Network Working Group                                       G. Pelletier
Request for Comments: 4164                                      Ericsson
Category: Standards Track                                    August 2005

RObust Header Compression (ROHC):
Context Replication for ROHC Profiles

Status of This Memo

   This document specifies an Internet standards track protocol for the
   Internet community, and requests discussion and suggestions for
   improvements.  Please refer to the current edition of the "Internet
   Official Protocol Standards" (STD 1) for the standardization state
   and status of this protocol.  Distribution of this memo is unlimited.

Copyright Notice

Copyright (C) The Internet Society (2005).

Abstract

   This document defines context replication, a complement to the
   context initialization procedure found in Robust Header Compression
   (ROHC), as specified in RFC 3095.  Profiles defining support for
   context replication may use the mechanism described herein to
   establish a new context based on another already existing context.
   Context replication is introduced to reduce the overhead of the
   context establishment procedure.  It may be especially useful for the
   compression of multiple short-lived flows that may be occurring
   simultaneously or near-simultaneously, such as short-lived TCP flows.

   1. Introduction ....................................................3
   2. Terminology .....................................................4
   3. Context Replication for ROHC Profiles ...........................5
      3.1. Robustness Considerations ..................................5
      3.2. Replication of Control Fields ..............................5
      3.3. Compressor States and Logic ................................6
           3.3.1. Context Replication (CR) State ......................6
           3.3.2. State Machine with Context Replication ..............7
           3.3.3. State Transition Logic ..............................7
                  3.3.3.1. Selection of Base Context, Upward
                           Transition .................................8
                  3.3.3.2. Optimistic Approach, Upward Transition .....9
                  3.3.3.3. Optional Acknowledgements (ACKs),
                           Upward Transition ..........................9
                  3.3.3.4. Negative ACKs (NACKs), Downward
                           Transition .................................9
      3.4. Decompressor Logic ........................................10
           3.4.1. Replication and Context Initialization .............10
           3.4.2. Reconstruction and Verification ....................10
           3.4.3. Actions upon Failure ...............................11
           3.4.4. Feedback Logic .....................................11
      3.5. Packet Formats ............................................11
           3.5.1. CRCs in the IR-CR Packet ...........................12
                  3.5.1.1. 7-bit CRC .................................13
                  3.5.1.2. 8-bit CRC .................................13
           3.5.2. General Format of the IR-CR Packet .................13
           3.5.3. Properties of the Base Context Identifier (BCID) ...15
   4. Security Considerations ........................................15
   5. Acknowledgements ...............................................15
   6. References .....................................................16
      6.1. Normative References ......................................16
      6.2. Informative References ....................................16
   Appendix A: General Format of the IR-CR Packet (Informative).......17
      A.1.  General Structure (Informative) ..........................17
      A.2.  Profile-Specific Replication Information (Informative) ...17
   Appendix B: Inter-Profile Context Replication (Informative)........18
      B.1.  Defining Support for Inter-Profile Context Replication ...18
      B.2.  Compatibility between Different Profiles (Informative) ...19

1. Introduction

   There is often some redundancy between header fields of different
   flows that pass through the same compressor-decompressor pair.  This
   means that some of the information needed to initialize the context
   for decompressing the headers of a new flow may already be present at
   the decompressor.  It may be desirable to reuse this information and
   remove some of the overhead normally required for the initialization
   of a new header compression context at both the compressor and
   decompressor.

   Reducing the overhead of the context establishment procedure is
   particularly useful when multiple short-lived connections (or flows)
   occur simultaneously, or near-simultaneously, between the same
   compressor-decompressor pair.  Because each new packet stream
   requires most of the header information to be sent during the
   initialization phase before smaller compressed headers can be used, a
   multitude of short-lived connections may significantly reduce the
   overall gain from header compression.

   Context replication allows some header fields, such as the IP source
   and/or destination addresses (16 octets each for IPv6), to be omitted
   within the special Initiation and Refresh (IR) packet type
   specifically defined for replication.  It also allows other fields,
   such as source and/or destination ports, to be either omitted or sent
   in a compressed form from the very first packet of the header
   compressed flow.

   Context replication is herein defined as a general ROHC mechanism.
   The benefits of context replication are not limited to any particular
   protocol and its support may be defined for any ROHC profile.

