there is a trade-off between header size variations and
compression efficiency, and also due to events in the header flow and
Svanbro Informational [Page 3]
RFC 3409 Lower Layer Guidelines for Robust HC December 2002
on the channel. Variations in header sizes cause variations in
packet sizes depending on variations of payload size. The following
will only treat header size variations caused by ROHC and not packet
size variations due to variations of payload size.
The link layer must in some manner support varying header sizes from
40 bytes (full RTP/UDP/IPv4 header) or 60 bytes (full RTP/UDP/IPv6)
down to 1 byte for the minimal compressed header. It is likely that
the small compressed headers dominate the flow of headers, and that
the largest headers are sent rarely, e.g., only a few times in the
initialization phase of the header compression scheme.
Header size variations and thus packet size variations depend on
numerous factors. Unpredictable changes in the RTP, UDP or IP
headers may cause compressed headers to momentarily increase in size,
and header sizes may depend on packet loss rate at lower layers.
Header size distributions depend also on the mode ROHC operates in.
However, for e.g., a voice application, carried by RTP/UDP/IPv4, with
a constant speech frame size and silence suppression, the following
basic header size changes may be considered as typical:
In the very beginning of the speech session, the ROHC scheme is
initialized by sending full headers called IR/DYN. These are the
largest headers, with sizes depending basically on the IP-version.
For IPv4 the size is approximately 40 bytes, and for IPv6
approximately 60 bytes. The IR/DYN headers are used typically during
one round trip time, possible interleaved with compressed headers.
After that, usually only compressed headers are sent. Compressed
headers may vary in size from 1 byte up to several bytes. The
smallest compressed headers are used when there is no unpredictable
changes in header fields, typically during a talk spurt. In the
beginning of a talk spurt, compressed header sizes may increase by
one or a few bytes momentarily. Apart from increases due to new talk
spurts, compressed headers may increase in size momentarily due to
unpredictable changes in header fields.
ROHC provides some means to limit the amount of produced header
sizes. In some cases a larger header than needed may be used to
limit the number of header sizes used. Padding octets may also be
used to fill up to a desired size. Chapter 6.3 (Implementation
parameters) in [RFC3095] provides optional implementation parameters
that make it possible to mandate how a ROHC implementation should
operate, for instance to mandate how many header sizes that may be
used.
Svanbro Informational [Page 4]
RFC 3409 Lower Layer Guidelines for Robust HC December 2002
ROHC has some parameters that need to be configured in an initial
setup phase. Which header compression profiles are allowed may have
to be determined and also what kind of context identification (CID)
mechanism to use.
The lower layers supporting ROHC should thus include mechanisms for
negotiation of header compression parameters such as CID usage and
header compression profile support. In certain environments, it
might also be desirable to have mechanisms for re-negotiation of
these parameters.
The negotiation must also make sure that compressor and decompressor
use exactly the same profile, i.e. that the set of profiles available
after negotiation must not include two profile identifiers with the
same 8-bit LSB value.
For unidirectional links, this configuration might have to be
performed out-of-band or a priori, and similar methods could of
course also be used for bi-directional links if direct negotiation is
not possible.
In some cellular technologies flows are demultiplexed onto radio
bearers suitable to the particular flows, i.e., onto logically
separated channels. For instance, real-time flows such as voice and
video may be carried on logically separated bearers. It is
recommended that this kind of demultiplexing is done in the lower
layers supporting robust header compression. By doing so, the need
for context identification in the header compression scheme is
reduced. If there is a one to one mapping between flow and logical
channel, there is no need at all for context identification at the
header compression level.
Header compression schemes like [RFC2507, RFC2508] have relied on the
underlying link layer to identify different kinds of headers by means
of packet type identifiers on link layers. This kind of mechanism is
not necessarily needed for ROHC since a ROHC packet type identifier
is included in all compressed ROHC headers. Only if ROHC packets are
to be mixed with other packets, such as packets compressed by other
header compression schemes, must the link layer provide a packet type
identifier. In such cases, or if ROHC is used on top of link layers
already providing packet type identification, one (1) packet type
identifier must be reserved for identification of ROHC packets. Thus,
Svanbro Informational [Page 5]
RFC 3409 Lower Layer Guidelines for Robust HC December 2002
only one ROHC packet type is needed to mix ROHC and e.g., RFC 2507
flows, or to support ROHC on links where packet type identifiers are
already present.
