Internet Engineering Task Force (IETF)                       A. Minaburo
Request for Comments: 8724                                        Acklio
Category: Standards Track                                     L. Toutain
ISSN: 2070-1721                                           IMT Atlantique
                                                                C. Gomez
                                    Universitat Politecnica de Catalunya
                                                              D. Barthel
                                                             Orange Labs
                                                              JC. Zuniga
                                                                  SIGFOX
                                                              March
                                                              April 2020

   SCHC: Generic Framework for Static Context Header Compression and
                             Fragmentation (SCHC)

Abstract

   This document defines the Static Context Header Compression and
   fragmentation (SCHC) framework, which provides both a header
   compression mechanism and an optional fragmentation mechanism.  SCHC
   has been designed with Low-Power Wide Area Networks (LPWANs) in mind.

   SCHC compression is based on a common static context stored both in
   the LPWAN device and in the network infrastructure side.  This
   document defines a generic header compression mechanism and its
   application to compress IPv6/UDP headers.

   This document also specifies an optional fragmentation and reassembly
   mechanism.  It can be used to support the IPv6 MTU requirement over
   the LPWAN technologies.  Fragmentation is needed for IPv6 datagrams
   that, after SCHC compression or when such compression was not
   possible, still exceed the Layer 2 maximum payload size.

   The SCHC header compression and fragmentation mechanisms are
   independent of the specific LPWAN technology over which they are
   used.  This document defines generic functionalities and offers
   flexibility with regard to parameter settings and mechanism choices.
   This document standardizes the exchange over the LPWAN between two
   SCHC entities.  Settings and choices specific to a technology or a
   product are expected to be grouped into profiles, which are specified
   in other documents.  Data models for the context and profiles are out
   of scope.

Status of This Memo

   This is an Internet Standards Track document.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Further information on
   Internet Standards is available in Section 2 of RFC 7841.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   https://www.rfc-editor.org/info/rfc8724.

Copyright Notice

   Copyright (c) 2020 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (https://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.

Table of Contents

   1.  Introduction
   2.  Requirements Notation
   3.  LPWAN Architecture
   4.  Terminology
   5.  SCHC Overview
     5.1.  SCHC Packet Format
     5.2.  Functional Mapping
   6.  RuleID
   7.  Compression/Decompression
     7.1.  SCHC C/D Rules
     7.2.  Packet Processing
     7.3.  Matching Operators
     7.4.  Compression/Decompression Actions (CDA)
       7.4.1.  Processing Fixed-Length Fields
       7.4.2.  Processing Variable-Length Fields
       7.4.3.  Not-Sent CDA
       7.4.4.  Value-Sent CDA
       7.4.5.  Mapping-Sent CDA
       7.4.6.  LSB CDA
       7.4.7.  DevIID, AppIID CDA
       7.4.8.  Compute-*
   8.  Fragmentation/Reassembly
     8.1.  Overview
     8.2.  SCHC F/R Protocol Elements
       8.2.1.  Messages
       8.2.2.  Tiles, Windows, Bitmaps, Timers, Counters
       8.2.3.  Integrity Checking
       8.2.4.  Header Fields
     8.3.  SCHC F/R Message Formats
       8.3.1.  SCHC Fragment Format
       8.3.2.  SCHC ACK Format
       8.3.3.  SCHC ACK REQ Format
       8.3.4.  SCHC Sender-Abort Format
       8.3.5.  SCHC Receiver-Abort Format
     8.4.  SCHC F/R Modes
       8.4.1.  No-ACK Mode
       8.4.2.  ACK-Always Mode
       8.4.3.  ACK-on-Error Mode
   9.  Padding Management
   10. SCHC Compression for IPv6 and UDP Headers
     10.1.  IPv6 Version Field
     10.2.  IPv6 Traffic Class Field
     10.3.  Flow Label Field
     10.4.  Payload Length Field
     10.5.  Next Header Field
     10.6.  Hop Limit Field
     10.7.  IPv6 Addresses Fields
       10.7.1.  IPv6 Source and Destination Prefixes
       10.7.2.  IPv6 Source and Destination IID
     10.8.  IPv6 Extension Headers
     10.9.  UDP Source and Destination Ports
     10.10. UDP Length Field
     10.11. UDP Checksum Field
   11. IANA Considerations
   12. Security Considerations
     12.1.  Security Considerations for SCHC Compression/Decompression
       12.1.1.  Forged SCHC Packet
       12.1.2.  Compressed Packet Size as a Side Channel to Guess a
               Secret Token
       12.1.3.  Decompressed Packet Different from the Original Packet
     12.2.  Security Considerations for SCHC Fragmentation/Reassembly
       12.2.1.  Buffer Reservation Attack
       12.2.2.  Corrupt Fragment Attack
       12.2.3.  Fragmentation as a Way to Bypass Network Inspection
       12.2.4.  Privacy Issues Associated with SCHC Header Fields
   13. References
     13.1.  Normative References
     13.2.  Informative References
   Appendix A.  Compression Examples
   Appendix B.  Fragmentation Examples
   Appendix C.  Fragmentation State Machines
   Appendix D.  SCHC Parameters
   Appendix E.  Supporting Multiple Window Sizes for Fragmentation
   Appendix F.  ACK-Always and ACK-on-Error on Quasi-Bidirectional
           Links
   Acknowledgements
   Authors' Addresses

1.  Introduction

   This document defines the Static Context Header Compression and
   fragmentation (SCHC) framework, which provides both a header
   compression mechanism and an optional fragmentation mechanism.  SCHC
   has been designed with Low-Power Wide Area Networks (LPWANs) in mind.

   LPWAN technologies impose some strict limitations on traffic.  For
   instance, devices sleep most of the time and may only receive data
   during short periods of time after transmission, in order to preserve
   battery.  LPWAN technologies are also characterized by a greatly
   reduced data unit and/or payload size (see [RFC8376]).

   Header compression is needed for efficient Internet connectivity to a
   node within an LPWAN.  The following properties of LPWANs can be
   exploited to get an efficient header compression:

   *  The network topology is star-oriented, which means that all
      packets between the same source-destination pair follow the same
      path.  For the needs of this document, the architecture can simply
      be described as Devices (Dev) exchanging information with LPWAN
      Application Servers (Apps) through a Network Gateway (NGW).

   *  Because devices embed built-in applications, the traffic flows to
      be compressed are known in advance.  Indeed, new applications are
      less frequently installed in an LPWAN device than they are in a
      general-purpose computer or smartphone.

   SCHC compression uses a Context (a set of Rules) in which information
   about header fields is stored.  This Context is static: the values of
   the header fields and the actions to do compression/decompression do
   not change over time.  This avoids the need for complex
   resynchronization mechanisms.  Indeed, a return path may be more
   restricted/expensive, or may sometimes be completely unavailable
   [RFC8376].  A compression protocol that relies on feedback is not
   compatible with the characteristics of such LPWANs.

   In most cases, a small Rule identifier is enough to represent the
   full IPv6/UDP headers.  The SCHC header compression mechanism is
   independent of the specific LPWAN technology over which it is used.

   Furthermore, some LPWAN technologies do not provide a fragmentation
   functionality; to support the IPv6 MTU requirement of 1280 bytes
   [RFC8200], they require a fragmentation protocol at the adaptation
   layer below IPv6.  Accordingly, this document defines an optional
   fragmentation/reassembly mechanism to help LPWAN technologies support
   the IPv6 MTU requirement.

   This document defines generic functionality and offers flexibility
   with regard to parameter settings and mechanism choices.  Technology-
   specific settings are expected to be grouped into Profiles specified
   in other documents.

2.  Requirements Notation

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
   "OPTIONAL" in this document are to be interpreted as described in
   BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all
   capitals, as shown here.

3.  LPWAN Architecture

   LPWAN architectures are similar among them, but each LPWAN technology
   names architecture elements differently.  In this document, we use
   terminology from [RFC8376], which identifies the following entities
   in a typical LPWAN (see Figure 1):

   *  Devices (Dev) are the end-devices or hosts (e.g., sensors,
      actuators, etc.).  There can be a very high density of devices per
      Radio Gateway.

   *  The Radio Gateway (RGW) is the endpoint of the constrained link.

   *  The Network Gateway (NGW) is the interconnection node between the
      Radio Gateway and the Internet.

   *  The Application Server (App) is the endpoint of the application-
      level protocol on the Internet side.

    ()   ()   ()       |
     ()  () () ()     / \       +---------+
   () () () () () () /   \======|    ^    |             +-----------+
    ()  ()   ()     |           | <--|--> |             |Application|
   ()  ()  ()  ()  / \==========|    v    |=============|   Server  |
     ()  ()  ()   /   \         +---------+             +-----------+
    Dev            RGWs             NGW                      App

   Figure 1: LPWAN Architecture (Simplified from That Shown in RFC 8376)

4.  Terminology

   This section defines terminology and abbreviations used in this
   document.  It extends the terminology of [RFC8376].

   The SCHC acronym is pronounced like "sheek" in English (or "chic" in
   French).  Therefore, this document writes "a SCHC Packet" instead of
   "an SCHC Packet".

   App:     LPWAN Application Server, as defined by [RFC8376].  It runs
            an application sending/receiving packets to/from the Dev.

   AppIID:  Application Interface Identifier.  The IID that identifies
            the App interface.

   Compression Residue:  The bits that remain to be sent (beyond the
            RuleID itself) after applying the SCHC compression.

   Context:  A set of Rules used to compress/decompress headers, or to
            fragment/reassemble a packet.

   Dev:     Device, as defined by [RFC8376].

   DevIID:  Device Interface Identifier.  The IID that identifies the
            Dev interface.

   Downlink:  From the App to the Dev.

   IID:     Interface Identifier.  See the IPv6 addressing architecture
            [RFC7136].

   L2:      Layer 2.  The immediate lower layer that SCHC interfaces
            with, for example an underlying LPWAN technology.  It does
            not necessarily correspond to the OSI model definition of
            Layer 2.

   L2 Word:  This is the minimum subdivision of payload data that the L2
            will carry.  In most L2 technologies, the L2 Word is an
            octet.  In bit-oriented radio technologies, the L2 Word
            might be a single bit.  The L2 Word size is assumed to be
            constant over time for each device.

   Padding:  Extra bits that may be appended by SCHC to a data unit that
            it passes down to L2 for transmission.  SCHC itself operates
            on bits, not bytes, and does not have any alignment
            prerequisite.  See Section 9.

   Profile:  SCHC offers variations in the way it is operated, with a
            number of parameters listed in Appendix D.  A Profile
            indicates a particular setting of all these parameters.
            Both ends of a SCHC communication must be provisioned with
            the same Profile information and with the same set of Rules
            before the communication starts, so that there is no
            ambiguity in how they expect to communicate.

   Rule:    Part of the Context that describes how a packet is
            compressed/decompressed or fragmented/reassembled.

   RuleID:  Rule Identifier.  An identifier for a Rule.

   SCHC:    Static Context Header Compression and fragmentation (SCHC),
            a generic framework.

   SCHC C/D:  SCHC Compressor/Decompressor, or SCHC Compression/
            Decompression.  The SCHC entity or mechanism used on both
            sides, at the Dev and at the network, to achieve
            compression/decompression of headers.

   SCHC F/R:  SCHC Fragmenter/Reassembler or SCHC Fragmentation/
            Reassembly.  The SCHC entity or mechanism used on both
            sides, at the Dev and at the network, to achieve
            fragmentation/reassembly of SCHC Packets.

   SCHC Packet:  A packet (e.g., an IPv6 packet) whose header has been
            compressed as per the header compression mechanism defined
            in this document.  If the header compression process is
            unable to actually compress the packet header, the packet
            with the uncompressed header is still called a SCHC Packet
            (in this case, a RuleID is used to indicate that the packet
            header has not been compressed).  See Section 7 for more
            details.

   Uplink:  From the Dev to the App.

   Additional terminology for the optional SCHC F/R is found in
   Section 8.2.

   Additional terminology for SCHC C/D is found in Section 7.1.

5.  SCHC Overview

   SCHC can be characterized as an adaptation layer between an upper
   layer (for example, IPv6) and an underlying layer (for example, an
   LPWAN technology).  SCHC comprises two sublayers (i.e., the
   Compression sublayer and the Fragmentation sublayer), as shown in
   Figure 2.

                +----------------+
                |      IPv6      |
             +- +----------------+
             |  |   Compression  |
       SCHC <   +----------------+
             |  |  Fragmentation |
             +- +----------------+
                |LPWAN technology|
                +----------------+

     Figure 2: Example of Protocol Stack Comprising IPv6, SCHC, and an
                              LPWAN Technology

   Before an upper layer packet (e.g., an IPv6 packet) is transmitted to
   the underlying layer, header compression is first attempted.  The
   resulting packet is called a "SCHC Packet", whether or not any
   compression is performed.  If needed by the underlying layer, the
   optional SCHC fragmentation MAY be applied to the SCHC Packet.  The
   inverse operations take place at the receiver.  This process is
   illustrated in Figure 3.

   A packet (e.g., an IPv6 packet)
            |                                           ^
            v                                           |
   +------------------+                      +--------------------+
   | SCHC Compression |                      | SCHC Decompression |
   +------------------+                      +--------------------+
            |                                           ^
            |   If no fragmentation (*)                 |
            +-------------- SCHC Packet  -------------->|
            |                                           |
            v                                           |
   +--------------------+                       +-----------------+
   | SCHC Fragmentation |                       | SCHC Reassembly |
   +--------------------+                       +-----------------+
         |     ^                                     |     ^
         |     |                                     |     |
         |     +---------- SCHC ACK (+) -------------+     |
         |                                                 |
         +-------------- SCHC Fragments -------------------+

           Sender                                    Receiver

   *: the decision not to use SCHC fragmentation is left to each Profile
   +: optional, depends on Fragmentation mode

          Figure 3: SCHC Operations at the Sender and the Receiver

5.1.  SCHC Packet Format

   The SCHC Packet is composed of the Compressed Header followed by the
   payload from the original packet (see Figure 4).  The Compressed
   Header itself is composed of the RuleID and a Compression Residue,
   which is the output of compressing the packet header with the Rule
   identified by that RuleID (see Section 7).  The Compression Residue
   may be empty.  Both the RuleID and the Compression Residue
   potentially have a variable size, and are not necessarily a multiple
   of bytes in size.

   |------- Compressed Header -------|
   +---------------------------------+--------------------+
   |  RuleID  |  Compression Residue |      Payload       |
   +---------------------------------+--------------------+

                           Figure 4: SCHC Packet

5.2.  Functional Mapping

   Figure 5 maps the functional elements of Figure 3 onto the LPWAN
   architecture elements of Figure 1.

           Dev                                               App
   +----------------+                               +----+ +----+ +----+
   | App1 App2 App3 |                               |App1| |App2| |App3|
   |                |                               |    | |    | |    |
   |       UDP      |                               |UDP | |UDP | |UDP |
   |      IPv6      |                               |IPv6| |IPv6| |IPv6|
   |                |                               |    | |    | |    |
   |SCHC C/D and F/R|                               |    | |    | |    |
   +--------+-------+                               +----+ +----+ +----+
            |  +---+     +---+    +----+    +----+    .      .      .
            +~ |RGW| === |NGW| == |SCHC| == |SCHC|..... Internet ....
               +---+     +---+    |F/R |    |C/D |
                                  +----+    +----+

                      Figure 5: Architectural Mapping

   SCHC C/D and SCHC F/R are located on both sides of the LPWAN
   transmission, hereafter called the "Dev side" and the "Network
   Infrastructure side".

   The operation in the Uplink direction is as follows.  The Device
   application uses IPv6 or IPv6/UDP protocols.  Before sending the
   packets, the Dev compresses their headers using SCHC C/D; if the SCHC
   Packet resulting from the compression needs to be fragmented by SCHC,
   SCHC F/R is performed (see Section 8).  The resulting SCHC Fragments
   are sent to an LPWAN Radio Gateway (RGW), which forwards them to a
   Network Gateway (NGW).  The NGW sends the data to a SCHC F/R for
   reassembly (if needed) and then to the SCHC C/D for decompression.
   After decompression, the packet can be sent over the Internet to one
   or several Apps.

   The SCHC F/R and SCHC C/D on the Network Infrastructure side can be
   part of the NGW or located in the Internet as long as a tunnel is
   established between them and the NGW.  For some LPWAN technologies,
   it may be suitable to locate the SCHC F/R functionality nearer the
   NGW, in order to better deal with time constraints of such
   technologies.

   The SCHC C/Ds on both sides MUST share the same set of Rules.  So
   MUST the SCHC F/Rs on both sides.

   The operation in the Downlink direction is similar to that in the
   Uplink direction, only reversing the order in which the architecture
   elements are traversed.

6.  RuleID

   RuleIDs identify the Rules used for compression/decompression or for
   fragmentation/reassembly.