   In particular, context replication is applicable to TCP compression
   because many TCP transfers are short-lived; a behavior analysis of
   TCP/IP header fields among multiple short-lived connections may be
   found in [5].  In addition, [4] introduces considerations and
   requirements for the ROHC-TCP profile [3] to efficiently compress
   such short-lived TCP transfers.

   For profiles supporting this mechanism, the compressor performs
   context replication by reusing or creating a copy of an existing
   context, i.e., a base context, to create the replicated context.  The
   replicated context is then updated to match the header fields of the
   new flow.  The compressor then sends to the decompressor a packet
   that contains a reference to the selected base context, along with
   some data for the fields that need to be updated when creating the

   replicated context.  Finally, the decompressor creates the replicated
   context based on the reference to the base context along with the
   uncompressed and compressed data from the received packet.

   This document specifies the context replication procedure for ROHC
   profiles.  It defines the general compressor and decompressor logic
   used during context replication, as well as the general format of the
   special IR packet required for this procedure.  Profiles defining
   support for context replication must further specify the specific
   format(s) of this packet.

   The fundamentals of the ROHC framework may be found in [2].  It is
   assumed throughout this document that these are understood.

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 [1].

   This document reuses some of the terminology found in [2].  In
   addition, this document defines the following terms:

   Base context

      A base context is a context that has been validated by both the
      compressor and the decompressor.  The compressor can use a base
      context as the reference when building a new context using
      replication.

   Base CID (BCID)

      The Base Context Identifier is the CID used to identify the base
      context, from which information needed for context replication can
      be extracted.

   Context replication

      Context replication is the mechanism that initializes a new
      context based on another already existing context (a base
      context).

3. Context Replication for ROHC Profiles

   For profiles defining its support, context replication may be used as
   an alternative to the context initialization procedure found in [2].
   Note that for such profiles, only the decompressor is mandated to
   support context replication; the use of the IR-CR packet is optional
   for the compressor.

   This section describes the compressor and decompressor logic as well
   as the general format of the IR packet used with context replication.

3.1. Robustness Considerations

   Context replication deviates from the initialization procedure
   defined in [2] in that it is able to achieve a certain level of
   compression from the first packet used to initialize the context for
   a new flow.  Therefore, it is of particular importance that the
   context replication procedure be robust.  This requires that a base
   context suitable for replication be used, that the integrity of the
   initialization packet be guaranteed, and finally that the outcome of
   the replication process be verified.

   The primary mechanisms used to achieve robustness of the context
   replication procedure are the selection of the base context (based on
   prior feedback from the decompressor) and the use of checksums.
   Specifically, the compressor must obtain enough confidence that the
   base context selected for replication is valid and available at the
   decompressor before initiating the replication procedure.  Thus, the
   most reliable way to select the base context is to choose a context
   for which at least the static part to be replicated has previously
   been acknowledged by the decompressor.

   In addition, the presence of a CRC covering the information that
   initializes the context ensures the integrity of the IR header used
   for replication.  Finally, an additional CRC calculated over the
   original uncompressed header allows the decompressor to validate the
   reconstructed header and the outcome of the replication process.

3.2. Replication of Control Fields

   Control fields are fields that are either transmitted from a ROHC
   compressor to a ROHC decompressor or inferred based on the behavior
   of other fields, but are not part of the uncompressed header itself.

   They can be used to control compression and decompression behavior,
   in particular, the set of packet formats to be used.  Control fields
   are profile-specific.  Examples of such fields include the NBO and
   RND flags [2], which indicate whether the IP-ID field is in Network

   Byte Order and the type of behavior of the field, respectively.
   Another example is the parameter indicating the mode of operation
   [2].

   The IR-CR differs from the IR packet [2] in that its purpose is to
   entirely specify what part of the base context is replicated, and to
   convey the complementary information needed to create a new context.
   Because of this, a profile supporting the use of the IR-CR packet
   SHOULD define for each control field if the value of the field is
   replicated from the base context to the new context, or if its value
   is reinitialized.

   In addition, a compressor MUST NOT initiate context replication while
   a control field that is not reinitialized by replication is being
   updated, e.g., during the handshake for a mode transition [2].