Exact duplications of one and the same packet may waste transmission
resources and is in contradiction to compression. Even so, packet
duplication may occur for various reasons. Packet duplication may
also occur in different places along the path for a packet.
ROHC can handle packet duplication before the compressor but such
packet duplications should be avoided for optimal compression
efficiency. For correct ROHC operation, lower layers are not allowed
to duplicate packets on the ROHC compressor-decompressor path.
Lower layers between compressor and decompressor are assumed not to
reorder packets, i.e., the decompressor must receive packets in the
same order as the compressor sends them. ROHC handles, however,
reordering before the compression point. That is, there is no
assumption that the compressor will only receive packets in sequence.
ROHC may operate in three different modes; Unidirectional mode (U-
mode), bidirectional optimistic mode (O-mode) and bidirectional
reliable mode (R-mode). A brief description of the modes can be
found in chapter 4.4 of [RFC3095].
In U-mode it is not necessary to send any feedback from the
decompressor to the compressor. O-mode and R-mode requires however
that feedback messages from the decompressor to the compressor be
sent. Feedback messages consist of small ROHC internal packets
without any application payload. It is possible in ROHC to piggy-
back feedback packets onto regular packets with ROHC compressed
headers and payload, if there is ROHC type of compression in both the
forward and reverse direction. However, this piggy-backing may not
be desired or possible in some cases.
To support ROHC O-mode or R-mode operation, lower layers must provide
transport of feedback packets from decompressor to compressor. If
piggybacking of feedback packets is not used, lower layers must be
able to handle feedback as small stand-alone packets. For optimal
compression efficiency, feedback packets from the decompressor should
be delivered as soon as possible to the compressor.
Svanbro Informational [Page 6]
RFC 3409 Lower Layer Guidelines for Robust HC December 2002
An important group of link layer technologies where robust header
compression will be needed are future cellular systems, which may
have a very large number of users in some years. The need for header
compression is large in these kinds of systems to achieve spectrum
efficiency. Hence, it is important that future cellular systems can
efficiently incorporate the robust header compression scheme.
One cellular specific property that may affect header compression is
mobility and thus, handover (i.e., change of serving base station or
radio network controller).
The main characteristics of handovers relevant for robust header
compression are: the length of the longest packet loss event due to
handover (i.e., the number of consecutive packet losses), and
relocation of header compression context when necessary.
Depending on the location of the header compressor/decompressor in
the radio access network and the type of handover, handover may or
may not cause disruptions or packet loss events in the (compressed)
header flow relevant for the header compression scheme. For
instance, if soft handover is used and if the header
compressor/decompressor reside above the combining point for soft
handover, there will be no extra packet losses visible to the
decompressor due to handover. In hard handovers, where packet loss
events due to handover is introduced, the length of the longest
consecutive packet loss is most relevant and thus should be
minimized.
To maintain efficient ROHC operation, it should be ensured that
handover events do not cause significant long events of consecutive
packet loss. The term "significant" in this context relates to the
kind of loss tolerable for the carried real-time application.
If hard handovers are performed, which may cause significant long
events of consecutive packet loss, the radio access network should
notify the compressor when such a handover has started and completed.
The compressor could then be implemented to take proper actions and
prevent consequences from such long loss events.
Cellular systems supporting robust header compression may have
internal mechanisms for transferring the header compression context
between nodes where contexts may reside, at or before handover. If
no such mechanism for transferring header compression context between
nodes is available, the contexts may be resynchronized by the header
Svanbro Informational [Page 7]
RFC 3409 Lower Layer Guidelines for Robust HC December 2002
compression scheme itself by means of a context refresh. The header
compressor will then perform a new header compression initialization,
e.g., by sending full headers. This will, however, introduce an
increase in the average header size dependent on how often a transfer
of context is needed. To reinitialize the context in such cases, the
lower layers must indicate to the header compressor when a handover
has occurred, so that it knows when to refresh the context. Chapter
6.3 (Implementation parameters) in [RFC3095] provides optional
implementation parameters that make it possible to trigger e.g., a
complete context refresh.