   The scope of the RuleID of a compression/decompression Rule is the
   link between the SCHC C/D in a given Dev and the corresponding SCHC
   C/D in the Network Infrastructure side.  The scope of the RuleID of a
   fragmentation/reassembly Rule is the link between the SCHC F/R in a
   given Dev and the corresponding SCHC F/R in the Network
   Infrastructure side.  If such a link is bidirectional, the scope
   includes both directions.

   The RuleIDs are therefore specific to the Context related to one Dev.
   Hence, multiple Dev instances, which refer to different Contexts, MAY
   reuse the same RuleID for different Rules.  On the Network
   Infrastructure side, in order to identify the correct Rule to be
   applied to Uplink traffic, the SCHC C/D or SCHC F/R needs to
   associate the RuleID with the Dev identifier.  Similarly, for
   Downlink traffic, the SCHC C/D or SCHC F/R on the Network
   Infrastructure side first needs to identify the destination Dev
   before looking for the appropriate Rule (and associated RuleID) in
   the Context of that Dev.

   Inside their scopes, Rules for compression/decompression and Rules
   for fragmentation/reassembly share the same RuleID space.

   The size of the RuleIDs is not specified in this document, as it is
   implementation-specific and can vary according to the LPWAN
   technology and the number of Rules, among other things.  It is
   defined in Profiles.

   The RuleIDs are used:

   *  For SCHC C/D, to identify the Rule that is used to compress a
      packet header.

      -  At least one RuleID MUST be allocated to tagging packets for
         which SCHC compression was not possible (i.e., no matching
         compression Rule was found).

   *  In SCHC F/R, to identify the specific mode and settings of
      fragmentation/reassembly for one direction of data traffic (Uplink
      or Downlink).

      -  When SCHC F/R is used for both communication directions, at
         least two RuleID values are needed for fragmentation/
         reassembly: one per direction of data traffic.  This is because
         fragmentation/reassembly may entail control messages flowing in
         the reverse direction compared to data traffic.

7.  Compression/Decompression

   Compression with SCHC is based on using a set of Rules, which
   constitutes the Context of SCHC C/D, to compress or decompress
   headers.  SCHC avoids Context synchronization traffic, which consumes
   considerable bandwidth in other header compression mechanisms such as
   RObust Header Compression (RoHC) [RFC5795].  Since the content of
   packets is highly predictable in LPWANs, static Contexts can be
   stored beforehand.  The Contexts MUST be stored at both ends, and
   they can be learned by a provisioning protocol, by out-of-band means,
   or by pre-provisioning.  The way the Contexts are provisioned is out
   of the scope of this document.

7.1.  SCHC C/D Rules

   The main idea of the SCHC compression scheme is to transmit the
   RuleID to the other end instead of sending known field values.  This
   RuleID identifies a Rule that matches the original packet values.
   Hence, when a value is known by both ends, it is only necessary to
   send the corresponding RuleID over the LPWAN.  The manner by which
   Rules are generated is out of the scope of this document.  The Rules
   MAY be changed at run-time, but the mechanism is out of scope of this
   document.

   The SCHC C/D Context is a set of Rules.  See Figure 6 for a high-
   level, abstract representation of the Context.  The formal
   specification of the representation of the Rules is outside the scope
   of this document.

   Each Rule itself contains a list of Field Descriptors composed of a
   Field Identifier (FID), a Field Length (FL), a Field Position (FP), a
   Direction Indicator (DI), a Target Value (TV), a Matching Operator
   (MO), and a Compression/Decompression Action (CDA).

     /-----------------------------------------------------------------\
     |                         Rule N                                  |
    /-----------------------------------------------------------------\|
    |                       Rule i                                    ||
   /-----------------------------------------------------------------\||
   |  (FID)            Rule 1                                        |||
   |+-------+--+--+--+------------+-----------------+---------------+|||
   ||Field 1|FL|FP|DI|Target Value|Matching Operator|Comp/Decomp Act||||
   |+-------+--+--+--+------------+-----------------+---------------+|||
   ||Field 2|FL|FP|DI|Target Value|Matching Operator|Comp/Decomp Act||||
   |+-------+--+--+--+------------+-----------------+---------------+|||
   ||...    |..|..|..|   ...      | ...             | ...           ||||
   |+-------+--+--+--+------------+-----------------+---------------+||/
   ||Field N|FL|FP|DI|Target Value|Matching Operator|Comp/Decomp Act|||
   |+-------+--+--+--+------------+-----------------+---------------+|/
   |                                                                 |
   \-----------------------------------------------------------------/

                        Figure 6: A SCHC C/D Context

   A Rule does not describe how the compressor parses a packet header to
   find and identify each field (e.g., the IPv6 Source Address, the UDP
   Destination Port, or a CoAP URI path option).  It is assumed that
   there is a protocol parser alongside SCHC that is able to identify
   all the fields encountered in the headers to be compressed, and to
   label them with a Field ID.  Rules only describe the compression/
   decompression behavior for each header field, after it has been
   identified.

   In a Rule, the Field Descriptors are listed in the order in which the
   fields appear in the packet header.  The Field Descriptors describe
   the header fields with the following entries:

   *  Field Identifier (FID) designates a protocol and field (e.g., UDP
      Destination Port), unambiguously among all protocols that a SCHC
      compressor processes.  In the presence of protocol nesting, the
      Field ID also identifies the nesting.

   *  Field Length (FL) represents the length of the original field.  It
      can be either a fixed value (in bits) if the length is known when
      the Rule is created or a type if the length is variable.  The
      length of a header field is defined by its own protocol
      specification (e.g., IPv6 or UDP).  If the length is variable, the
      type defines the process to compute the length and its unit (bits,
      bytes...).

   *  Field Position (FP): most often, a field only occurs once in a
      packet header.  However, some fields may occur multiple times.  An
      example is the uri-path of CoAP.  FP indicates which occurrence
      this Field Descriptor applies to.  The default value is 1.  The
      value 1 designates the first occurrence.  The value 0 is special.
      It means "don't care", see Section 7.2.

   *  A Direction Indicator (DI) indicates the packet direction(s) this
      Field Descriptor applies to.  It allows for asymmetric processing,
      using the same Rule.  Three values are possible:

      Up:
         this Field Descriptor is only applicable to packets traveling
         Uplink.

      Dw:
         this Field Descriptor is only applicable to packets traveling
         Downlink.

      Bi:
         this Field Descriptor is applicable to packets traveling Uplink
         or Downlink.

   *  Target Value (TV) is the value used to match against the packet
      header field.  The Target Value can be a scalar value of any type
      (integer, strings, etc.) or a more complex structure (array, list,
      etc.).  The types and representations are out of scope for this
      document.

   *  Matching Operator (MO) is the operator used to match the field
      value and the Target Value.  The Matching Operator may require
      some parameters.  The set of MOs defined in this document can be
      found in Section 7.3.

   *  Compression/Decompression Action (CDA) describes the pair of
      actions that are performed at the compressor to compress a header
      field and at the decompressor to recover the original value of the
      header field.  Some CDAs might use parameter values for their
      operation.  The set of CDAs defined in this document can be found
      in Section 7.4.

7.2.  Packet Processing

   The compression/decompression process follows several phases:

   Compression Rule selection:  the general idea is to browse the Rule
      set to find a Rule that has a matching Field Descriptor (given the
      DI and FP) for all and only those header fields that appear in the
      packet being compressed.  The detailed algorithm is the following:

      *  The first step is to check the FIDs.  If any header field of
         the packet being examined cannot be matched with a Field
         Descriptor with the correct FID, the Rule MUST be disregarded.
         If any Field Descriptor in the Rule has a FID that cannot be
         matched to one of the header fields of the packet being
         examined, the Rule MUST be disregarded.

      *  The next step is to match the Field Descriptors by their
         direction, using the DI.  If any field of the packet header
         cannot be matched with a Field Descriptor with the correct FID
         and DI, the Rule MUST be disregarded.

      *  Then, the Field Descriptors are further selected according to
         FP.  If any field of the packet header cannot be matched with a
         Field Descriptor with the correct FID, DI and FP, the Rule MUST
         be disregarded.

         The value 0 for FP means "don't care", i.e., the comparison of
         this Field Descriptor's FP with the position of the field of
         the packet header being compressed returns True, whatever that
         position.  FP=0 can be useful to build compression Rules for
         protocol headers in which some fields order is irrelevant.  An
         example could be uri-queries in CoAP.  Care needs to be
         exercised when writing Rules containing FP=0 values.  Indeed,
         it may result in decompressed packets having fields ordered
         differently compared to the original packet.

      *  Once each header field has been associated with a Field
         Descriptor with matching FID, DI, and FP, each packet field's
         value is then compared to the corresponding TV stored in the
         Rule for that specific field, using the MO.  If every field in
         the packet header satisfies the corresponding MOs of a Rule
         (i.e., all MO results are True), that Rule is valid for use to
         compress the header.  Otherwise, the Rule MUST be disregarded.

         This specification does not prevent multiple Rules from
         matching the above steps and, therefore, being valid for use.
         Which Rule to use among multiple valid Rules is left to the
         implementation.  As long as the same Rule set is installed at
         both ends, this degree of freedom does not constitute an
         interoperability issue.

      *  If no valid compression Rule is found, then the packet MUST be
         sent uncompressed using the RuleID dedicated to this purpose
         (see Section 6).  The entire packet header is the Compression
         Residue (see Figure 4).  Sending an uncompressed header is
         likely to require SCHC F/R.

   Compression:  if a valid Rule is found, each field of the header is
      compressed according to the CDAs of the Rule.  The fields are
      compressed in the order that the Field Descriptors appear in the
      Rule.  The compression of each field results in a residue, which
      may be empty.  The Compression Residue for the packet header is
      the concatenation of the non-empty residues for each field of the
      header, in the order the Field Descriptors appear in the Rule.
      The order in which the Field Descriptors appear in the Rule is
      therefore semantically important.

       |------------------- Compression Residue -------------------|
       +-----------------+-----------------+-----+-----------------+
       | field 1 residue | field 2 residue | ... | field N residue |
       +-----------------+-----------------+-----+-----------------+

                  Figure 7: Compression Residue Structure

   Sending:  The RuleID is sent to the other end jointly with the
      Compression Residue (which could be empty) or the uncompressed
      header, and directly followed by the payload (see Figure 4).  The
      way the RuleID is sent will be specified in the Profile and is out
      of the scope of the present document.  For example, it could be
      included in an L2 header or sent as part of the L2 payload.

   Decompression:  when decompressing, on the Network Infrastructure
      side, the SCHC C/D needs to find the correct Rule based on the L2
      address of the Dev. On the Dev side, only the RuleID is needed to
      identify the correct Rule since the Dev typically only holds Rules
      that apply to itself.

      This Rule describes the compressed header format.  From this, the
      decompressor determines the order of the residues, the fixed-size
      or variable-size nature of each residue (see Section 7.4.2), and
      the size of the fixed-size residues.

      Therefore, from the received compressed header, it can retrieve
      all the residue values and associate them to the corresponding
      header fields.

      For each field in the header, the receiver applies the CDA action
      associated with that field in order to reconstruct the original
      header field value.  The CDA application order can be different
      from the order in which the fields are listed in the Rule.  In
      particular, Compute-* MUST be applied after the application of the
      CDAs of all the fields it computes on.

7.3.  Matching Operators

   MOs are functions used at the compression side of SCHC C/D.  They are
   not typed and can be applied to integer, string or any other data
   type.  The result of the operation can either be True or False.  The
   following MOs are defined:

   equal:  The match result is True if the field value in the packet
      matches the TV.

   ignore:  No matching is attempted between the field value in the
      packet and the TV in the Rule.  The result is always True.

   MSB(x):  A match is obtained if the most significant (leftmost) x
      bits of the packet header field value are equal to the TV in the
      Rule.  The x parameter of the MSB MO indicates how many bits are
      involved in the comparison.  If the FL is described as variable,
      the x parameter must be a multiple of the FL unit.  For example, x
      must be multiple of 8 if the unit of the variable length is bytes.

   match-mapping:  With match-mapping, TV is a list of values.  Each
      value of the list is identified by an index.  Compression is
      achieved by sending the index instead of the original header field
      value.  This operator matches if the header field value is equal
      to one of the values in the target list.

7.4.  Compression/Decompression Actions (CDA)

   The CDA specifies the actions taken during the compression of header
   fields and the inverse action taken by the decompressor to restore
   the original value.  The CDAs defined by this document are described
   in detail in Section 7.4.3 to Section 7.4.8.  They are summarized in
   Table 1.

     +--------------+------------------------+-----------------------+
     | Action       | Compression            | Decompression         |
     +==============+========================+=======================+
     | not-sent     | elided                 | use TV stored in Rule |
     +--------------+------------------------+-----------------------+
     | value-sent   | send                   | use received value    |
     +--------------+------------------------+-----------------------+
     | mapping-sent | send index             | retrieve value from   |
     |              |                        | TV list               |
     +--------------+------------------------+-----------------------+
     | LSB          | send least significant | concatenate TV and    |
     |              | bits (LSB)             | received value        |
     +--------------+------------------------+-----------------------+
     | compute-*    | elided                 | recompute at          |
     |              |                        | decompressor          |
     +--------------+------------------------+-----------------------+
     | DevIID       | elided                 | build IID from L2 Dev |
     |              |                        | addr                  |
     +--------------+------------------------+-----------------------+
     | AppIID       | elided                 | build IID from L2 App |
     |              |                        | addr                  |
     +--------------+------------------------+-----------------------+

               Table 1: Compression and Decompression Actions

   The first column shows the action's name.  The second and third
   columns show the compression and decompression behaviors for each
   action.

7.4.1.  Processing Fixed-Length Fields

   If the field is identified in the Field Descriptor as being of fixed
   length, then applying the CDA to compress this field results in a
   fixed amount of bits.  The residue for that field is simply the bits
   resulting from applying the CDA to the field.  This value may be
   empty (e.g., not-sent CDA), in which case the field residue is absent
   from the Compression Residue.

   |- field residue -|
   +-----------------+
   |      value      |
   +-----------------+

                Figure 8: Fixed-Size Field Residue Structure

7.4.2.  Processing Variable-Length Fields

   If the field is identified in the Field Descriptor as being of
   variable length, then applying the CDA to compress this field may
   result in a value of fixed size (e.g., not-sent or mapping-sent) or
   of variable size (e.g., value-sent or LSB).  In the latter case, the
   residue for that field is the bits that result from applying the CDA
   to the field, preceded with the size of the value.  The most
   significant bit of the size is stored to the left (leftmost bit of
   the residue field).

   |--- field residue ---|
   +-------+-------------+
   |  size |    value    |
   +-------+-------------+

              Figure 9: Variable-Size Field Residue Structure

   The size (using the unit defined in the FL) is encoded on 4, 12, or
   28 bits as follows:

   *  If the size is between 0 and 14, it is encoded as a 4-bit unsigned
      integer.

   *  Sizes between 15 and 254 are encoded as 0b1111 followed by the
      8-bit unsigned integer.

   *  Larger sizes are encoded as 0xfff followed by the 16-bit unsigned
      integer.

   If the field is identified in the Field Descriptor as being of
   variable length and this field is not present in the packet header
   being compressed, size 0 MUST be sent to denote its absence.

7.4.3.  Not-Sent CDA

   The not-sent action can be used when the field value is specified in
   a Rule and, therefore, known by both the Compressor and the
   Decompressor.  This action SHOULD be used with the "equal" MO.  If MO
   is "ignore", there is a risk of having a decompressed field value
   that is different from the original field that was compressed.

   The compressor does not send any residue for a field on which not-
   sent compression is applied.

   The decompressor restores the field value with the TV stored in the
   matched Rule identified by the received RuleID.

7.4.4.  Value-Sent CDA

   The value-sent action can be used when the field value is not known
   by both the Compressor and the Decompressor.  The field is sent in
   its entirety, using the same bit order as in the original packet
   header.

   If this action is performed on a variable-length field, the size of
   the residue value (using the units defined in FL) MUST be sent as
   described in Section 7.4.2.

   This action is generally used with the "ignore" MO.

7.4.5.  Mapping-Sent CDA

   The mapping-sent action is used to send an index (the index into the
   TV list of values) instead of the original value.  This action is
   used together with the "match-mapping" MO.

   On the compressor side, the match-mapping MO searches the TV for a
   match with the header field value.  The mapping-sent CDA then sends
   the corresponding index as the field residue.  The most significant
   bit of the index is stored to the left (leftmost bit of the residue
   field).

   On the decompressor side, the CDA uses the received index to restore
   the field value by looking up the list in the TV.

   The number of bits sent is the minimal size for coding all the
   possible indices.

   The first element in the list MUST be represented by index value 0,
   and successive elements in the list MUST have indices incremented by
   1.

7.4.6.  LSB CDA

   The LSB action is used together with the "MSB(x)" MO to avoid sending
   the most significant part of the packet field if that part is already
   known by the receiving end.