3.3. Compressor States and Logic

   Compression with ROHC normally starts in the IR state, where IR
   packets must be sent to initialize a new context at the decompressor.
   IR packets include all static and non-static fields of the original
   header in uncompressed form plus some additional information.  The
   compressor stays in the IR state until it obtains confidence that the
   decompressor has received the information.

   Context replication provides an optional mechanism to complement the
   ROHC initialization procedure.  It defines a packet type, the IR
   packet for Context Replication (IR-CR), which can be used to
   initialize a new context.  Consequently, the Context Replication (CR)
   state is introduced to the compressor state machine to encompass the
   additional logic required for the use of the IR-CR packet.

   For profiles defining support for context replication, the compressor
   may thus transit directly from the IR state to the CR state if an
   already existing context can be selected as a base context for
   replication.  This effectively replaces any IR/IR-DYN packets sent
   during the context establishment procedure with an IR-CR packet.

3.3.1. Context Replication (CR) State

   The purpose of the CR state is to initialize a new context by reusing
   an already existing context.  In this state, the compressor sends a
   combination of uncompressed and compressed information, along with a
   reference to a base context plus some additional information.
   Therefore, header information pertaining to fields that are being
   replicated is not sent.

The compressor stays in the CR state until it is confident that the
decompressor has received the replication information correctly.

3.3.2. State Machine with Context Replication

   The compressor always starts in the lower compression state (IR), and
   transits to the context replication state (CR) under the constraint
   that the compressor can select a base context that is suitable for
   the flow being compressed (see also Section 3.3.3.1).

   The transition from the CR state to a higher compression state (e.g.,
   the CO state for [3]) is based on the optimistic approach principle
   or feedback received from the decompressor.

   The figure below shows the additional state for the compressor.  The
   details of the state transitions and compression logic are given in
   sub-sections following the figure.

              BCID selection       Optimistic approach / ACK
           +----->----->------+    +----->----->----->-----+
           |                  |    |                       |
           |                  v    |                       v
      +---------+          +---------+              +-------------+
      |   IR    |          |   CR    |              |   Higher    |
      |  state  |          |  state  |              | order state |
      +---------+          +---------+              +-------------+
           ^                    |
           | NACK / STATIC-NACK |
           +---<-----<-----<----+

   Note that context replication is a complement to the normal
   initialization procedure for ROHC profiles that support it.
   Therefore, the compressor transition to the CR state is an optional
   addition to the state machine, and does not affect already existing
   transitions between the IR state and higher order state(s).

3.3.3. State Transition Logic

   Decisions about transition to and from the CR state are taken by the
   compressor on the basis of:

   - availability of a base context
   - positive feedback from the decompressor (Acknowledgements -- ACKs)
   - negative feedback from the decompressor (Negative ACKs -- NACKs)
   - confidence level regarding error-free decompression of a packet

   Context replication is designed to operate over links where a
   feedback channel is available.  This is necessary to ensure that the
   information used to create a new context is synchronized between the
   compressor and the decompressor.  In addition, context replication
   may also make use of feedback from decompressor to compressor for
   transition back to the IR state and for OPTIONAL improved forward
   transition towards a state with a higher compression ratio.

   The format that must be used by all profiles for the feedback field
   within the general ROHC format is specified in Section 5.2.2 of [2];
   the feedback information is structured using two possible formats:
   FEEDBACK-1 and FEEDBACK-2.  In particular, FEEDBACK-2 can carry one
   of three possible types of feedback information: ACK, NACK, or
   STATIC-NACK.

3.3.3.1. Selection of Base Context, Upward Transition

   The compressor may initiate a transition from the IR state to the CR
   state when a suitable base context can be identified.  To perform
   this transition, the compressor selects a context that has previously
   been acknowledged by the decompressor as the base context.  The
   selected context MUST have been acknowledged by the decompressor
   using the CRC option (see also [2], Section 5.7.6.3) in the feedback
   message.  The static part of the base context to be replicated MUST
   have been acknowledged by the decompressor and the base context MUST
   be valid at replication time.

   This also implies that a compressor is not allowed to use the context
   replication mechanism if a feedback channel is not present.  However,
   note that the presence of the feedback channel cannot provide the
   guarantee that a base context selected for replication has not been
   corrupted after it has been acknowledged, or that it is still part of
   the state managed by the decompressor when the IR-CR will be
   received.