Section 3.1 states that ROHC requires error detection from lower
layers for at least the compressed header. However, some cellular
technologies may differentiate the amount of error detection for
different parts of a packet. For instance, it could be possible to
have a stronger error detection for the header part of a packet, if
the application payload part of the packet is less sensitive to
errors, e.g., some cellular types of speech codes.
ROHC does not require UED from lower layers, ROHC requires only an
error detection mechanism that detects errors in at least the header
part of the packet. Thus there is no requirement on lower layers to
provide separate error detection for the header and payload part of a
packet. However, overall performance may be increased if UED is
used.
For example, if equal error detection is used in the form of one link
layer checksum covering the entire packet including both header and
payload part, any bit error will cause the packet to be discarded at
the ROHC decompressor. It is not possible to distinguish between
errors in the header and the payload part of the packet with this
error detection mechanism and the ROHC decompressor must assume that
the header is damaged, even if the bit error hit the payload part of
the packet. If the header is assumed to be damaged, it is not
possible to ensure correct decompression and that packet will thus be
discarded. If the application is such that it tolerates some errors
in the payload, it could have been better to deliver that packet to
the application and let the application judge whether the payload was
usable or not. Hence, with an unequal error detection scheme where
it is possible to separate detection of errors in the header and
payload part of a packet, more packets may be delivered to
applications in some cases for the same lower layer error rates. The
final benefit depends of course on the cost of UED for the radio
interface and related protocols.
Svanbro Informational [Page 8]
RFC 3409 Lower Layer Guidelines for Robust HC December 2002
Some cellular technologies can provide different error probabilities
for different parts of a packet, unequal error protection (UEP). For
instance, the lower layers may provide a stronger error protection
for the header part of a packet compared to the payload part of the
packet.
ROHC does not require UEP. UEP may be beneficial in some cases to
reduce the error rate in ROHC headers, but only if it is possible to
distinguish between errors in header and payload parts of a packet,
i.e., only if unequal error detection (UED) is used. The benefit of
UEP depends of course on the cost of UEP for the radio interface and
related protocols.
A protocol which follows these guidelines, e.g., [RFC3095], will
require the IANA to assign various numbers. This document by itself,
however, does not require IANA involvement.
A protocol which follows these guidelines, e.g., [RFC3095], must be
able to compress packets containing IPSEC headers according to
[RFC3096]. There may be other security aspects to consider in such
protocols. This document by itself, however, does not add security
risks.
[RFC1144] Jacobson, V., "Compressing TCP/IP Headers for Low-Speed
Serial Links", RFC 1144, February 1990.
[RFC1661] Simpson, W., Ed., "The Point-To-Point Protocol (PPP)",
STD 51, RFC 1661, July 1994.
[RFC2507] Degermark, M., Nordgren, B. and S. Pink, "IP Header
Compression", RFC 2507, February 1999.
[RFC2508] Casner, S. and V. Jacobson, "Compressing IP/UDP/RTP
Headers for Low-Speed Serial Links", RFC 2508, February
1999.
[RFC2509] Engan, M., Casner, S. and C. Bormann, "IP Header
Compression over PPP", RFC 2509, February 1999.
Svanbro Informational [Page 9]
RFC 3409 Lower Layer Guidelines for Robust HC December 2002
[RFC3095] Borman, C., Burmeister, C., Degermark, M., Fukushima, H.,
Hannu, H., Jonsson, L-E., Hakenberg, R., Koren, T., Le,
K., Liu, Z., Martensson, A., Miyazaki, A., Svanbro, K.,
Wiebke, T., Yoshimura, T. and H. Zheng, "Robust Header
Compression (ROHC)", RFC 3095, July 2001.
[RFC3096] Degermark, M., "Requirements for robust IP/UDP/RTP header
compression", RFC 3096, July 2001.
Copyright (C) The Internet Society (2002). 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 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.
Acknowledgement
Funding for the RFC Editor function is currently provided by the
Internet Society.
Svanbro Informational [Page 11]