   The compressor sends the LSBs as the field residue value.  The number
   of bits sent is the original header field length minus the length
   specified in the MSB(x) MO.  The bits appear in the residue in the
   same bit order as in the original packet header.

   The decompressor concatenates the x most significant bits of the TV
   and the received residue value.

   If this action is performed on a variable-length field, the size of
   the residue value (using the units defined in FL) MUST be sent as
   described in Section 7.4.2.

7.4.7.  DevIID, AppIID CDA

   These actions are used to process the DevIID and AppIID of the IPv6
   addresses, respectively.  AppIID CDA is less common since most
   current LPWAN technologies frames contain a single L2 address, which
   is the Dev's address.

   The DevIID value MAY be computed from the Dev ID present in the L2
   header, or from some other stable identifier.  The computation is
   specific to each Profile and MAY depend on the Dev ID size.

   In the Downlink direction, at the compressor, the DevIID CDA may be
   used to generate the L2 addresses on the LPWAN, based on the packet's
   Destination Address.

7.4.8.  Compute-*

   Some fields can be elided at the compressor and recomputed locally at
   the decompressor.

   Because the field is uniquely identified by its FID (e.g., UDP IPv6
   length), the relevant protocol specification unambiguously defines
   the algorithm for such computation.

   Examples

   An example of fields a field that know knows how to recompute themselves are UDP
   length, itself is IPv6 length and UDP checksum.
   length.

8.  Fragmentation/Reassembly

8.1.  Overview

   In LPWAN technologies, the L2 MTU typically ranges from tens to
   hundreds of bytes.  Some of these technologies do not have an
   internal fragmentation/reassembly mechanism.

   The optional SCHC F/R functionality enables such LPWAN technologies
   to comply with the IPv6 MTU requirement of 1280 bytes [RFC8200].  It
   is OPTIONAL to implement per this specification, but Profiles may
   specify that it is REQUIRED.

   This specification includes several SCHC F/R modes, which allow for a
   range of reliability options such as optional SCHC Fragment
   retransmission.  More modes may be defined in the future.

   The same SCHC F/R mode MUST be used for all SCHC Fragments of a given
   SCHC Packet.  This document does not specify which mode(s) must be
   implemented and used over a specific LPWAN technology.  That
   information will be given in Profiles.

   SCHC allows transmitting non-fragmented SCHC Packet concurrently with
   fragmented SCHC Packets.  In addition, SCHC F/R provides protocol
   elements that allow transmitting several fragmented SCHC Packets
   concurrently, i.e., interleaving the transmission of fragments from
   different fragmented SCHC Packets.  A Profile MAY restrict the latter
   behavior.

   The L2 Word size (see Section 4) determines the encoding of some
   messages.  SCHC F/R usually generates SCHC Fragments and SCHC ACKs
   that are multiples of L2 Words.

8.2.  SCHC F/R Protocol Elements

   This subsection describes the different elements that are used to
   enable the SCHC F/R functionality defined in this document.  These
   elements include the SCHC F/R messages, tiles, windows, bitmaps,
   counters, timers, and header fields.

   The elements are described here in a generic manner.  Their
   application to each SCHC F/R mode is found in Section 8.4.

8.2.1.  Messages

   SCHC F/R defines the following messages:

   SCHC Fragment:  A message that carries part of a SCHC Packet from the
      sender to the receiver.

   SCHC ACK:  An acknowledgement for fragmentation, by the receiver to
      the sender.  This message is used to indicate whether or not the
      reception of pieces of, or the whole of, the fragmented SCHC
      Packet was successful.

   SCHC ACK REQ:  A request by the sender for a SCHC ACK from the
      receiver.

   SCHC Sender-Abort:  A message by the sender telling the receiver that
      it has aborted the transmission of a fragmented SCHC Packet.

   SCHC Receiver-Abort:  A message by the receiver to tell the sender to
      abort the transmission of a fragmented SCHC Packet.

   The format of these messages is provided in Section 8.3.

8.2.2.  Tiles, Windows, Bitmaps, Timers, Counters

8.2.2.1.  Tiles

   The SCHC Packet is fragmented into pieces, hereafter called "tiles".
   The tiles MUST be non-empty and pairwise disjoint.  Their union MUST
   be equal to the SCHC Packet.

   See Figure 10 for an example.

                                   SCHC Packet
           +----+--+-----+---+----+-+---+-----+...-----+----+---+------+
   Tiles   |    |  |     |   |    | |   |     |        |    |   |      |
           +----+--+-----+---+----+-+---+-----+...-----+----+---+------+

                 Figure 10: SCHC Packet Fragmented in Tiles

   Modes (see Section 8.4) MAY place additional constraints on tile
   sizes.

   Each SCHC Fragment message carries at least one tile in its Payload,
   if the Payload field is present.

8.2.2.2.  Windows

   Some SCHC F/R modes may handle successive tiles in groups, called
   windows.

   If windows are used:

   *  all the windows of a SCHC Packet, except the last one, MUST
      contain the same number of tiles.  This number is WINDOW_SIZE.

   *  WINDOW_SIZE MUST be specified in a Profile.

   *  the windows are numbered.

   *  their numbers MUST increment by 1 from 0 upward, from the start of
      the SCHC Packet to its end.

   *  the last window MUST contain WINDOW_SIZE tiles or less.

   *  tiles are numbered within each window.

   *  the tile indices MUST decrement by 1 from WINDOW_SIZE - 1
      downward, looking from the start of the SCHC Packet toward its
      end.

   *  therefore, each tile of a SCHC Packet is uniquely identified by a
      window number and a tile index within this window.

   See Figure 11 for an example.

           +---------------------------------------------...-----------+
           |                       SCHC Packet                         |
           +---------------------------------------------...-----------+

   Tile#   | 4 | 3 | 2 | 1 | 0 | 4 | 3 | 2 | 1 | 0 | 4 |     | 0 | 4 |3|
   Window# |-------- 0 --------|-------- 1 --------|- 2  ... 27 -|- 28-|

     Figure 11: SCHC Packet Fragmented in Tiles Grouped in 29 Windows,
                            with WINDOW_SIZE = 5

   Appendix E discusses the benefits of selecting one among multiple
   window sizes depending on the size of the SCHC Packet to be
   fragmented.

   When windows are used:

   *  Bitmaps (see Section 8.2.2.3) MAY be sent back by the receiver to
      the sender in a SCHC ACK message.

   *  A Bitmap corresponds to exactly one Window.

8.2.2.3.  Bitmaps

   Each bit in the Bitmap for a window corresponds to a tile in the
   window.  Therefore, each Bitmap has WINDOW_SIZE bits.  The bit at the
   leftmost position corresponds to the tile numbered WINDOW_SIZE - 1.
   Consecutive bits, going right, correspond to sequentially decreasing
   tile indices.  In Bitmaps for windows that are not the last one of a
   SCHC Packet, the bit at the rightmost position corresponds to the
   tile numbered 0.  In the Bitmap for the last window, the bit at the
   rightmost position corresponds either to the tile numbered 0 or to a
   tile that is sent/received as "the last one of the SCHC Packet"
   without explicitly stating its number (see Section 8.3.1.2).

   At the receiver:

   *  a bit set to 1 in the Bitmap indicates that a tile associated with
      that bit position has been correctly received for that window.

   *  a bit set to 0 in the Bitmap indicates that there has been no tile
      correctly received, associated with that bit position, for that
      window.  Possible reasons include that the tile was not sent at
      all, not received, or received with errors.

8.2.2.4.  Timers and Counters

   Some SCHC F/R modes can use the following timers and counters:

   Inactivity Timer:  a SCHC Fragment receiver uses this timer to abort
      waiting for a SCHC F/R message.

   Retransmission Timer:  a SCHC Fragment sender uses this timer to
      abort waiting for an expected SCHC ACK.

   Attempts:  this counter counts the requests for SCHC ACKs, up to
      MAX_ACK_REQUESTS.

8.2.3.  Integrity Checking

   The integrity of the fragmentation-reassembly process of a SCHC
   Packet MUST be checked at the receive end.  A Profile MUST specify
   how integrity checking is performed.

   It is RECOMMENDED that integrity checking be performed by computing a
   Reassembly Check Sequence (RCS) based on the SCHC Packet at the
   sender side and transmitting it to the receiver for comparison with
   the RCS locally computed after reassembly.

   The RCS supports UDP checksum elision by SCHC C/D (see
   Section 10.11).

   The CRC32 polynomial 0xEDB88320 (i.e., the reversed polynomial
   representation, which is used in the Ethernet standard [ETHERNET]) is
   RECOMMENDED as the default algorithm for computing the RCS.

   The RCS MUST be computed on the full SCHC Packet concatenated with
   the padding bits, if any, of the SCHC Fragment carrying the last
   tile.  The rationale is that the SCHC reassembler has no way of
   knowing the boundary between the last tile and the padding bits.
   Indeed, this requires decompressing the SCHC Packet, which is out of
   the scope of the SCHC reassembler.

   The concatenation of the complete SCHC Packet and any padding bits,
   if present, of the last SCHC Fragment does not generally constitute
   an integer number of bytes.  CRC libraries are usually byte oriented.
   It is RECOMMENDED that the concatenation of the complete SCHC Packet
   and any last fragment padding bits be zero-extended to the next byte
   boundary and that the RCS be computed on that byte array.

8.2.4.  Header Fields

   The SCHC F/R messages contain the following fields (see the formats
   in Section 8.3):

   RuleID:  this field is present in all the SCHC F/R messages.  The
      Rule identifies:

      *  that a SCHC F/R message is being carried, as opposed to an
         unfragmented SCHC Packet,

      *  which SCHC F/R mode is used,

      *  in case this mode uses windows, what the value of WINDOW_SIZE
         is, and

      *  what other optional fields are present and what the field sizes
         are.

      The Rule tells apart a non-fragmented SCHC Packet from SCHC
      Fragments.  It will also tell apart SCHC Fragments of fragmented
      SCHC Packets that use different SCHC F/R modes or different
      parameters.  Therefore, interleaved transmission of these is
      possible.

      All SCHC F/R messages pertaining to the same SCHC Packet MUST bear
      the same RuleID.

   Datagram Tag (DTag):  This field allows differentiating SCHC F/R
      messages belonging to different SCHC Packets that may be using the
      same RuleID simultaneously.  Hence, it allows interleaving
      fragments of a new SCHC Packet with fragments of a previous SCHC
      Packet under the same RuleID.

      The size of the DTag field (called "T", in bits) is defined by
      each Profile for each RuleID.  When T is 0, the DTag field does
      not appear in the SCHC F/R messages and the DTag value is defined
      as 0.

      When T is 0, there can be no more than one fragmented SCHC Packet
      in transit for each fragmentation RuleID.

      If T is not 0, DTag:

      *  MUST be set to the same value for all the SCHC F/R messages
         related to the same fragmented SCHC Packet, and

      *  MUST be set to different values for SCHC F/R messages related
         to different SCHC Packets that are being fragmented under the
         same RuleID and whose transmission may overlap.

   W:  The W field is optional.  It is only present if windows are used.
      Its presence and size (called "M", in bits) is defined by each
      SCHC F/R mode and each Profile for each RuleID.

      This field carries information pertaining to the window a SCHC F/R
      message relates to.  If present, W MUST carry the same value for
      all the SCHC F/R messages related to the same window.  Depending
      on the mode and Profile, W may carry the full window number, or
      just the LSB or any other partial representation of the window
      number.

   Fragment Compressed Number (FCN):  The FCN field is present in the
      SCHC Fragment Header.  Its size (called "N", in bits) is defined
      by each Profile for each RuleID.

      This field conveys information about the progress in the sequence
      of tiles being transmitted by SCHC Fragment messages.  For
      example, it can contain a partial, efficient representation of a
      larger-sized tile index.  The description of the exact use of the
      FCN field is left to each SCHC F/R mode.  However, two values are
      reserved for special purposes.  They help control the SCHC F/R
      process:

      *  The FCN value with all the bits equal to 1 (called "All-1")
         signals that the very last tile of a SCHC Packet has been
         transmitted.  By extension, if windows are used, the last
         window of a packet is called the "All-1" window.

      *  If windows are used, the FCN value with all the bits equal to 0
         (called "All-0") signals the last tile of a window that is not
         the last one of the SCHC packet.  By extension, such a window
         is called an "All-0 window".

   Reassembly Check Sequence (RCS):  This field only appears in the
      All-1 SCHC Fragments.  Its size (called "U", in bits) is defined
      by each Profile for each RuleID.

      See Section 8.2.3 for the RCS default size, default polynomial and
      details on RCS computation.

   C (integrity Check):  C is a 1-bit field.  This field is used in the
      SCHC ACK message to report on the reassembled SCHC Packet
      integrity check (see Section 8.2.3).

      A value of 1 tells that the integrity check was performed and is
      successful.  A value of 0 tells that the integrity check was not
      performed or that it was a failure.

   Compressed Bitmap:  The Compressed Bitmap is used together with
      windows and Bitmaps (see Section 8.2.2.3).  Its presence and size
      is defined for each SCHC F/R mode for each RuleID.

      This field appears in the SCHC ACK message to report on the
      receiver Bitmap (see Section 8.3.2.1).

8.3.  SCHC F/R Message Formats

   This section defines the SCHC Fragment formats, the SCHC ACK format,
   the SCHC ACK REQ format and the SCHC Abort formats.

8.3.1.  SCHC Fragment Format

   A SCHC Fragment conforms to the general format shown in Figure 12.
   It comprises a SCHC Fragment Header and a SCHC Fragment Payload.  The
   SCHC Fragment Payload carries one or several tile(s).

   +-----------------+-----------------------+~~~~~~~~~~~~~~~~~~~~~
   | Fragment Header | Fragment Payload      | padding (as needed)
   +-----------------+-----------------------+~~~~~~~~~~~~~~~~~~~~~

                  Figure 12: SCHC Fragment General Format

8.3.1.1.  Regular SCHC Fragment

   The Regular SCHC Fragment format is shown in Figure 13.  Regular SCHC
   Fragments are generally used to carry tiles that are not the last one
   of a SCHC Packet.  The DTag field and the W field are OPTIONAL, their
   presence is specified by each mode and Profile.

   |-- SCHC Fragment Header ----|
            |-- T --|-M-|-- N --|
   +-- ... -+- ... -+---+- ... -+--------...-------+~~~~~~~~~~~~~~~~~~~~
   | RuleID | DTag  | W |  FCN  | Fragment Payload | padding (as needed)
   +-- ... -+- ... -+---+- ... -+--------...-------+~~~~~~~~~~~~~~~~~~~~

        Figure 13: Detailed Header Format for Regular SCHC Fragments

   The FCN field MUST NOT contain all bits set to 1.

   Profiles MUST ensure that a SCHC Fragment with FCN equal to 0 (called
   an "All-0 SCHC Fragment") is distinguishable by size, even in the
   presence of padding, from a SCHC ACK REQ message (see Section 8.3.3)
   with the same RuleID value and with the same T, M, and N values.
   This condition is met if the Payload is at least the size of an L2
   Word.  This condition is also met if the SCHC Fragment Header is a
   multiple of L2 Words.

8.3.1.2.  All-1 SCHC Fragment

   The All-1 SCHC Fragment format is shown in Figure 14.  The sender
   uses the All-1 SCHC Fragment format for the message that completes
   the emission of a fragmented SCHC Packet.  The DTag field, the W
   field, the RCS field and the Payload are OPTIONAL, their presence is
   specified by each mode and Profile.  At least one of RCS field or
   Fragment Payload MUST be present.  The FCN field is all ones.

   |------- SCHC Fragment Header -------|
            |-- T --|-M-|-- N --|-- U --|
   +-- ... -+- ... -+---+- ... -+- ... -+-----...-----+~~~~~~~~~~~~~~~~~
   | RuleID | DTag  | W | 11..1 |  RCS  | FragPayload | pad. (as needed)
   +-- ... -+- ... -+---+- ... -+- ... -+-----...-----+~~~~~~~~~~~~~~~~~
                          (FCN)

       Figure 14: Detailed Header Format for the All-1 SCHC Fragment

   Profiles MUST ensure that an All-1 SCHC Fragment message is
   distinguishable by size, even in the presence of padding, from a SCHC
   Sender-Abort message (see Section 8.3.4) with the same RuleID value
   and with the same T, M, and N values.  This condition is met if the
   RCS is present and is at least the size of an L2 Word or if the
   Payload is present and is at least the size an L2 Word.  This
   condition is also met if the SCHC Sender-Abort Header is a multiple
   of L2 Words.