   More specifically, RFC 3095 [2] defines the context identifier (CID)
   as a reference to the state information (i.e., the context) used for
   compression and decompression.  Multiple packet streams, each having
   its own context, may thus share a channel; and the CID space along
   with its representation within packet formats may be negotiated as
   part of the channel state.  However, because RFC 3095 [2] does not
   explicitly define context state management between compressor and
   decompressor, in particular for connection-oriented flows (e.g.,
   TCP), no more than a high degree of confidence can be achieved when
   selecting a base context.

   In the case where feedback is not used by the decompressor, the
   compressor may have to periodically transit back to the IR state.  In
   such a case, the same logic applies for the transition back to the
   higher order state via the CR state: a base context, previously
   acknowledged and suitable for replication, must be re-selected.

   The criteria for whether an existing context is a suitable base
   context for replication for a new flow are left to implementations.

   Whenever the sequencing information from the last acknowledgement
   received is available, the compressor MAY use it to determine what
   fields can be replicated to avoid replicating any fields that have
   changed significantly from the state corresponding to the
   acknowledged packet.

3.3.3.2. Optimistic Approach, Upward Transition

   Transition to a higher order state can be carried out according to
   the optimistic approach principle.  This means that the compressor
   may perform an upward state transition when it is fairly confident
   that the decompressor has received enough information to correctly
   decompress packets sent according to the higher compression state.

   In general, there are many approaches where the compressor can obtain
   such information.  The compressor may obtain its confidence by
   sending several IR-CR packets with the same information.

3.3.3.3. Optional Acknowledgements (ACKs), Upward Transition

   An ACK may be sent by the decompressor to indicate that a context has
   been successfully initialized during context replication.

   Upon reception of an ACK, the compressor may assume that the context
   replication procedure was successful and transit from its initial
   state (e.g., IR state) to a higher compression state.

3.3.3.4. Negative ACKs (NACKs), Downward Transition

   A STATIC-NACK sent by the decompressor may indicate that the
   decompressor could not initialize a valid context during context
   replication, and that the corresponding context has been invalidated.

   Upon reception of a STATIC-NACK, the compressor MUST transit back to
   its initial no context state.  The compressor SHOULD also refrain
   from sending IR-CR packets using the same base context, at least
   until an acknowledgement subsequent to the reception of the

   STATIC-NACK makes this context suitable for replication (Section
   3.3.3.1).  The compressor SHOULD re-initialize the decompressor
   context using an IR packet.

   A NACK sent by the decompressor may indicate that a valid context has
   been successfully initialized but that the decompression of one or
   more subsequent packets has failed.

   Upon reception of a NACK, the compressor MAY assume that the static
   part of the decompressor context is valid, but that the dynamic part
   is invalid; the compressor may take actions accordingly.

3.4. Decompressor Logic

3.4.1. Replication and Context Initialization

   Upon reception of an IR-CR packet, the decompressor first determines
   its content ([2], Section 5.2.6).  The profile indicated in the IR-CR
   packet determines how it is to be processed.  If the CRC (8-bit CRC)
   fails to verify the packet, the packet MUST be discarded.

   If the profile as indicated in the IR-CR packet defines the use of
   the Base CID, and if its corresponding field is present within the
   packet format, this field is used to identify the base context;
   otherwise, the CID is used.

3.4.2. Reconstruction and Verification

   The decompressor creates a new context using the information present
   in the IR-CR packet together with the identified base context, and
   decompresses the original header.

   The CRC calculated over the original uncompressed header and carried
   within the profile-specific part of the IR-CR headers (7-bit CRC)
   MUST be used to verify decompression.

   When the decompression is verified and successful, the decompressor
   initializes or updates the context with the information received in
   the current header.  The decompressor SHOULD send an ACK when it
   successfully validates the context as a result of the decompression
   of one or more IR-CR packets.

   Otherwise, if the reconstructed header fails the CRC check, changes
   (either initialization or update) to the context MUST NOT be
   performed.  When the decompressor fails to validate the header,
   actions as specified in Section 3.4.3 are taken.

3.4.3. Actions upon Failure

   For profiles supporting context replication, the feedback logic of a
   decompressor is similar to the logic used for context initialization,
   as described in [2].

   Specifically, when the decompressor fails to validate the context
   following the decompression of one or more initial IR-CR packets, it
   MUST invalidate the context and remain in its initial state.  In
   addition, the decompressor SHOULD send a STATIC-NACK.  In particular,
   a decompressor implementation performing strict memory management,
   such as deleting context state information when a connection-oriented
   flow (e.g., TCP) is known to have terminated, SHOULD send STATIC-NACK
   in this case.  Otherwise, there is a risk that the compressor will
   maintain a specific CID as a potential candidate for a later
   replication attempt, while actually there is insufficient state left
   in the decompressor for this CID to act as a Base CID.

   If the context has been successfully validated from the decompression
   of one or more initial IR-CR packets, the decompressor SHOULD send a
   NACK when it fails to verify the context following the decompression
   of one or more subsequent IR-CR packets.

3.4.4. Feedback Logic

The decompressor SHOULD use the CRC option (see [2], Section 5.7.6.3)
when sending feedback corresponding to an IR or an IR-CR packet.

3.5. Packet Formats

   The format of the IR-CR packet has been designed under the following
   constraints:

   a) it must be possible to either overwrite a CID during context
      replication, or to use a different CID than the Base CID for the
      replicated context;
   b) it must be possible to selectively include or exclude from the
      packet format some fields that may be replicable;
   c) it must be possible for some fields that may be replicable to be
      represented within the packet format using either a compressed or
      an uncompressed form;
   d) it must be possible for the decompressor to verify the success of
      the replication procedure;
   e) it is anticipated that profiles, other than ROHC-TCP [3], will
      also define support for context replication.  Therefore it is
      desirable that the packet format be profile independent.

3.5.1. CRCs in the IR-CR Packet

   The IR packet, as defined in [2], is used to communicate static
   and/or dynamic parts of a context, and typically initialize the
   context.  For example, the static and dynamic chains of IR packets
   may contain an uncompressed representation of the original header.

   The IR packet format includes an 8-bit CRC, calculated over the
   initial part of the IR packet.  This CRC is meant to protect any
   information that initializes the context.  In particular, its
   coverage always includes any CID information as well as the profile
   used to interpret the remainder of the IR packet.

   The purpose of the 8-bit CRC is to ensure the integrity of the IR
   header itself.  Profiles may extend the coverage of this CRC to
   include the entire IR header, thus allowing the verification of the
   integrity of the entire uncompressed header.  However, because the
   format of the IR packet is common to all ROHC profiles and verified
   as part of the initial processing of a ROHC decompressor (see  [2],
   Section 5.2.6.), profiles may not redefine this CRC beyond the extent
   of its coverage.

   RFC 3095 [2] also defines a 3-bit CRC and a 7-bit CRC for compressed
   headers, used to verify proper decompression and validate the
   context.  This type of CRC is calculated over the original
   uncompressed header, as it is not sufficient to protect only the
   compressed data being exchanged between compressor and decompressor
   for the purpose of ensuring a robust reconstruction of the original
   header.

   Thus, there is a clear distinction in purpose between the 8-bit CRC
   found in the IR packet and the 3-bit or 7-bit CRC found in compressed
   headers.  With context replication, where the IR-CR packet may
   contain both compressed as well as uncompressed information and omit
   entirely replicable fields, this distinction in no longer present.

   Profiles supporting context replication MUST define a CRC over the
   original uncompressed header as part of the profile-specific
   information in the IR-CR packet.  This is necessary to allow a
   decompressor to verify that the replication process has succeeded.

3.5.1.1. 7-bit CRC

   The 7-bit CRC in the IR-CR packet is calculated over all octets of
   the entire original header, before replication, in the same manner as
   described in Section 5.9.2 of [2].