8.3.2.  SCHC ACK Format

   The SCHC ACK message is shown in Figure 15.  The DTag field and the W
   field are OPTIONAL, their presence is specified by each mode and
   Profile.  The Compressed Bitmap field MUST be present in SCHC F/R
   modes that use windows and MUST NOT be present in other modes.

   |--- SCHC ACK Header ----|
            |-- T --|-M-| 1 |
   +-- ... -+- ... -+---+---+~~~~~~~~~~~~~~~~~~
   | RuleID |  DTag | W |C=1| padding as needed                (success)
   +-- ... -+- ... -+---+---+~~~~~~~~~~~~~~~~~~

   +-- ... -+- ... -+---+---+------ ... ------+~~~~~~~~~~~~~~~
   | RuleID |  DTag | W |C=0|Compressed Bitmap| pad. as needed (failure)
   +-- ... -+- ... -+---+---+------ ... ------+~~~~~~~~~~~~~~~

                 Figure 15: Format of the SCHC ACK Message

   The SCHC ACK Header contains a C bit (see Section 8.2.4).

   If the C bit is set to 1 (integrity check successful), no Bitmap is
   carried.

   If the C bit is set to 0 (integrity check not performed or failed)
   and if windows are used, a Compressed Bitmap for the window referred
   to by the W field is transmitted as specified in Section 8.3.2.1.

8.3.2.1.  Bitmap Compression

   For transmission, the Compressed Bitmap in the SCHC ACK message is
   defined by the following algorithm (see Figure 16 for a follow-along
   example):

   *  Build a temporary SCHC ACK message that contains the Header
      followed by the original Bitmap (see Section 8.2.2.3 for a
      description of Bitmaps).

   *  Position scissors at the end of the Bitmap, after its last bit.

   *  While the bit on the left of the scissors is 1 and belongs to the
      Bitmap, keep moving left, then stop.

   *  Then, while the scissors are not on an L2 Word boundary of the
      SCHC ACK message and there is a Bitmap bit on the right of the
      scissors, keep moving right, then stop.

   *  At this point, cut and drop off any bits to the right of the
      scissors.

   When one or more bits have effectively been dropped off as a result
   of the above algorithm, the SCHC ACK message is a multiple of L2
   Words; no padding bits will be appended.

   Because the SCHC Fragment sender knows the size of the original
   Bitmap, it can reconstruct the original Bitmap from the Compressed
   Bitmap received in the SCHC ACK message.

   Figure 16 shows an example where L2 Words are actually bytes and
   where the original Bitmap contains 17 bits, the last 15 of which are
   all set to 1.

   |--- SCHC ACK Header ----|--------      Bitmap     --------|
            |-- T --|-M-| 1 |
   +-- ... -+- ... -+---+---+---------------------------------+
   | RuleID |  DTag | W |C=0|1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1|
   +-- ... -+- ... -+---+---+---------------------------------+
          next L2 Word boundary ->|

            Figure 16: SCHC ACK Header Plus Uncompressed Bitmap

   Figure 17 shows that the last 14 bits are not sent.

   |--- SCHC ACK Header ----|CpBmp|
            |-- T --|-M-| 1 |
   +-- ... -+- ... -+---+---+-----+
   | RuleID |  DTag | W |C=0|1 0 1|
   +-- ... -+- ... -+---+---+-----+
          next L2 Word boundary ->|

        Figure 17: Resulting SCHC ACK Message with Compressed Bitmap

   Figure 18 shows an example of a SCHC ACK with tile indices ranging
   from 6 down to 0, where the Bitmap indicates that the second and the
   fourth tile of the window have not been correctly received.

   |--- SCHC ACK Header ----|--- Bitmap --|
            |-- T --|-M-| 1 |6 5 4 3 2 1 0| (tile #)
   +--------+-------+---+---+-------------+
   | RuleID |  DTag | W |C=0|1 0 1 0 1 1 1|     uncompressed Bitmap
   +--------+-------+---+---+-------------+
      next L2 Word boundary ->|<-- L2 Word --->|

   +--------+-------+---+---+-------------+~~~~+
   | RuleID |  DTag | W |C=0|1 0 1 0 1 1 1|pad.| transmitted SCHC ACK
   +--------+-------+---+---+-------------+~~~~+
      next L2 Word boundary ->|<-- L2 Word --->|

          Figure 18: Example of a SCHC ACK Message, Missing Tiles

   Figure 19 shows an example of a SCHC ACK with tile indices ranging
   from 6 down to 0, where integrity check has not been performed or has
   failed and the Bitmap indicates that there is no missing tile in that
   window.

   |--- SCHC ACK Header ----|--- Bitmap --|
            |-- T --|-M-| 1 |6 5 4 3 2 1 0| (tile #)
   +--------+-------+---+---+-------------+
   | RuleID |  DTag | W |C=0|1 1 1 1 1 1 1|  with uncompressed Bitmap
   +--------+-------+---+---+-------------+
      next L2 Word boundary ->|

   +-- ... -+- ... -+---+---+-+
   | RuleID |  DTag | W |C=0|1|                  transmitted SCHC ACK
   +-- ... -+- ... -+---+---+-+
      next L2 Word boundary ->|

         Figure 19: Example of a SCHC ACK Message, No Missing Tile

8.3.3.  SCHC ACK REQ Format

   The SCHC ACK REQ is used by a sender to request a SCHC ACK from the
   receiver.  Its format is shown in Figure 20.  The DTag field and the
   W field are OPTIONAL, their presence is specified by each mode and
   Profile.  The FCN field is all zero.

   |--- SCHC ACK REQ Header ----|
            |-- T --|-M-|-- N --|
   +-- ... -+- ... -+---+- ... -+~~~~~~~~~~~~~~~~~~~~~
   | RuleID | DTag  | W |  0..0 | padding (as needed)      (no payload)
   +-- ... -+- ... -+---+- ... -+~~~~~~~~~~~~~~~~~~~~~

                       Figure 20: SCHC ACK REQ Format

8.3.4.  SCHC Sender-Abort Format

   When a SCHC Fragment sender needs to abort an ongoing fragmented SCHC
   Packet transmission, it sends a SCHC Sender-Abort message to the SCHC
   Fragment receiver.

   The SCHC Sender-Abort format is shown in Figure 21.  The DTag field
   and the W field are OPTIONAL, their presence is specified by each
   mode and Profile.  The FCN field is all ones.

   |--- Sender-Abort Header ----|
            |-- T --|-M-|-- N --|
   +-- ... -+- ... -+---+- ... -+~~~~~~~~~~~~~~~~~~~~~
   | RuleID | DTag  | W | 11..1 | padding (as needed)
   +-- ... -+- ... -+---+- ... -+~~~~~~~~~~~~~~~~~~~~~

                    Figure 21: SCHC Sender-Abort Format

   If the W field is present:

   *  the fragment sender MUST set it to all ones.  Other values are
      RESERVED.

   *  the fragment receiver MUST check its value.  If the value is
      different from all ones, the message MUST be ignored.

   The SCHC Sender-Abort MUST NOT be acknowledged.

8.3.5.  SCHC Receiver-Abort Format

   When a SCHC Fragment receiver needs to abort an ongoing fragmented
   SCHC Packet transmission, it transmits a SCHC Receiver-Abort message
   to the SCHC Fragment sender.

   The SCHC Receiver-Abort format is shown in Figure 22.  The DTag field
   and the W field are OPTIONAL, their presence is specified by each
   mode and Profile.

   |-- Receiver-Abort Header ---|
              |--- T ---|-M-| 1 |
   +--- ... --+-- ... --+---+---+-+-+-+-+-+-+-+-+-+-+-+
   |  RuleID  |   DTag  | W |C=1| 1..1|      1..1     |
   +--- ... --+-- ... --+---+---+-+-+-+-+-+-+-+-+-+-+-+
              next L2 Word boundary ->|<-- L2 Word -->|

                   Figure 22: SCHC Receiver-Abort Format

   If the W field is present:

   *  the fragment receiver MUST set it to all ones.  Other values are
      RESERVED.

   *  if the value is different from all ones, the fragment sender MUST
      ignore the message.

   The SCHC Receiver-Abort has the same header as a SCHC ACK message.
   The bits that follow the SCHC Receiver-Abort Header MUST be as
   follows:

   *  if the Header does not end at an L2 Word boundary, append bits set
      to 1 as needed to reach the next L2 Word boundary.

   *  append exactly one more L2 Word with bits all set to ones.

   Such a bit pattern never occurs in a legitimate SCHC ACK.  This is
   how the fragment sender recognizes a SCHC Receiver-Abort.

   The SCHC Receiver-Abort MUST NOT be acknowledged.

8.4.  SCHC F/R Modes

   This specification includes several SCHC F/R modes that:

   *  allow for a range of reliability options, such as optional SCHC
      Fragment retransmission.

   *  support various LPWAN characteristics, such as links with variable
      MTU or unidirectional links.

   More modes may be defined in the future.

   Appendix B provides examples of fragmentation sessions based on the
   modes described hereafter.

   Appendix C provides examples of Finite State Machines implementing
   the SCHC F/R modes described hereafter.

8.4.1.  No-ACK Mode

   The No-ACK mode has been designed under the assumption that data unit
   out-of-sequence delivery does not occur between the entity performing
   fragmentation and the entity performing reassembly.  This mode
   supports L2 technologies that have a variable MTU.

   In No-ACK mode, there is no communication from the fragment receiver
   to the fragment sender.  The sender transmits all the SCHC Fragments
   without expecting any acknowledgement.  Therefore, No-ACK does not
   require bidirectional links: unidirectional links are just fine.

   In No-ACK mode, only the All-1 SCHC Fragment is padded as needed.
   The other SCHC Fragments are intrinsically aligned to L2 Words.

   The tile sizes are not required to be uniform.  Windows are not used.
   The Retransmission Timer is not used.  The Attempts counter is not
   used.

   Each Profile MUST specify which RuleID value(s) corresponds to SCHC
   F/R messages operating in this mode.

   The W field MUST NOT be present in the SCHC F/R messages.  SCHC ACK
   MUST NOT be sent.  SCHC ACK REQ MUST NOT be sent.  SCHC Sender-Abort
   MAY be sent.  SCHC Receiver-Abort MUST NOT be sent.

   The value of N (size of the FCN field) is RECOMMENDED to be 1.

   Each Profile, for each RuleID value, MUST define:

   *  the size of the DTag field,

   *  the size and algorithm for the RCS field, and

   *  the expiration time of the Inactivity Timer.

   Each Profile, for each RuleID value, MAY define

   *  a value of N different from the recommended one, and

   *  the meaning of values sent in the FCN field, for values different
      from the All-1 value.

   For each active pair of RuleID and DTag values, the receiver MUST
   maintain an Inactivity Timer.  If the receiver is under-resourced to
   do this, it MUST silently drop the related messages.

8.4.1.1.  Sender Behavior

   At the beginning of the fragmentation of a new SCHC Packet, the
   fragment sender MUST select a RuleID and DTag value pair for this
   SCHC Packet.

   Each SCHC Fragment MUST contain exactly one tile in its Payload.  The
   tile MUST be at least the size of an L2 Word.  The sender MUST
   transmit the SCHC Fragments messages in the order that the tiles
   appear in the SCHC Packet.  Except for the last tile of a SCHC
   Packet, each tile MUST be of a size that complements the SCHC
   Fragment Header so that the SCHC Fragment is a multiple of L2 Words
   without the need for padding bits.  Except for the last one, the SCHC
   Fragments MUST use the Regular SCHC Fragment format specified in
   Section 8.3.1.1.  The SCHC Fragment that carries the last tile MUST
   be an All-1 SCHC Fragment, described in Section 8.3.1.2.

   The sender MAY transmit a SCHC Sender-Abort.

   Figure 39 shows an example of a corresponding state machine.

8.4.1.2.  Receiver Behavior

   Upon receiving each Regular SCHC Fragment:

   *  the receiver MUST reset the Inactivity Timer.

   *  the receiver assembles the payloads of the SCHC Fragments.

   On receiving an All-1 SCHC Fragment:

   *  the receiver MUST append the All-1 SCHC Fragment Payload and the
      padding bits to the previously received SCHC Fragment Payloads for
      this SCHC Packet.

   *  the receiver MUST perform the integrity check.

   *  if integrity checking fails, the receiver MUST drop the
      reassembled SCHC Packet.

   *  the reassembly operation concludes.

   On expiration of the Inactivity Timer, the receiver MUST drop the
   SCHC Packet being reassembled.

   On receiving a SCHC Sender-Abort, the receiver MAY drop the SCHC
   Packet being reassembled.

   Figure 40 shows an example of a corresponding state machine.

8.4.2.  ACK-Always Mode

   The ACK-Always mode has been designed under the following
   assumptions:

   *  Data unit out-of-sequence delivery does not occur between the
      entity performing fragmentation and the entity performing
      reassembly,

   *  The L2 MTU value does not change while the fragments of a SCHC
      Packet are being transmitted, and

   *  There is a feedback path from the reassembler to the fragmenter.
      See Appendix F for a discussion on using ACK-Always mode on quasi-
      bidirectional links.

   In ACK-Always mode, windows are used.  An acknowledgement, positive
   or negative, is transmitted by the fragment receiver to the fragment
   sender at the end of the transmission of each window of SCHC
   Fragments.

   The tiles are not required to be of uniform size.  In ACK-Always
   mode, only the All-1 SCHC Fragment is padded as needed.  The other
   SCHC Fragments are intrinsically aligned to L2 Words.

   Briefly, the algorithm is as follows: after a first blind
   transmission of all the tiles of a window, the fragment sender
   iterates retransmitting the tiles that are reported missing until the
   fragment receiver reports that all the tiles belonging to the window
   have been correctly received or until too many attempts were made.
   The fragment sender only advances to the next window of tiles when it
   has ascertained that all the tiles belonging to the current window
   have been fully and correctly received.  This results in a per-window
   lock-step behavior between the sender and the receiver.

   Each Profile MUST specify which RuleID value(s) correspond to SCHC F/
   R messages operating in this mode.

   The W field MUST be present and its size M MUST be 1 bit.

   Each Profile, for each RuleID value, MUST define:

   *  the value of N,

   *  the value of WINDOW_SIZE, which MUST be strictly less than 2^N,

   *  the size and algorithm for the RCS field,

   *  the value of T,

   *  the value of MAX_ACK_REQUESTS,

   *  the expiration time of the Retransmission Timer, and

   *  the expiration time of the Inactivity Timer.

   For each active pair of RuleID and DTag values, the sender MUST
   maintain:

   *  one Attempts counter

   *  one Retransmission Timer

   For each active pair of RuleID and DTag values, the receiver MUST
   maintain

   *  one Inactivity Timer, and

   *  one Attempts counter.

8.4.2.1.  Sender Behavior

   At the beginning of the fragmentation of a new SCHC Packet, the
   fragment sender MUST select a RuleID and DTag value pair for this
   SCHC Packet.

   Each SCHC Fragment MUST contain exactly one tile in its Payload.  All
   tiles with the index 0, as well as the last tile, MUST be at least
   the size of an L2 Word.

   In all SCHC Fragment messages, the W field MUST be filled with the
   LSB of the window number that the sender is currently processing.

   For a SCHC Fragment that carries a tile other than the last one of
   the SCHC Packet:

   *  the Fragment MUST be of the Regular type specified in
      Section 8.3.1.1.

   *  the FCN field MUST contain the tile index.

   *  each tile MUST be of a size that complements the SCHC Fragment
      Header so that the SCHC Fragment is a multiple of L2 Words without
      the need for padding bits.

   The SCHC Fragment that carries the last tile MUST be an All-1 SCHC
   Fragment, described in Section 8.3.1.2.

   The fragment sender MUST start by transmitting the window numbered 0.

   All message receptions being discussed in the rest of this section
   are to be understood as "matching the RuleID and DTag pair being
   processed", even if not spelled out, for brevity.

   The sender starts by a "blind transmission" phase, in which it MUST
   transmit all the tiles composing the window, in decreasing tile index
   order.

   Then, it enters a "retransmission phase" in which it MUST initialize
   an Attempts counter to 0, it MUST start a Retransmission Timer and it
   MUST await a SCHC ACK.

   *  Then, upon receiving a SCHC ACK:

      -  if the SCHC ACK indicates that some tiles are missing at the
         receiver, then the sender MUST transmit all the tiles that have
         been reported missing, it MUST increment Attempts, it MUST
         reset the Retransmission Timer, and MUST await the next SCHC
         ACK.

      -  if the current window is not the last one and the SCHC ACK
         indicates that all tiles were correctly received, the sender
         MUST stop the Retransmission Timer, it MUST advance to the next
         fragmentation window, and it MUST start a blind transmission
         phase as described above.

      -  if the current window is the last one and the SCHC ACK
         indicates that more tiles were received than the sender sent,
         the fragment sender MUST send a SCHC Sender-Abort, and it MAY
         exit with an error condition.

      -  if the current window is the last one and the SCHC ACK
         indicates that all tiles were correctly received, yet the
         integrity check was a failure, the fragment sender MUST send a
         SCHC Sender-Abort, and it MAY exit with an error condition.