   The initial content of the CRC register is to be preset to all 1's.
   The CRC polynomial used for the 7-bit CRC in the IR-CR is:

      C(x) = 1 + x + x^2 + x^3 + x^6 + x^7

3.5.1.2. 8-bit CRC

   The coverage of the 8-bit CRC in the IR-CR packet is not profile
   dependent, as opposed to the ROHC IR(-DYN) packet (see [2], Sections
   5.2.3 and 5.2.4).  It MUST cover the entire packet, excluding the
   payload.  In particular, this includes the CID or any add-CID octet
   as well as the Base CID field, if present.  For profiles that define
   the usage of the Base CID within the packet format of the IR-CR as
   optional, this CRC MUST also cover the information used to indicate
   the presence of this field within the packet.

   The initial content of the CRC register is to be preset to all 1's.
   The CRC polynomial used for the 8-bit CRC in the IR-CR is:

      C(x) = 1 + x + x^2 + x^8

3.5.2. General Format of the IR-CR Packet

   The context replication mechanism requires a dedicated IR packet
   format that uniquely identifies the IR-CR packet.  This packet
   communicates the static and the dynamic parts of the replicated
   context.  It may also communicate a reference to a base context.

   With consideration to the extensibility of the IR packet type defined
   in [2], support for replication can be added using the profile-
   specific part of the IR packet.  Note that there is one bit, (x),
   left in the IR header for "Profile specific information".  The
   definition of this bit is profile specific.  Thus, profiles
   supporting context replication MAY use this bit as a flag indicating
   whether the packet is an IR packet or an IR-CR packet.  Note also
   that profiles may define an alternative method to identify the IR-CR
   packet within the profile-specific information, instead of using this
   bit.

   The IR-CR header associates a CID with a profile, and initializes the
   context using the context replication mechanism.  It is not
   recommended to use this packet to repair a damaged context.

      The IR-CR has the following general format:

        0   1   2   3   4   5   6   7
       --- --- --- --- --- --- --- ---
      :         Add-CID octet         : if for small CIDs and (CID != 0)
      +---+---+---+---+---+---+---+---+
      | 1   1   1   1   1   1   0   x | IR type octet
      +---+---+---+---+---+---+---+---+
      :                               :
      /      0-2 octets of CID        / 1-2 octets if for large CIDs
      :                               :
      +---+---+---+---+---+---+---+---+
      |            Profile            | 1 octet
      +---+---+---+---+---+---+---+---+
      |              CRC              | 1 octet
      +---+---+---+---+---+---+---+---+
      |                               |
      / Profile-specific information  / variable length
      |                               |
       - - - - - - - - - - - - - - - -
      |                               |
      /           Payload             / variable length
      |                               |
       - - - - - - - - - - - - - - - -

      x:        Profile-specific information.  Interpreted according to
                the profile indicated in the Profile field.

      Profile:  The profile to be associated with the CID.  In the IR-CR
                packet, the profile identifier is abbreviated to the 8
                least significant bits (LSBs).  It selects the highest-
                number profile in the channel state parameter PROFILES
                that matches the 8 LSBs given (see also [2]).

      CRC:      8-bit CRC computed using the polynomial of Section
                3.5.1.2.

      Profile-specific information:  The contents of this part of the
                IR-CR packet are defined by the individual profiles.
                This information is interpreted according to the profile
                indicated in the Profile field.  It MUST include a 7-bit
                CRC over the original uncompressed header using the
                polynomial of Section 3.5.1.1.  It also includes the
                static and dynamic subheader information used for
                replication; thus, which header fields are replicated
                and their respective encoding methods are outside the
                scope of this document.

Payload: The payload of the corresponding original packet, if
any.

3.5.3. Properties of the Base Context Identifier (BCID)

   The Base CID within the packet format of the IR-CR may be assigned a
   different value than the context identifier associated with the new
   flow (i.e., BCID != CID); otherwise, the base context is overwritten
   with the new context by the replication process.

   When the channel uses small CIDs, a four-bit field within the packet
   format of the IR-CR minimally represents the BCID with a value from 0
   to 15.  In particular, the four bits of Add-CID used with small CIDs
   [2] are not needed for the BCID, as this information is already
   provided by the CID of the IR-CR packet itself.  When large CIDs are
   used, the BCID is represented in the IR-CR with one or two octets,
   and it is coded in the same way as a large CID [2].

4. Security Considerations

   This document adds an alternative mechanism for ROHC profiles to
   increase the compression efficiency when initializing a new context,
   by reusing information already existing at the decompressor.  This is
   achieved by introducing new state transition logic, new feedback
   logic, and a new packet type -- all based on logic and packet formats
   already defined in RFC 3095 [2].

   In this respect, this document is not believed to bring any
   additional weakness to potential attacks to those already listed in
   [2].  However, it does increase the potential impacts of these
   attacks by creating dependencies between multiple contexts.
   Specifically, corruption of one context can fail compressor attempts
   to initialize another context at the decompressor, or to propagate to
   another context, if the compressor uses a corrupted context as a base
   for replication.

5. Acknowledgements

   The author would like to thank Richard Price, Kristofer Sandlund,
   Fredrik Lindstroem, Zhigang Liu, and HongBin Liao for valuable input,
   as well as Mark West and Lars-Erik Jonsson who also served as
   committed working group document reviewers.

6. References

6.1. Normative References

   [1]  Bradner, S., "Key words for use in RFCs to Indicate Requirement
        Levels", BCP 14, RFC 2119, March 1997.

   [2]  Bormann, 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):
        Framework and four profiles: RTP, UDP, ESP, and uncompressed",
        RFC 3095, July 2001.

6.2. Informative References

   [3]  Pelletier, G., Jonsson, L-E., Sandlund, K., and M. West, "RObust
        Header Compression (ROHC): A Profile for TCP/IP (ROHC-TCP)",
        Work in Progress, July 2005.

   [4]  Jonsson, L-E., "RObust Header Compression (ROHC): Requirements
        on TCP/IP Header Compression", RFC 4163, August 2005.

   [5]  West, M. and S. McCann, "TCP/IP Field Behavior", Work in
        Progress, October 2004.

   [6]  Finking, R. and G. Pelletier, "Formal Notation for Robust Header
        Compression (ROHC-FN)", Work in Progress, June 2005.

Appendix A: General Format of the IR-CR Packet (Informative)

A.1. General Structure (Informative)

   This section provides an example of the format of the profile-
   specific information within the general format of the IR-CR.

     0   1   2   3   4   5   6   7
   +---+---+---+---+---+---+---+---+
   |                               |
   / replication base information  / variable length
   |                               |
   +---+---+---+---+---+---+---+---+
   |                               |
   /    replication information    / variable length
   |                               |
    - - - - - - - - - - - - - - - -

   Replication base information: The contents of this part of the IR-CR
      packet are defined by the individual profiles.  This information
      is interpreted according to the profile indicated in the Profile
      field.  It MUST include a 7-bit CRC over the original uncompressed
      header using the polynomial of Section 3.4.1.1.  See Appendix A.2.

   Replication information: The contents of this part of the IR-CR
      packet are also defined by the individual profiles.  This part
      contains the static and dynamic subheader information used for
      replication.  How this information is structured is profile
      specific; profiles may define the contents of this field using a
      chain structure (static and dynamic replication chains) or by
      defining header formats for replication (e.g., ROHC-TCP [3]).

A.2. Profile-Specific Replication Information (Informative)

   This section provides a more detailed example of the possible format
   of the replication information field described in Appendix A.1:

   +---+---+---+---+---+---+---+---+
   | B |          CRC7             |  1 octet
   +---+---+---+---+---+---+---+---+
   |                               |  present if B = 1,
   /           Base CID            /  1 octet if for small CIDs, or
   |                               |  1-2 octets if for large CIDs
   +---+---+---+---+---+---+---+---+

B: B = 1 indicates that the Base CID field is present.

   CRC7:    The CRC over the original, uncompressed, header.  This 7-bit
            CRC is computed according to Section 3.4.1.1.

   Base CID: The CID identifying the base context used for replication.

Appendix B: Inter-Profile Context Replication (Informative)

   Context replication as defined in this document does not explicitly
   support the concept of context replication between profiles.
   However, it might be of interest when developing new compression
   profiles.

   Inter-profile context replication would require that the decompressor
   have access to data structures from the base context, which belongs
   to a profile different than the profile using replication.  This
   information would have to be made available in a format consistent
   with the data structures and encoding method(s) in use for all header
   fields that are being replicated.