      -  if the current window is the last one and the SCHC ACK
         indicates that integrity checking was successful, the sender
         exits successfully.

   *  on Retransmission Timer expiration:

      -  if Attempts is strictly less that MAX_ACK_REQUESTS, the
         fragment sender MUST send a SCHC ACK REQ and MUST increment the
         Attempts counter.

      -  otherwise, the fragment sender MUST send a SCHC Sender-Abort,
         and it MAY exit with an error condition.

   At any time:

   *  on receiving a SCHC Receiver-Abort, the fragment sender MAY exit
      with an error condition.

   *  on receiving a SCHC ACK that bears a W value different from the W
      value that it currently uses, the fragment sender MUST silently
      discard and ignore that SCHC ACK.

   Figure 41 shows an example of a corresponding state machine.

8.4.2.2.  Receiver Behavior

   On receiving a SCHC Fragment with a RuleID and DTag pair not being
   processed at that time:

   *  the receiver SHOULD check if the DTag value has not recently been
      used for that RuleID value, thereby ensuring that the received
      SCHC Fragment is not a remnant of a prior fragmented SCHC Packet
      transmission.  The initial value of the Inactivity Timer is the
      RECOMMENDED lifetime for the DTag value at the receiver.  If the
      SCHC Fragment is determined to be such a remnant, the receiver MAY
      silently ignore it and discard it.

   *  the receiver MUST start a process to assemble a new SCHC Packet
      with that RuleID and DTag value pair.

   *  the receiver MUST start an Inactivity Timer for that RuleID and
      DTag pair.  It MUST initialize an Attempts counter to 0 for that
      RuleID and DTag pair.  It MUST initialize a window counter to 0.
      If the receiver is under-resourced to do this, it MUST respond to
      the sender with a SCHC Receiver-Abort.

   In the rest of this section, "local W bit" means the least
   significant bit of the window counter of the receiver.

   On reception of any SCHC F/R message for the RuleID and DTag pair
   being processed, the receiver MUST reset the Inactivity Timer
   pertaining to that RuleID and DTag pair.

   All message receptions being discussed in the rest of this section
   are to be understood as "matching the RuleID and DTag pair being
   processed", even if not spelled out, for brevity.

   The receiver MUST first initialize an empty Bitmap for the first
   window then enter an "acceptance phase", in which:

   *  on receiving a SCHC Fragment or a SCHC ACK REQ, either one having
      the W bit different from the local W bit, the receiver MUST
      silently ignore and discard that message.

   *  on receiving a SCHC ACK REQ with the W bit equal to the local W
      bit, the receiver MUST send a SCHC ACK for this window.

   *  on receiving a SCHC Fragment with the W bit equal to the local W
      bit, the receiver MUST assemble the received tile based on the
      window counter and on the FCN field in the SCHC Fragment, and it
      MUST update the Bitmap.

      -  if the SCHC Fragment received is an All-0 SCHC Fragment, the
         current window is determined to be a not-last window, the
         receiver MUST send a SCHC ACK for this window and it MUST enter
         the "retransmission phase" for this window.

      -  if the SCHC Fragment received is an All-1 SCHC Fragment, the
         current window is determined to be the last window, the padding
         bits of the All-1 SCHC Fragment MUST be assembled after the
         received tile, the receiver MUST perform the integrity check
         and it MUST send a SCHC ACK for this window.  Then:

         o  If the integrity check indicates that the full SCHC Packet
            has been correctly reassembled, the receiver MUST enter the
            "clean-up phase" for this window.

         o  If the integrity check indicates that the full SCHC Packet
            has not been correctly reassembled, the receiver enters the
            "retransmission phase" for this window.

   In the "retransmission phase":

   *  if the window is a not-last window:

      -  on receiving a SCHC Fragment that is not All-0 or All-1 and
         that has a W bit different from the local W bit, the receiver
         MUST increment its window counter and allocate a fresh Bitmap,
         it MUST assemble the tile received and update the Bitmap, and
         it MUST enter the "acceptance phase" for that new window.

      -  on receiving a SCHC ACK REQ with a W bit different from the
         local W bit, the receiver MUST increment its window counter and
         allocate a fresh Bitmap, it MUST send a SCHC ACK for that new
         window, and it MUST enter the "acceptance phase" for that new
         window.

      -  on receiving a SCHC All-0 Fragment with a W bit different from
         the local W bit, the receiver MUST increment its window counter
         and allocate a fresh Bitmap, it MUST assemble the tile received
         and update the Bitmap, it MUST send a SCHC ACK for that new
         window, and it MUST stay in the "retransmission phase" for that
         new window.

      -  on receiving a SCHC All-1 Fragment with a W bit different from
         the local W bit, the receiver MUST increment its window counter
         and allocate a fresh Bitmap; it MUST assemble the tile
         received, including the padding bits; it MUST update the Bitmap
         and perform the integrity check; it MUST send a SCHC ACK for
         the new window, which is determined to be the last window.
         Then:

         o  If the integrity check indicates that the full SCHC Packet
            has been correctly reassembled, the receiver MUST enter the
            "clean-up phase" for that new window.

         o  If the integrity check indicates that the full SCHC Packet
            has not been correctly reassembled, the receiver enters the
            "retransmission phase" for that new window.

      -  on receiving a SCHC Fragment with a W bit equal to the local W
         bit:

         o  if the SCHC Fragment received is an All-1 SCHC Fragment, the
            receiver MUST silently ignore it and discard it.

         o  otherwise, the receiver MUST assemble the tile received and
            update the Bitmap.  If the Bitmap becomes fully populated
            with 1's or if the SCHC Fragment is an All-0, the receiver
            MUST send a SCHC ACK for this window.

      -  on receiving a SCHC ACK REQ with the W bit equal to the local W
         bit, the receiver MUST send a SCHC ACK for this window.

   *  if the window is the last window:

      -  on receiving a SCHC Fragment or a SCHC ACK REQ, either one
         having a W bit different from the local W bit, the receiver
         MUST silently ignore and discard that message.

      -  on receiving a SCHC ACK REQ with the W bit equal to the local W
         bit, the receiver MUST send a SCHC ACK for this window.

      -  on receiving a SCHC Fragment with a W bit equal to the local W
         bit:

         o  if the SCHC Fragment received is an All-0 SCHC Fragment, the
            receiver MUST silently ignore it and discard it.

         o  otherwise, the receiver MUST update the Bitmap, and it MUST
            assemble the tile received.  If the SCHC Fragment received
            is an All-1 SCHC Fragment, the receiver MUST assemble the
            padding bits of the All-1 SCHC Fragment after the received
            tile, it MUST perform the integrity check and:

            +  if the integrity check indicates that the full SCHC
               Packet has been correctly reassembled, the receiver MUST
               send a SCHC ACK and it enters the "clean-up phase".

            +  if the integrity check indicates that the full SCHC
               Packet has not been correctly reassembled:

               *  if the SCHC Fragment received was an All-1 SCHC
                  Fragment, the receiver MUST send a SCHC ACK for this
                  window.

   In the "clean-up phase":

   *  On receiving an All-1 SCHC Fragment or a SCHC ACK REQ, either one
      having the W bit equal to the local W bit, the receiver MUST send
      a SCHC ACK.

   *  Any other SCHC Fragment received MUST be silently ignored and
      discarded.

   At any time, on sending a SCHC ACK, the receiver MUST increment the
   Attempts counter.

   At any time, on incrementing its window counter, the receiver MUST
   reset the Attempts counter.

   At any time, on expiration of the Inactivity Timer, on receiving a
   SCHC Sender-Abort or when Attempts reaches MAX_ACK_REQUESTS, the
   receiver MUST send a SCHC Receiver-Abort, and it MAY exit the receive
   process for that SCHC Packet.

   Figure 42 shows an example of a corresponding state machine.

8.4.3.  ACK-on-Error Mode

   The ACK-on-Error mode supports L2 technologies that have variable MTU
   and out-of-order delivery.  It requires an L2 that provides a
   feedback path from the reassembler to the fragmenter.  See Appendix F
   for a discussion on using ACK-on-Error mode on quasi-bidirectional
   links.

   In ACK-on-Error mode, windows are used.

   All tiles except the last one and the penultimate one MUST be of
   equal size, hereafter called "regular".  The size of the last tile
   MUST be smaller than or equal to the regular tile size.  Regarding
   the penultimate tile, a Profile MUST pick one of the following two
   options:

   *  The penultimate tile size MUST be the regular tile size, or

   *  the penultimate tile size MUST be either the regular tile size or
      the regular tile size minus one L2 Word.

   A SCHC Fragment message carries one or several contiguous tiles,
   which may span multiple windows.  A SCHC ACK reports on the reception
   of exactly one window of tiles.

   See Figure 23 for an example.

           +---------------------------------------------...-----------+
           |                       SCHC Packet                         |
           +---------------------------------------------...-----------+

   Tile#   | 4 | 3 | 2 | 1 | 0 | 4 | 3 | 2 | 1 | 0 | 4 |     | 0 | 4 |3|
   Window# |-------- 0 --------|-------- 1 --------|- 2  ... 27 -|- 28-|

   SCHC Fragment msg   |-----------|

       Figure 23: SCHC Packet Fragmented in Tiles, ACK-on-Error Mode

   The W field is wide enough that it unambiguously represents an
   absolute window number.  The fragment receiver sends SCHC ACKs to the
   fragment sender about windows for which tiles are missing.  No SCHC
   ACK is sent by the fragment receiver for windows that it knows have
   been fully received.

   The fragment sender retransmits SCHC Fragments for tiles that are
   reported missing.  It can advance to next windows even before it has
   ascertained that all tiles belonging to previous windows have been
   correctly received, and it can still later retransmit SCHC Fragments
   with tiles belonging to previous windows.  Therefore, the sender and
   the receiver may operate in a decoupled fashion.  The fragmented SCHC
   Packet transmission concludes when:

   *  integrity checking shows that the fragmented SCHC Packet has been
      correctly reassembled at the receive end, and this information has
      been conveyed back to the sender, or

   *  too many retransmission attempts were made, or

   *  the receiver determines that the transmission of this fragmented
      SCHC Packet has been inactive for too long.

   Each Profile MUST specify which RuleID value(s) corresponds to SCHC
   F/R messages operating in this mode.

   The W field MUST be present in the SCHC F/R messages.

   Each Profile, for each RuleID value, MUST define:

   *  the tile size (a tile does not need to be multiple of an L2 Word,
      but it MUST be at least the size of an L2 Word),

   *  the value of M,

   *  the value of N,

   *  the value of WINDOW_SIZE, which MUST be strictly less than 2^N,

   *  the size and algorithm for the RCS field,

   *  the value of T,

   *  the value of MAX_ACK_REQUESTS,

   *  the expiration time of the Retransmission Timer,

   *  the expiration time of the Inactivity Timer,

   *  if the last tile is carried in a Regular SCHC Fragment or an All-1
      SCHC Fragment (see Section 8.4.3.1), and

   *  if the penultimate tile MAY be one L2 Word smaller than the
      regular tile size.  In this case, the regular tile size MUST be at
      least twice the L2 Word size.

   For each active pair of RuleID and DTag values, the sender MUST
   maintain:

   *  one Attempts counter, and

   *  one Retransmission Timer.

   For each active pair of RuleID and DTag values, the receiver MUST
   maintain:

   *  one Inactivity Timer, and

   *  one Attempts counter.

8.4.3.1.  Sender Behavior

   At the beginning of the fragmentation of a new SCHC Packet:

   *  the fragment sender MUST select a RuleID and DTag value pair for
      this SCHC Packet.  A Rule MUST NOT be selected if the values of M
      and WINDOW_SIZE for that Rule are such that the SCHC Packet cannot
      be fragmented in (2^M) * WINDOW_SIZE tiles or less.

   *  the fragment sender MUST initialize the Attempts counter to 0 for
      that RuleID and DTag value pair.

   A Regular SCHC Fragment message carries in its payload one or more
   tiles.  If more than one tile is carried in one Regular SCHC
   Fragment:

   *  the selected tiles MUST be contiguous in the original SCHC Packet,
      and

   *  they MUST be placed in the SCHC Fragment Payload adjacent to one
      another, in the order they appear in the SCHC Packet, from the
      start of the SCHC Packet toward its end.

   Tiles that are not the last one MUST be sent in Regular SCHC
   Fragments specified in Section 8.3.1.1.  The FCN field MUST contain
   the tile index of the first tile sent in that SCHC Fragment.

   In a Regular SCHC Fragment message, the sender MUST fill the W field
   with the window number of the first tile sent in that SCHC Fragment.

   A Profile MUST define if the last tile of a SCHC Packet is sent:

   *  in a Regular SCHC Fragment, alone or as part of a multi-tiles
      Payload,

   *  alone in an All-1 SCHC Fragment, or

   *  with any of the above two methods.

   In an All-1 SCHC Fragment message, the sender MUST fill the W field
   with the window number of the last tile of the SCHC Packet.

   The fragment sender MUST send SCHC Fragments such that, all together,
   they contain all the tiles of the fragmented SCHC Packet.

   The fragment sender MUST send at least one All-1 SCHC Fragment.

   In doing the two items above, the sender MUST ascertain that the
   receiver will not receive the last tile through both a Regular SCHC
   Fragment and an All-1 SCHC Fragment.

   The fragment sender MUST listen for SCHC ACK messages after having
   sent:

   *  an All-1 SCHC Fragment, or

   *  a SCHC ACK REQ.

   A Profile MAY specify other times at which the fragment sender MUST
   listen for SCHC ACK messages.  For example, this could be after
   sending a complete window of tiles.

   Each time a fragment sender sends an All-1 SCHC Fragment or a SCHC
   ACK REQ:

   *  it MUST increment the Attempts counter, and

   *  it MUST reset the Retransmission Timer.

   On Retransmission Timer expiration:

   *  if the Attempts counter is strictly less than MAX_ACK_REQUESTS,
      the fragment sender MUST send either the All-1 SCHC Fragment or a
      SCHC ACK REQ with the W field corresponding to the last window,

   *  otherwise, the fragment sender MUST send a SCHC Sender-Abort, and
      it MAY exit with an error condition.

   All message receptions being discussed in the rest of this section
   are to be understood as "matching the RuleID and DTag pair being
   processed", even if not spelled out, for brevity.

   On receiving a SCHC ACK:

   *  if the W field in the SCHC ACK corresponds to the last window of
      the SCHC Packet:

      -  if the C bit is set, the sender MAY exit successfully.

      -  otherwise:

         o  if the Profile mandates that the last tile be sent in an
            All-1 SCHC Fragment:

            +  if the SCHC ACK shows no missing tile at the receiver,
               the sender:

               *  MUST send a SCHC Sender-Abort, and

               *  MAY exit with an error condition.

            +  otherwise:

               *  the fragment sender MUST send SCHC Fragment messages
                  containing all the tiles that are reported missing in
                  the SCHC ACK.

               *  if the last of these SCHC Fragment messages is not an
                  All-1 SCHC Fragment, then the fragment sender MUST in
                  addition send after it a SCHC ACK REQ with the W field
                  corresponding to the last window.

               *  in doing the two items above, the sender MUST
                  ascertain that the receiver will not receive the last
                  tile through both a Regular SCHC Fragment and an All-1
                  SCHC Fragment.

         o  otherwise:

            +  if the SCHC ACK shows no missing tile at the receiver,
               the sender MUST send the All-1 SCHC Fragment

            +  otherwise:

               *  the fragment sender MUST send SCHC Fragment messages
                  containing all the tiles that are reported missing in
                  the SCHC ACK.

               *  the fragment sender MUST then send either the All-1
                  SCHC Fragment or a SCHC ACK REQ with the W field
                  corresponding to the last window.

   *  otherwise, the fragment sender:

      -  MUST send SCHC Fragment messages containing the tiles that are
         reported missing in the SCHC ACK.

      -  then, it MAY send a SCHC ACK REQ with the W field corresponding
         to the last window.

   See Figure 43 for one among several possible examples of a Finite
   State Machine implementing a sender behavior obeying this
   specification.

8.4.3.2.  Receiver Behavior

   On receiving a SCHC Fragment with a RuleID and DTag pair not being
   processed at that time:

   *  the receiver SHOULD check if the DTag value has not recently been
      used for that RuleID value, thereby ensuring that the received
      SCHC Fragment is not a remnant of a prior fragmented SCHC Packet
      transmission.  The initial value of the Inactivity Timer is the
      RECOMMENDED lifetime for the DTag value at the receiver.  If the
      SCHC Fragment is determined to be such a remnant, the receiver MAY
      silently ignore it and discard it.

   *  the receiver MUST start a process to assemble a new SCHC Packet
      with that RuleID and DTag value pair.  The receiver MUST start an
      Inactivity Timer for that RuleID and DTag value pair.  It MUST
      initialize an Attempts counter to 0 for that RuleID and DTag value
      pair.  If the receiver is under-resourced to do this, it MUST
      respond to the sender with a SCHC Receiver-Abort.