B.1. Defining Support for Inter-Profile Context Replication

   A ROHC profile describes how to compress a specific protocol stack,
   and includes one or more sets of packet formats.  The packet formats
   will typically compress the protocol headers relative to a context of
   field values from previous headers in a flow.  This context may also
   contain some control data.  Thus, the packet formats specify a
   mapping between the uncompressed and compressed version of a protocol
   field.

   This mapping is achieved through the use of one or more encoding
   methods, which are simply functions applied to compress or decompress
   a field.  An encoding method is in turn defined using a name, a set
   of function parameters, and a formal expression (i.e., using the
   ROHC-FN [6]) or a textual description (i.e., a la RFC 3095 [2]) of
   its behaviour.

   To compress one or more fields of a specific protocol stack,
   different profiles may define their packet formats using different
   encoding methods, or using a variant of a similar technique.  A
   typical example of the latter is list compression, such as used for
   IP extension headers.  This implies that context entries for a field
   belonging to a specific protocol stack may differ in their content,
   representation, and structure from one profile to another.

   As a consequence of the above, a profile that supports context
   replication can only use a base context from another profile
   explicitly supporting the concept of a base context.  That is,
   existing profiles not supporting this concept must be updated first
   to ensure that they can export the necessary context data entries
   that use a meaningful representation during replication.

   Specifically, inter-profile context replication would require that
   decompressor implementations (including existing ones) of other
   profiles be updated when adding support for a profile that uses
   context replication.  Therefore, inter-profile context replication
   cannot be seen as an implementation-specific issue.

   The compressor must know if the decompressor supports inter-profile
   context replication before initiating the procedure.  The compressor
   must also know which contexts (belonging to which profile) may be
   used as a base context.  Therefore, a compressor cannot initiate
   context replication using a base context belonging to a different
   profile, unless that profile explicitly provides the proper mapping
   for its context entries or that profile is defined formally using
   ROHC-FN [6] in a manner that makes both profiles compatible.  The set
   of profiles negotiated for the channel (see also RFC 3095 [2]) can
   then be used to determine if a context for a specific profile can be
   used as a base context.

B.2. Compatibility between Different Profiles (Informative)

   Compatibility between profiles, when replicating a field for a
   particular protocol stack, can be expressed as follow: a field that
   is compressed by different profiles is compatible for inter-profile
   replication if it is defined in the set of packet formats using the
   same mapping function between its uncompressed and compressed
   version.

   For example, the IP Destination Address field which, based on the
   packet formats and compression strategies defined in RFC 3095 [2], is
   implicitly compressed using an encoding method equivalent to the
   static() method defined in ROHC-FN [6].

   In particular, for profiles that define their packet formats using a
   formal notation such as ROHC-FN [6], two different encoding methods
   may not have the same name.  Thus, a field from a protocol stack is
   said to be compatible for replication between two different profiles
   if it has an equivalent definition within respective packet formats.

Author's Address

   Ghyslain Pelletier
   Box 920
   Ericsson AB
   SE-971 28 Lulea, Sweden

   Phone: +46 8 404 29 43
   Fax:   +46 920 996 21
   EMail: ghyslain.pelletier@ericsson.com

Full Copyright Statement

   Copyright (C) The Internet Society (2005).

   This document is subject to the rights, licenses and restrictions
   contained in BCP 78, and except as set forth therein, the authors
   retain all their rights.

   This document and the information contained herein are provided on an
   "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
   OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
   ENGINEERING TASK FORCE DISCLAIM 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.

Intellectual Property

   The IETF takes no position regarding the validity or scope of any
   Intellectual Property Rights or other rights that might be claimed to
   pertain to the implementation or use of the technology described in
   this document or the extent to which any license under such rights
   might or might not be available; nor does it represent that it has
   made any independent effort to identify any such rights.  Information
   on the procedures with respect to rights in RFC documents can be
   found in BCP 78 and BCP 79.

   Copies of IPR disclosures made to the IETF Secretariat 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 on-line IPR repository at
   http://www.ietf.org/ipr.

   The IETF invites any interested party to bring to its attention any
   copyrights, patents or patent applications, or other proprietary
   rights that may cover technology that may be required to implement
   this standard.  Please address the information to the IETF at ietf-
   ipr@ietf.org.

Acknowledgement

Funding for the RFC Editor function is currently provided by the
Internet Society.