   On reception of any SCHC F/R message for the RuleID and DTag pair
   being processed, the receiver MUST reset the Inactivity Timer
   pertaining to that RuleID and DTag pair.

   All message receptions being discussed in the rest of this section
   are to be understood as "matching the RuleID and DTag pair being
   processed", even if not spelled out, for brevity.

   On receiving a SCHC Fragment message, the receiver determines what
   tiles were received, based on the payload length and on the W and FCN
   fields of the SCHC Fragment.

   *  if the FCN is All-1, if a Payload is present, the full SCHC
      Fragment Payload MUST be assembled including the padding bits.
      This is because the size of the last tile is not known by the
      receiver; therefore, padding bits are indistinguishable from the
      tile data bits, at this stage.  They will be removed by the SCHC
      C/D sublayer.  If the size of the SCHC Fragment Payload exceeds or
      equals the size of one regular tile plus the size of an L2 Word,
      this SHOULD raise an error flag.

   *  otherwise, tiles MUST be assembled based on the a priori known
      tile size.

      -  If allowed by the Profile, the end of the payload MAY contain
         the last tile, which may be shorter.  Padding bits are
         indistinguishable from the tile data bits, at this stage.

      -  The payload may contain the penultimate tile that, if allowed
         by the Profile, MAY be exactly one L2 Word shorter than the
         regular tile size.

      -  Otherwise, padding bits MUST be discarded.  This is possible
         because:

         o  the size of the tiles is known a priori,

         o  tiles are larger than an L2 Word, and

         o  padding bits are always strictly less than an L2 Word.

   On receiving a SCHC ACK REQ or an All-1 SCHC Fragment:

   *  if the receiver knows of any windows with missing tiles for the
      packet being reassembled, it MUST return a SCHC ACK for the
      lowest-numbered such window:

   *  otherwise:

      -  if it has received at least one tile, it MUST return a SCHC ACK
         for the highest-numbered window it currently has tiles for,

      -  otherwise, it MUST return a SCHC ACK for window numbered 0.

   A Profile MAY specify other times and circumstances at which a
   receiver sends a SCHC ACK, and which window the SCHC ACK reports
   about in these circumstances.

   Upon sending a SCHC ACK, the receiver MUST increase the Attempts
   counter.

   After receiving an All-1 SCHC Fragment, a receiver MUST check the
   integrity of the reassembled SCHC Packet at least every time it
   prepares for sending a SCHC ACK for the last window.

   Upon receiving a SCHC Sender-Abort, the receiver MAY exit with an
   error condition.

   Upon expiration of the Inactivity Timer, the receiver MUST send a
   SCHC Receiver-Abort, and it MAY exit with an error condition.

   On the Attempts counter exceeding MAX_ACK_REQUESTS, the receiver MUST
   send a SCHC Receiver-Abort, and it MAY exit with an error condition.

   Reassembly of the SCHC Packet concludes when:

   *  a Sender-Abort has been received, or

   *  the Inactivity Timer has expired, or

   *  the Attempts counter has exceeded MAX_ACK_REQUESTS, or

   *  at least an All-1 SCHC Fragment has been received and integrity
      checking of the reassembled SCHC Packet is successful.

   See Figure 44 for one among several possible examples of a Finite
   State Machine implementing a receiver behavior obeying this
   specification.  The example provided is meant to match the sender
   Finite State Machine of Figure 43.

9.  Padding Management

   SCHC C/D and SCHC F/R operate on bits, not bytes.  SCHC itself does
   not have any alignment prerequisite.  The size of SCHC Packets can be
   any number of bits.

   If the L2 constrains the payload to align to coarser boundaries (for
   example, bytes), the SCHC messages MUST be padded.  When padding
   occurs, the number of appended bits MUST be strictly less than the L2
   Word size.

   If a SCHC Packet is sent unfragmented (see Figure 24), it is padded
   as needed for transmission.

   If a SCHC Packet needs to be fragmented for transmission, it is not
   padded in itself.  Only the SCHC F/R messages are padded as needed
   for transmission.  Some SCHC F/R messages are intrinsically aligned
   to L2 Words.

   A packet (e.g., an IPv6 packet)
            |                                           ^ (padding bits
            v                                           |       dropped)
   +------------------+                      +--------------------+
   | SCHC Compression |                      | SCHC Decompression |
   +------------------+                      +--------------------+
            |                                           ^
            |   If no fragmentation,                    |
            +---- SCHC Packet + padding as needed ----->|
            |                                           | (integrity
            v                                           |  checked)
   +--------------------+                       +-----------------+
   | SCHC Fragmentation |                       | SCHC Reassembly |
   +--------------------+                       +-----------------+
        |       ^                                   |       ^
        |       |                                   |       |
        |       +--- SCHC ACK + padding as needed --+       |
        |                                                   |
        +------- SCHC Fragments + padding as needed---------+

           Sender                                    Receiver

          Figure 24: SCHC Operations, Including Padding as Needed

   Each Profile MUST specify the size of the L2 Word.  The L2 Word might
   actually be a single bit, in which case no padding will take place at
   all.

   A Profile MUST define the value of the padding bits if the L2 Word is
   wider than a single bit.  The RECOMMENDED value is 0.

10.  SCHC Compression for IPv6 and UDP Headers

   This section lists the IPv6 and UDP header fields and describes how
   they can be compressed.  An example of a set of Rules for UDP/IPv6
   header compression is provided in Appendix A.

10.1.  IPv6 Version Field

   The IPv6 version field is labeled by the protocol parser as being the
   "version" field of the IPv6 protocol.  Therefore, it only exists for
   IPv6 packets.  In the Rule, TV is set to 6, MO to "ignore" and CDA to
   "not-sent".

10.2.  IPv6 Traffic Class Field

   If the Diffserv field does not vary and is known by both sides, the
   Field Descriptor in the Rule SHOULD contain a TV with this well-known
   value, an "equal" MO, and a "not-sent" CDA.

   Otherwise (e.g., ECN bits are to be transmitted), two possibilities
   can be considered depending on the variability of the value:

   *  One possibility is to not compress the field and send the original
      value.  In the Rule, TV is not set to any particular value, MO is
      set to "ignore", and CDA is set to "value-sent".

   *  If some upper bits in the field are constant and known, a better
      option is to only send the LSBs.  In the Rule, TV is set to a
      value with the stable known upper part, MO is set to MSB(x), and
      CDA to LSB.

      ECN functionality depends on both bits of the ECN field, which are
      the 2 LSBs of this field; hence, sending only a single LSB of this
      field is NOT RECOMMENDED.

10.3.  Flow Label Field

   If the flow label is not set, i.e., its value is zero, the Field
   Descriptor in the Rule SHOULD contain a TV set to zero, an "equal"
   MO, and a "not-sent" CDA.

   If the flow label is set to a pseudorandom value according to
   [RFC6437], in the Rule, TV is not set to any particular value, MO is
   set to "ignore", and CDA is set to "value-sent".

   If the flow label is set according to some prior agreement, i.e., by
   a flow state establishment method as allowed by [RFC6437], the Field
   Descriptor in the Rule SHOULD contain a TV with this agreed-upon
   value, an "equal" MO, and a "not-sent" CDA.

10.4.  Payload Length Field

   This field can be elided for the transmission on the LPWAN.  The SCHC
   C/D recomputes the original payload length value.  In the Field
   Descriptor, TV is not set, MO is set to "ignore", and CDA is
   "compute-*".

10.5.  Next Header Field

   If the Next Header field does not vary and is known by both sides,
   the Field Descriptor in the Rule SHOULD contain a TV with this Next
   Header value, the MO SHOULD be "equal", and the CDA SHOULD be "not-
   sent".

   Otherwise, TV is not set in the Field Descriptor, MO is set to
   "ignore", and CDA is set to "value-sent".  Alternatively, a matching-
   list MAY also be used.

10.6.  Hop Limit Field

   The field behavior for this field is different for Uplink and
   Downlink.  In Uplink, since there is no IP forwarding between the Dev
   and the SCHC C/D, the value is relatively constant.  On the other
   hand, the Downlink value depends on Internet routing and can change
   more frequently.  The DI can be used to distinguish both directions:

   *  in an Up Field Descriptor, elide the field: the TV is set to the
      known constant value, the MO is set to "equal" and the CDA is set
      to "not-sent".

   *  in a Dw Field Descriptor, the Hop Limit is elided for transmission
      and forced to 1 at the receiver, by setting TV to 1, MO to
      "ignore" and CDA to "not-sent".  This prevents any further
      forwarding.

10.7.  IPv6 Addresses Fields

   As in 6LoWPAN [RFC4944], IPv6 addresses are split into two 64-bit-
   long fields; one for the prefix and one for the Interface Identifier
   (IID).  These fields SHOULD be compressed.  To allow for a single
   Rule being used for both directions, these values are identified by
   their role (Dev or App) and not by their position in the header
   (source or destination).

10.7.1.  IPv6 Source and Destination Prefixes

   Both ends MUST be configured with the appropriate prefixes.  For a
   specific flow, the source and destination prefixes can be unique and
   stored in the Context.  In that case, the TV for the source and
   destination prefixes contain the values, the MO is set to "equal" and
   the CDA is set to "not-sent".

   If the Rule is intended to compress packets with different prefix
   values, match-mapping SHOULD be used.  The different prefixes are
   listed in the TV, the MO is set to "match-mapping" and the CDA is set
   to "mapping-sent".  See Figure 26.

   Otherwise, the TV is not set, the MO is set to "ignore", and the CDA
   is set to "value-sent".

10.7.2.  IPv6 Source and Destination IID

   If the Dev or App IID are based on an L2 address, then the IID can be
   reconstructed with information coming from the L2 header.  In that
   case, the TV is not set, the MO is set to "ignore" and the CDA is set
   to "DevIID" or "AppIID".  On LPWAN technologies where the frames
   carry a single identifier (corresponding to the Dev), AppIID cannot
   be used.

   As described in [RFC8065], it may be undesirable to build the Dev
   IPv6 IID out of the Dev address.  Another static value is used
   instead.  In that case, the TV contains the static value, the MO
   operator is set to "equal" and the CDA is set to "not-sent".

   If several IIDs are possible, then the TV contains the list of
   possible IIDs, the MO is set to "match-mapping" and the CDA is set to
   "mapping-sent".

   It may also happen that the IID variability only expresses itself on
   a few bytes.  In that case, the TV is set to the stable part of the
   IID, the MO is set to "MSB" and the CDA is set to "LSB".

   Finally, the IID can be sent in its entirety on the L2.  In that
   case, the TV is not set, the MO is set to "ignore", and the CDA is
   set to "value-sent".

10.8.  IPv6 Extension Headers

   This document does not provide recommendations on how to compress
   IPv6 extension headers.

10.9.  UDP Source and Destination Ports

   To allow for a single Rule being used for both directions, the UDP
   port values are identified by their role (Dev or App) and not by
   their position in the header (source or destination).  The SCHC C/D
   MUST be aware of the traffic direction (Uplink, Downlink) to select
   the appropriate field.  The following Rules apply for Dev and App
   port numbers.

   If both ends know the port number, it can be elided.  The TV contains
   the port number, the MO is set to "equal", and the CDA is set to
   "not-sent".

   If the port variation is on few bits, the TV contains the stable part
   of the port number, the MO is set to "MSB", and the CDA is set to
   "LSB".

   If some well-known values are used, the TV can contain the list of
   these values, the MO is set to "match-mapping", and the CDA is set to
   "mapping-sent".

   Otherwise, the port numbers are sent over the L2.  The TV is not set,
   the MO is set to "ignore" and the CDA is set to "value-sent".

10.10.  UDP Length Field

   The parser MUST NOT label this field unless the UDP length can be computed Length value
   matches the Payload Length value from the received data. IPv6 header.  The TV is not
   set, the MO is set to "ignore", and the CDA is set to "compute-*".

10.11.  UDP Checksum Field

   The UDP checksum operation is mandatory with IPv6 for most packets,
   but there are exceptions [RFC8200].

   For instance, protocols that use UDP as a tunnel encapsulation may
   enable zero-checksum mode for a specific port (or set of ports) for
   sending and/or receiving.  [RFC8200] requires any node implementing
   zero-checksum mode to follow the requirements specified in
   "Applicability Statement for the Use of IPv6 UDP Datagrams with Zero
   Checksums" [RFC6936].

   6LoWPAN Header Compression [RFC6282] also specifies that a UDP
   checksum can be elided by the compressor and recomputed by the
   decompressor when an upper layer guarantees the integrity of the UDP
   payload and pseudo-header.  A specific example of this is when a
   message integrity check protects the compressed message between the
   compressor that elides the UDP checksum and the decompressor that
   computes it, with a strength that is identical or better to the UDP
   checksum.

   Similarly, a SCHC compressor MAY elide the UDP checksum when another
   layer guarantees at least equal integrity protection for the UDP
   payload and the pseudo-header.  In this case, the TV is not set, the
   MO is set to "ignore", and the CDA is set to "compute-*".

   In particular, when SCHC fragmentation is used, a fragmentation RCS
   of 2 bytes or more provides equal or better protection than the UDP
   checksum; in that case, if the compressor is collocated with the
   fragmentation point and the decompressor is collocated with the
   packet reassembly point, and if the SCHC Packet is fragmented even
   when it would fit unfragmented in the L2 MTU, then the compressor MAY
   verify and then elide the UDP checksum.  Whether and when the UDP
   Checksum is elided is to be specified in the Profile.

   Since the compression happens before the fragmentation, implementers
   should understand the risks when dealing with unprotected data below
   the transport layer and take special care when manipulating that
   data.

   In other cases, the checksum SHOULD be explicitly sent.  The TV is
   not set, the MO is set to "ignore" and the CDA is set to "value-
   sent".

11.  IANA Considerations

   This document has no IANA actions.

12.  Security Considerations

   As explained in Section 5, SCHC is expected to be implemented on top
   of LPWAN technologies, which are expected to implement security
   measures.

   In this section, we analyze the potential security threats that could
   be introduced into an LPWAN by adding the SCHC functionalities.

12.1.  Security Considerations for SCHC Compression/Decompression

12.1.1.  Forged SCHC Packet

   Let's assume that an attacker is able to send a forged SCHC Packet to
   a SCHC decompressor.

   Let's first consider the case where the RuleID contained in that
   forged SCHC Packet does not correspond to a Rule allocated in the
   Rule table.  An implementation should detect that the RuleID is
   invalid and should silently drop the offending SCHC Packet.

   Let's now consider that the RuleID corresponds to a Rule in the
   table.  With the CDAs defined in this document, the reconstructed
   packet is, at most, a constant number of bits bigger than the SCHC
   Packet that was received.  This assumes that the compute-*
   decompression actions produce a bounded number of bits, irrespective
   of the incoming SCHC Packet.  This property is true for IPv6 Length,
   UDP Length, and UDP Checksum, for which the compute-* CDA is
   recommended by this document.

   As a consequence, SCHC decompression does not amplify attacks, beyond
   adding a bounded number of bits to the SCHC Packet received.  This
   bound is determined by the Rule stored in the receiving device.

   As a general safety measure, a SCHC decompressor should never
   reconstruct a packet larger than MAX_PACKET_SIZE (defined in a
   Profile, with 1500 bytes as generic default).

12.1.2.  Compressed Packet Size as a Side Channel to Guess a Secret
         Token

   Some packet compression methods are known to be susceptible to
   attacks, such as BREACH and CRIME.  The attack involves injecting
   arbitrary data into the packet and observing the resulting compressed
   packet size.  The observed size potentially reflects correlation
   between the arbitrary data and some content that was meant to remain
   secret, such as a security token, thereby allowing the attacker to
   get at the secret.

   By contrast, SCHC compression takes place header field by header
   field, with the SCHC Packet being a mere concatenation of the
   compression residues of each of the individual field.  Any
   correlation between header fields does not result in a change in the
   SCHC Packet size compressed under the same Rule.

   If SCHC C/D is used to compress packets that include a secret
   information field, such as a token, the Rule set should be designed
   so that the size of the compression residue for the field to remain
   secret is the same irrespective of the value of the secret
   information.  This is achieved by, e.g., sending this field in
   extenso with the "ignore" MO and the "value-sent" CDA.  This
   recommendation is disputable if it is ascertained that the Rule set
   itself will remain secret.

12.1.3.  Decompressed Packet Different from the Original Packet

   As explained in Section 7.2, using FPs with value 0 in Field
   Descriptors in a Rule may result in header fields appearing in the
   decompressed packet in an order different from that in the original
   packet.  Likewise, as stated in Section 7.4.3, using an "ignore" MO
   together with a "not-sent" CDA will result in the header field taking
   the TV value, which is likely to be different from the original
   value.

   Depending on the protocol, the order of header fields in the packet
   may or may not be functionally significant.

   Furthermore, if the packet is protected by a checksum or a similar
   integrity protection mechanism, and if the checksum is transmitted
   instead of being recomputed as part of the decompression, these
   situations may result in the packet being considered corrupt and
   dropped.

12.2.  Security Considerations for SCHC Fragmentation/Reassembly

12.2.1.  Buffer Reservation Attack

   Let's assume that an attacker is able to send a forged SCHC Fragment
   to a SCHC reassembler.

   A node can perform a buffer reservation attack: the receiver will
   reserve buffer space for the SCHC Packet.  If the implementation has
   only one buffer, other incoming fragmented SCHC Packets will be
   dropped while the reassembly buffer is occupied during the reassembly
   timeout.  Once that timeout expires, the attacker can repeat the same
   procedure, and iterate, thus, creating a denial-of-service attack.
   An implementation may have multiple reassembly buffers.  The cost to
   mount this attack is linear with the number of buffers at the target
   node.  Better, the cost for an attacker can be increased if
   individual fragments of multiple SCHC Packets can be stored in the
   reassembly buffer.  The finer grained the reassembly buffer (down to
   the smallest tile size), the higher the cost of the attack.  If
   buffer overload does occur, a smart receiver could selectively
   discard SCHC Packets being reassembled based on the sender behavior,
   which may help identify which SCHC Fragments have been sent by the
   attacker.  Another mild countermeasure is for the target to abort the
   fragmentation/reassembly session as early as it detects a non-
   identical SCHC Fragment duplicate, anticipating for an eventual
   corrupt SCHC Packet, so as to save the sender the hassle of sending
   the rest of the fragments for this SCHC Packet.

12.2.2.  Corrupt Fragment Attack

   Let's assume that an attacker is able to send a forged SCHC Fragment
   to a SCHC reassembler.  The malicious node is additionally assumed to
   be able to hear an incoming communication destined to the target
   node.

   It can then send a forged SCHC Fragment that looks like it belongs to
   a SCHC Packet already being reassembled at the target node.  This can
   cause the SCHC Packet to be considered corrupt and to be dropped by
   the receiver.  The amplification happens here by a single spoofed
   SCHC Fragment rendering a full sequence of legitimate SCHC Fragments
   useless.  If the target uses ACK-Always or ACK-on-Error mode, such a
   malicious node can also interfere with the acknowledgement and
   repetition algorithm of SCHC F/R.  A single spoofed ACK, with all
   Bitmap bits set to 0, will trigger the repetition of WINDOW_SIZE
   tiles.  This protocol loop amplification depletes the energy source
   of the target node and consumes the channel bandwidth.  Similarly, a
   spoofed ACK REQ will trigger the sending of a SCHC ACK, which may be
   much larger than the ACK REQ if WINDOW_SIZE is large.  These
   consequences should be borne in mind when defining profiles for SCHC
   over specific LPWAN technologies.

12.2.3.  Fragmentation as a Way to Bypass Network Inspection

   Fragmentation is known for potentially allowing one to force through
   a Network Inspection device (e.g., firewall) packets that would be
   rejected if unfragmented.  This involves sending overlapping
   fragments to rewrite fields whose initial value led the Network
   Inspection device to allow the flow to go through.

   SCHC F/R is expected to be used over one LPWAN link, where no Network
   Inspection device is expected to sit.  As described in Section 5.2,
   even if the SCHC F/R on the Network Infrastructure side is located in
   the Internet, a tunnel is to be established between it and the NGW.

12.2.4.  Privacy Issues Associated with SCHC Header Fields

   SCHC F/R allocates a DTag value to fragments belonging to the same
   SCHC Packet.  Concerns were raised that, if DTag is a wide counter
   that is incremented in a predictable fashion for each new fragmented
   SCHC Packet, it might lead to a privacy issue, such as enabling
   tracking of a device across LPWANs.

   However, SCHC F/R is expected to be used over exactly one LPWAN link.
   As described in Section 5.2, even if the SCHC F/R on the Network
   Infrastructure side is located in the Internet, a tunnel is to be
   established between it and the NGW.  Therefore, assuming the tunnel
   provides confidentiality, neither the DTag field nor any other SCHC-
   introduced field is visible over the Internet.

13.  References

13.1.  Normative References

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,
              <https://www.rfc-editor.org/info/rfc2119>.

   [RFC6936]  Fairhurst, G. and M. Westerlund, "Applicability Statement
              for the Use of IPv6 UDP Datagrams with Zero Checksums",
              RFC 6936, DOI 10.17487/RFC6936, April 2013,
              <https://www.rfc-editor.org/info/rfc6936>.

   [RFC8174]  Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
              2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
              May 2017, <https://www.rfc-editor.org/info/rfc8174>.

   [RFC8200]  Deering, S. and R. Hinden, "Internet Protocol, Version 6
              (IPv6) Specification", STD 86, RFC 8200,
              DOI 10.17487/RFC8200, July 2017,
              <https://www.rfc-editor.org/info/rfc8200>.

   [RFC8376]  Farrell, S., Ed., "Low-Power Wide Area Network (LPWAN)
              Overview", RFC 8376, DOI 10.17487/RFC8376, May 2018,
              <https://www.rfc-editor.org/info/rfc8376>.

13.2.  Informative References

   [ETHERNET] IEEE, "IEEE Standard for Ethernet",
              DOI 10.1109/IEEESTD.2012.6419735, IEEE
              Standard 802.3-2012, December 2012,
              <https://ieeexplore.ieee.org/document/6419735>.

   [RFC4944]  Montenegro, G., Kushalnagar, N., Hui, J., and D. Culler,
              "Transmission of IPv6 Packets over IEEE 802.15.4
              Networks", RFC 4944, DOI 10.17487/RFC4944, September 2007,
              <https://www.rfc-editor.org/info/rfc4944>.

   [RFC5795]  Sandlund, K., Pelletier, G., and L-E. Jonsson, "The RObust
              Header Compression (ROHC) Framework", RFC 5795,
              DOI 10.17487/RFC5795, March 2010,
              <https://www.rfc-editor.org/info/rfc5795>.

   [RFC6282]  Hui, J., Ed. and P. Thubert, "Compression Format for IPv6
              Datagrams over IEEE 802.15.4-Based Networks", RFC 6282,
              DOI 10.17487/RFC6282, September 2011,
              <https://www.rfc-editor.org/info/rfc6282>.

   [RFC6437]  Amante, S., Carpenter, B., Jiang, S., and J. Rajahalme,
              "IPv6 Flow Label Specification", RFC 6437,
              DOI 10.17487/RFC6437, November 2011,
              <https://www.rfc-editor.org/info/rfc6437>.

   [RFC7136]  Carpenter, B. and S. Jiang, "Significance of IPv6
              Interface Identifiers", RFC 7136, DOI 10.17487/RFC7136,
              February 2014, <https://www.rfc-editor.org/info/rfc7136>.

   [RFC8065]  Thaler, D., "Privacy Considerations for IPv6 Adaptation-
              Layer Mechanisms", RFC 8065, DOI 10.17487/RFC8065,
              February 2017, <https://www.rfc-editor.org/info/rfc8065>.

Appendix A.  Compression Examples

   This section gives some scenarios of the compression mechanism for
   IPv6/UDP.  The goal is to illustrate the behavior of SCHC.

   The mechanisms defined in this document can be applied to a Dev that
   embeds some applications running over CoAP.  In this example, three
   flows are considered.  The first flow is for the device management
   based on CoAP using Link Local IPv6 addresses and UDP ports 123 and
   124 for Dev and App, respectively.  The second flow is a CoAP server
   for measurements done by the Dev (using ports 5683) and Global IPv6
   Address prefixes alpha::IID/64 to beta::1/64.  The last flow is for
   legacy applications using different ports numbers, the destination
   IPv6 address prefix is gamma::1/64.

   Figure 25 presents the protocol stack.  IPv6 and UDP are represented
   with dotted lines since these protocols are compressed on the radio
   link.

    Management   Data
   +----------+---------+---------+
   |   CoAP   |  CoAP   | legacy  |
   +----||----+---||----+---||----+
   .   UDP    .  UDP    |   UDP   |
   ................................
   .   IPv6   .  IPv6   .  IPv6   .
   +------------------------------+
   |    SCHC Header compression   |
   |      and fragmentation       |
   +------------------------------+
   |      LPWAN L2 technologies   |
   +------------------------------+
            Dev or NGW

              Figure 25: Simplified Protocol Stack for LP-WAN

   Rule 0
     Special RuleID used to tag an uncompressed UDP/IPV6 packet.

   Rule 1
    +----------------+--+--+--+---------+--------+------------++------+
    |       FID      |FL|FP|DI|    TV   |   MO   |     CDA    || Sent |
    |                |  |  |  |         |        |            ||[bits]|
    +----------------+--+--+--+---------+---------------------++------+
    |IPv6 Version    |4 |1 |Bi|6        | ignore | not-sent   ||      |
    |IPv6 Diffserv   |8 |1 |Bi|0        | equal  | not-sent   ||      |
    |IPv6 Flow Label |20|1 |Bi|0        | equal  | not-sent   ||      |
    |IPv6 Length     |16|1 |Bi|         | ignore | compute-*  ||      |
    |IPv6 Next Header|8 |1 |Bi|17       | equal  | not-sent   ||      |
    |IPv6 Hop Limit  |8 |1 |Bi|255      | ignore | not-sent   ||      |
    |IPv6 DevPrefix  |64|1 |Bi|FE80::/64| equal  | not-sent   ||      |
    |IPv6 DevIID     |64|1 |Bi|         | ignore | DevIID     ||      |
    |IPv6 AppPrefix  |64|1 |Bi|FE80::/64| equal  | not-sent   ||      |
    |IPv6 AppIID     |64|1 |Bi|::1      | equal  | not-sent   ||      |
    +================+==+==+==+=========+========+============++======+
    |UDP DevPort     |16|1 |Bi|123      | equal  | not-sent   ||      |
    |UDP AppPort     |16|1 |Bi|124      | equal  | not-sent   ||      |
    |UDP Length      |16|1 |Bi|         | ignore | compute-*  ||      |
    |UDP checksum    |16|1 |Bi|         | ignore | compute-*  ||      |
    +================+==+==+==+=========+========+============++======+

                Figure 26: Context Rules - Rule 0 and Rule 1

    Rule 2
    +----------------+--+--+--+---------+--------+------------++------+
    |       FID      |FL|FP|DI|    TV   |   MO   |     CDA    || Sent |
    |                |  |  |  |         |        |            ||[bits]|
    +----------------+--+--+--+---------+--------+------------++------+
    |IPv6 Version    |4 |1 |Bi|6        | ignore | not-sent   ||      |
    |IPv6 Diffserv   |8 |1 |Bi|0        | equal  | not-sent   ||      |
    |IPv6 Flow Label |20|1 |Bi|0        | equal  | not-sent   ||      |
    |IPv6 Length     |16|1 |Bi|         | ignore | compute-*  ||      |
    |IPv6 Next Header|8 |1 |Bi|17       | equal  | not-sent   ||      |
    |IPv6 Hop Limit  |8 |1 |Bi|255      | ignore | not-sent   ||      |
    |IPv6 DevPrefix  |64|1 |Bi|[alpha/64, match- |mapping-sent||   1  |
    |                |  |  |  |fe80::/64] mapping|            ||      |
    |IPv6 DevIID     |64|1 |Bi|         | ignore | DevIID     ||      |
    |IPv6 AppPrefix  |64|1 |Bi|[beta/64,| match- |mapping-sent||   2  |
    |                |  |  |  |alpha/64,| mapping|            ||      |
    |                |  |  |  |fe80::64]|        |            ||      |
    |IPv6 AppIID     |64|1 |Bi|::1000   | equal  | not-sent   ||      |
    +================+==+==+==+=========+========+============++======+
    |UDP DevPort     |16|1 |Bi|5683     | equal  | not-sent   ||      |
    |UDP AppPort     |16|1 |Bi|5683     | equal  | not-sent   ||      |
    |UDP Length      |16|1 |Bi|         | ignore | compute-*  ||      |
    |UDP checksum    |16|1 |Bi|         | ignore | compute-*  ||      |
    +================+==+==+==+=========+========+============++======+

                     Figure 27: Context Rules - Rule 2

    Rule 3
    +----------------+--+--+--+---------+--------+------------++------+
    |       FID      |FL|FP|DI|    TV   |   MO   |     CDA    || Sent |
    |                |  |  |  |         |        |            ||[bits]|
    +----------------+--+--+--+---------+--------+------------++------+
    |IPv6 Version    |4 |1 |Bi|6        | ignore | not-sent   ||      |
    |IPv6 Diffserv   |8 |1 |Bi|0        | equal  | not-sent   ||      |
    |IPv6 Flow Label |20|1 |Bi|0        | equal  | not-sent   ||      |
    |IPv6 Length     |16|1 |Bi|         | ignore | compute-*  ||      |
    |IPv6 Next Header|8 |1 |Bi|17       | equal  | not-sent   ||      |
    |IPv6 Hop Limit  |8 |1 |Up|255      | ignore | not-sent   ||      |
    |IPv6 Hop Limit  |8 |1 |Dw|         | ignore | value-sent ||   8  |
    |IPv6 DevPrefix  |64|1 |Bi|alpha/64 | equal  | not-sent   ||      |
    |IPv6 DevIID     |64|1 |Bi|         | ignore | DevIID     ||      |
    |IPv6 AppPrefix  |64|1 |Bi|gamma/64 | equal  | not-sent   ||      |
    |IPv6 AppIID     |64|1 |Bi|::1000   | equal  | not-sent   ||      |
    +================+==+==+==+=========+========+============++======+
    |UDP DevPort     |16|1 |Bi|8720     | MSB(12)| LSB        ||   4  |
    |UDP AppPort     |16|1 |Bi|8720     | MSB(12)| LSB        ||   4  |
    |UDP Length      |16|1 |Bi|         | ignore | compute-*  ||      |
    |UDP checksum    |16|1 |Bi|         | ignore | compute-*  ||      |
    +================+==+==+==+=========+========+============++======+

                     Figure 28: Context Rules - Rule 3

   Figures 26 to 28 describe an example of a Rule set.

   In this example, 0 was chosen as the special RuleID that tags packets
   that cannot be compressed with any compression Rule.

   All the fields described in Rules 1-3 are present in the IPv6 and UDP
   headers.  The DevIID value is inferred from the L2 header.

   Rules 2-3 use global addresses.  The way the Dev learns the prefix is
   not in the scope of the document.

   Rule 3 compresses each port number to 4 bits.

Appendix B.  Fragmentation Examples

   This section provides examples for the various fragment reliability
   modes specified in this document.  In the drawings, Bitmaps are shown
   in their uncompressed form.

   Figure 29 illustrates the transmission in No-ACK mode of a SCHC
   Packet that needs 11 SCHC Fragments.  FCN is 1 bit wide.

           Sender               Receiver
             |-------FCN=0-------->|
             |-------FCN=0-------->|
             |-------FCN=0-------->|
             |-------FCN=0-------->|
             |-------FCN=0-------->|
             |-------FCN=0-------->|
             |-------FCN=0-------->|
             |-------FCN=0-------->|
             |-------FCN=0-------->|
             |-------FCN=0-------->|
             |-----FCN=1 + RCS --->| Integrity check: success
           (End)

                 Figure 29: No-ACK Mode, 11 SCHC Fragments

   In the following examples, N (the size of the FCN field) is 3 bits.
   The All-1 FCN value is therefore 7.

   Figure 30 illustrates the transmission in ACK-on-Error mode of a SCHC
   Packet fragmented in 11 tiles, with one tile per SCHC Fragment,
   WINDOW_SIZE=7 and no lost SCHC Fragment.

           Sender               Receiver
             |-----W=0, FCN=6----->|
             |-----W=0, FCN=5----->|
             |-----W=0, FCN=4----->|
             |-----W=0, FCN=3----->|
             |-----W=0, FCN=2----->|
             |-----W=0, FCN=1----->|
             |-----W=0, FCN=0----->|
         (no ACK)
             |-----W=1, FCN=6----->|
             |-----W=1, FCN=5----->|
             |-----W=1, FCN=4----->|
             |--W=1, FCN=7 + RCS-->| Integrity check: success
             |<-- ACK, W=1, C=1 ---| C=1
           (End)

         Figure 30: ACK-on-Error Mode, 11 Tiles, One Tile per SCHC
                      Fragment, No Lost SCHC Fragment

   Figure 31 illustrates the transmission in ACK-on-Error mode of a SCHC
   Packet fragmented in 11 tiles, with one tile per SCHC Fragment,
   WINDOW_SIZE=7, and three lost SCHC Fragments.

           Sender               Receiver
             |-----W=0, FCN=6----->|
             |-----W=0, FCN=5----->|
             |-----W=0, FCN=4--X-->|
             |-----W=0, FCN=3----->|
             |-----W=0, FCN=2--X-->|
             |-----W=0, FCN=1----->|
             |-----W=0, FCN=0----->|        6543210
             |<-- ACK, W=0, C=0 ---| Bitmap:1101011
             |-----W=0, FCN=4----->|
             |-----W=0, FCN=2----->|
         (no ACK)
             |-----W=1, FCN=6----->|
             |-----W=1, FCN=5----->|
             |-----W=1, FCN=4--X-->|
             |- W=1, FCN=7 + RCS ->| Integrity check: failure
             |<-- ACK, W=1, C=0 ---| C=0, Bitmap:1100001
             |-----W=1, FCN=4----->| Integrity check: success
             |<-- ACK, W=1, C=1 ---| C=1
           (End)

         Figure 31: ACK-on-Error Mode, 11 Tiles, One Tile per SCHC
                       Fragment, Lost SCHC Fragments

   Figure 32 shows an example of a transmission in ACK-on-Error mode of
   a SCHC Packet fragmented in 73 tiles, with N=5, WINDOW_SIZE=28, M=2,
   and three lost SCHC Fragments.

      Sender               Receiver
       |-----W=0, FCN=27----->| 4 tiles sent
       |-----W=0, FCN=23----->| 4 tiles sent
       |-----W=0, FCN=19----->| 4 tiles sent
       |-----W=0, FCN=15--X-->| 4 tiles sent (not received)
       |-----W=0, FCN=11----->| 4 tiles sent
       |-----W=0, FCN=7 ----->| 4 tiles sent
       |-----W=0, FCN=3 ----->| 4 tiles sent
       |-----W=1, FCN=27----->| 4 tiles sent
       |-----W=1, FCN=23----->| 4 tiles sent
       |-----W=1, FCN=19----->| 4 tiles sent
       |-----W=1, FCN=15----->| 4 tiles sent
       |-----W=1, FCN=11----->| 4 tiles sent
       |-----W=1, FCN=7 ----->| 4 tiles sent
       |-----W=1, FCN=3 --X-->| 4 tiles sent (not received)
       |-----W=2, FCN=27----->| 4 tiles sent
       |-----W=2, FCN=23----->| 4 tiles sent
   ^   |-----W=2, FCN=19----->| 1 tile sent
   |   |-----W=2, FCN=18----->| 1 tile sent
   |   |-----W=2, FCN=17----->| 1 tile sent
       |-----W=2, FCN=16----->| 1 tile sent
   s   |-----W=2, FCN=15----->| 1 tile sent
   m   |-----W=2, FCN=14----->| 1 tile sent
   a   |-----W=2, FCN=13--X-->| 1 tile sent (not received)
   l   |-----W=2, FCN=12----->| 1 tile sent
   l   |---W=2, FCN=31 + RCS->| Integrity check: failure
   e   |<--- ACK, W=0, C=0 ---| C=0, Bitmap:1111111111110000111111111111
   r   |-----W=0, FCN=15----->| 1 tile sent
       |-----W=0, FCN=14----->| 1 tile sent
   L   |-----W=0, FCN=13----->| 1 tile sent
   2   |-----W=0, FCN=12----->| 1 tile sent
       |<--- ACK, W=1, C=0 ---| C=0, Bitmap:1111111111111111111111110000
   M   |-----W=1, FCN=3 ----->| 1 tile sent
   T   |-----W=1, FCN=2 ----->| 1 tile sent
   U   |-----W=1, FCN=1 ----->| 1 tile sent
       |-----W=1, FCN=0 ----->| 1 tile sent
   |   |<--- ACK, W=2, C=0 ---| C=0, Bitmap:1111111111111101000000000001
   |   |-----W=2, FCN=13----->| Integrity check: success
   V   |<--- ACK, W=2, C=1 ---| C=1
     (End)

                 Figure 32: ACK-on-Error Mode, Variable MTU

   In this example, the L2 MTU becomes reduced just before sending the
   "W=2, FCN=19" fragment, leaving space for only one tile in each
   forthcoming SCHC Fragment.  Before retransmissions, the 73 tiles are
   carried by a total of 25 SCHC Fragments, the last nine being of
   smaller size.

   Note: other sequences of events (e.g., regarding when ACKs are sent
   by the Receiver) are also allowed by this specification.  Profiles
   may restrict this flexibility.

   Figure 33 illustrates the transmission in ACK-Always mode of a SCHC
   Packet fragmented in 11 tiles, with one tile per SCHC Fragment, with
   N=3, WINDOW_SIZE=7, and no loss.

           Sender               Receiver
             |-----W=0, FCN=6----->|
             |-----W=0, FCN=5----->|
             |-----W=0, FCN=4----->|
             |-----W=0, FCN=3----->|
             |-----W=0, FCN=2----->|
             |-----W=0, FCN=1----->|
             |-----W=0, FCN=0----->|
             |<-- ACK, W=0, C=0 ---| Bitmap:1111111
             |-----W=1, FCN=6----->|
             |-----W=1, FCN=5----->|
             |-----W=1, FCN=4----->|
             |--W=1, FCN=7 + RCS-->| Integrity check: success
             |<-- ACK, W=1, C=1 ---| C=1
           (End)

     Figure 33: ACK-Always Mode, 11 Tiles, One Tile per SCHC Fragment,
                                  No Loss

   Figure 34 illustrates the transmission in ACK-Always mode of a SCHC
   Packet fragmented in 11 tiles, with one tile per SCHC Fragment, N=3,
   WINDOW_SIZE=7 and three lost SCHC Fragments.

           Sender               Receiver
             |-----W=0, FCN=6----->|
             |-----W=0, FCN=5----->|
             |-----W=0, FCN=4--X-->|
             |-----W=0, FCN=3----->|
             |-----W=0, FCN=2--X-->|
             |-----W=0, FCN=1----->|
             |-----W=0, FCN=0----->|        6543210
             |<-- ACK, W=0, C=0 ---| Bitmap:1101011
             |-----W=0, FCN=4----->|
             |-----W=0, FCN=2----->|
             |<-- ACK, W=0, C=0 ---| Bitmap:1111111
             |-----W=1, FCN=6----->|
             |-----W=1, FCN=5----->|
             |-----W=1, FCN=4--X-->|
             |--W=1, FCN=7 + RCS-->| Integrity check: failure
             |<-- ACK, W=1, C=0 ---| C=0, Bitmap:11000001
             |-----W=1, FCN=4----->| Integrity check: success
             |<-- ACK, W=1, C=1 ---| C=1
           (End)

     Figure 34: ACK-Always Mode, 11 Tiles, One Tile per SCHC Fragment,
                         Three Lost SCHC Fragments

   Figure 35 illustrates the transmission in ACK-Always mode of a SCHC
   Packet fragmented in six tiles, with one tile per SCHC Fragment, N=3,
   WINDOW_SIZE=7, three lost SCHC Fragments, and only one retry needed
   to recover each lost SCHC Fragment.

             Sender                Receiver
                |-----W=0, FCN=6----->|
                |-----W=0, FCN=5----->|
                |-----W=0, FCN=4--X-->|
                |-----W=0, FCN=3--X-->|
                |-----W=0, FCN=2--X-->|
                |--W=0, FCN=7 + RCS-->| Integrity check: failure
                |<-- ACK, W=0, C=0 ---| C=0, Bitmap:1100001
                |-----W=0, FCN=4----->| Integrity check: failure
                |-----W=0, FCN=3----->| Integrity check: failure
                |-----W=0, FCN=2----->| Integrity check: success
                |<-- ACK, W=0, C=1 ---| C=1
              (End)

          Figure 35: ACK-Always Mode, Six Tiles, One Tile per SCHC
                    Fragment, Three Lost SCHC Fragments

   Figure 36 illustrates the transmission in ACK-Always mode of a SCHC
   Packet fragmented in six tiles, with one tile per SCHC Fragment, N=3,
   WINDOW_SIZE=7, three lost SCHC Fragments, and the second SCHC ACK
   lost.

             Sender                Receiver
                |-----W=0, FCN=6----->|
                |-----W=0, FCN=5----->|
                |-----W=0, FCN=4--X-->|
                |-----W=0, FCN=3--X-->|
                |-----W=0, FCN=2--X-->|
                |--W=0, FCN=7 + RCS-->| Integrity check: failure
                |<-- ACK, W=0, C=0 ---| C=0, Bitmap:1100001
                |-----W=0, FCN=4----->| Integrity check: failure
                |-----W=0, FCN=3----->| Integrity check: failure
                |-----W=0, FCN=2----->| Integrity check: success
                |<-X-ACK, W=0, C=1 ---| C=1
       timeout  |                     |
                |--- W=0, ACK REQ --->| ACK REQ
                |<-- ACK, W=0, C=1 ---| C=1
              (End)

          Figure 36: ACK-Always Mode, Six Tiles, One Tile per SCHC
                          Fragment, SCHC ACK Loss

   Figure 37 illustrates the transmission in ACK-Always mode of a SCHC
   Packet fragmented in six tiles, with N=3, WINDOW_SIZE=7, with three
   lost SCHC Fragments, and one retransmitted SCHC Fragment lost again.

              Sender                Receiver
                |-----W=0, FCN=6----->|
                |-----W=0, FCN=5----->|
                |-----W=0, FCN=4--X-->|
                |-----W=0, FCN=3--X-->|
                |-----W=0, FCN=2--X-->|
                |--W=0, FCN=7 + RCS-->| Integrity check: failure
                |<-- ACK, W=0, C=0 ---| C=0, Bitmap:1100001
                |-----W=0, FCN=4----->| Integrity check: failure
                |-----W=0, FCN=3----->| Integrity check: failure
                |-----W=0, FCN=2--X-->|
         timeout|                     |
                |--- W=0, ACK REQ --->| ACK REQ
                |<-- ACK, W=0, C=0 ---| C=0, Bitmap: 1111101
                |-----W=0, FCN=2----->| Integrity check: success
                |<-- ACK, W=0, C=1 ---| C=1
              (End)

         Figure 37: ACK-Always Mode, Six Tiles, Retransmitted SCHC
                            Fragment Lost Again

   Figure 38 illustrates the transmission in ACK-Always mode of a SCHC
   Packet fragmented in 28 tiles, with one tile per SCHC Fragment, N=5,
   WINDOW_SIZE=24, and two lost SCHC Fragments.

         Sender               Receiver
           |-----W=0, FCN=23----->|
           |-----W=0, FCN=22----->|
           |-----W=0, FCN=21--X-->|
           |-----W=0, FCN=20----->|
           |-----W=0, FCN=19----->|
           |-----W=0, FCN=18----->|
           |-----W=0, FCN=17----->|
           |-----W=0, FCN=16----->|
           |-----W=0, FCN=15----->|
           |-----W=0, FCN=14----->|
           |-----W=0, FCN=13----->|
           |-----W=0, FCN=12----->|
           |-----W=0, FCN=11----->|
           |-----W=0, FCN=10--X-->|
           |-----W=0, FCN=9 ----->|
           |-----W=0, FCN=8 ----->|
           |-----W=0, FCN=7 ----->|
           |-----W=0, FCN=6 ----->|
           |-----W=0, FCN=5 ----->|
           |-----W=0, FCN=4 ----->|
           |-----W=0, FCN=3 ----->|
           |-----W=0, FCN=2 ----->|
           |-----W=0, FCN=1 ----->|
           |-----W=0, FCN=0 ----->|
           |                      |
           |<--- ACK, W=0, C=0 ---| Bitmap:110111111111101111111111
           |-----W=0, FCN=21----->|
           |-----W=0, FCN=10----->|
           |<--- ACK, W=0, C=0 ---| Bitmap:111111111111111111111111
           |-----W=1, FCN=23----->|
           |-----W=1, FCN=22----->|
           |-----W=1, FCN=21----->|
           |--W=1, FCN=31 + RCS-->| Integrity check: success
           |<--- ACK, W=1, C=1 ---| C=1
         (End)

     Figure 38: ACK-Always Mode, 28 Tiles, One Tile per SCHC Fragment,
                            Lost SCHC Fragments

Appendix C.  Fragmentation State Machines

   The fragmentation state machines of the sender and the receiver, one
   for each of the different reliability modes, are described in the
   following figures:

                +===========+
   +------------+  Init     |
   |  FCN=0     +===========+
   |  No Window
   |  No Bitmap
   |                   +-------+
   |          +========+==+    | More Fragments
   |          |           | <--+ ~~~~~~~~~~~~~~~~~~~~
   +--------> |   Send    |      send Fragment (FCN=0)
              +===+=======+
                  |  last fragment
                  |  ~~~~~~~~~~~~
                  |  FCN = 1
                  v  send fragment+RCS
              +============+
              |    END     |
              +============+

            Figure 39: Sender State Machine for the No-ACK Mode

                         +------+ Not All-1
              +==========+=+    | ~~~~~~~~~~~~~~~~~~~
              |            + <--+ set Inactivity Timer
              |  RCV Frag  +-------+
              +=+===+======+       |All-1 &
      All-1 &   |   |              |RCS correct
    RCS wrong   |   |Inactivity    |
                |   |Timer Exp.    |
                v   |              |
     +==========++  |              v
     |   Error   |<-+     +========+==+
     +===========+        |    END    |
                          +===========+

           Figure 40: Receiver State Machine for the No-ACK Mode

                 +=======+
                 | INIT  |       FCN!=0 & more frags
                 |       |       ~~~~~~~~~~~~~~~~~~~~~~
                 +======++  +--+ send Window + frag(FCN)
                    W=0 |   |  | FCN-
     Clear lcl_bm       |   |  v set lcl_bm
          FCN=max value |  ++==+========+
                        +> |            |
   +---------------------> |    SEND    |
   |                       +==+===+=====+
   |      FCN==0 & more frags |   | last frag
   |    ~~~~~~~~~~~~~~~~~~~~~ |   | ~~~~~~~~~~~~~~~
   |               set lcl_bm |   | set lcl_bm
   |   send wnd + frag(all-0) |   | send wnd+frag(all-1)+RCS
   |       set Retrans_Timer  |   | set Retrans_Timer
   |                          |   |
   |Recv_wnd == wnd &         |   |
   |lcl_bm==recv_bm &         |   |  +----------------------+
   |more frag                 |   |  | lcl_bm!=rcv-bm       |
   |~~~~~~~~~~~~~~~~~~~~~~    |   |  | ~~~~~~~~~            |
   |Stop Retrans_Timer        |   |  | Attempt++            v
   |clear lcl_bm              v   v  |                +=====+=+
   |window=next_window   +====+===+==+===+            |Resend |
   +---------------------+               |            |Missing|
                    +----+     Wait      |            |Frag   |
   not expected wnd |    |    Bitmap     |            +=======+
   ~~~~~~~~~~~~~~~~ +--->+               ++Retrans_Timer Exp  |
       discard frag      +==+=+===+=+==+=+| ~~~~~~~~~~~~~~~~~ |
                            | |   | ^  ^  |reSend(empty)All-* |
                            | |   | |  |  |Set Retrans_Timer  |
                            | |   | |  +--+Attempt++          |
     C_bit==1 &             | |   | +-------------------------+
   Recv_window==window &    | |   |   all missing frags sent
                no more frag| |   |   ~~~~~~~~~~~~~~~~~~~~~~
    ~~~~~~~~~~~~~~~~~~~~~~~~| |   |   Set Retrans_Timer
          Stop Retrans_Timer| |   |
    +=============+         | |   |
    |     END     +<--------+ |   |
    +=============+           |   | Attempt > MAX_ACK_REQUESTS
               All-1 Window & |   | ~~~~~~~~~~~~~~~~~~
                  C_bit ==0 & |   v Send Abort
             lcl_bm==recv_bm  | +=+===========+
                 ~~~~~~~~~~~~ +>|    ERROR    |
                   Send Abort   +=============+

          Figure 41: Sender State Machine for the ACK-Always Mode

    Not All- & w=expected +---+   +---+w = Not expected
    ~~~~~~~~~~~~~~~~~~~~~ |   |   |   |~~~~~~~~~~~~~~~~
    Set lcl_bm(FCN)       |   v   v   |discard
                         ++===+===+===+=+
   +---------------------+     Rcv      +--->* ABORT
   |  +------------------+   Window     |
   |  |                  +=====+==+=====+
   |  |       All-0 & w=expect |  ^ w =next & not-All
   |  |     ~~~~~~~~~~~~~~~~~~ |  |~~~~~~~~~~~~~~~~~~~~~
   |  |    set lcl_bm(FCN)     |  |expected = next window
   |  |      send lcl_bm       |  |Clear lcl_bm
   |  |                        |  |
   |  | w=expected & not-All   |  |
   |  | ~~~~~~~~~~~~~~~~~~     |  |
   |  |     set lcl_bm(FCN)+-+ |  | +--+ w=next & All-0
   |  |     if lcl_bm full | | |  | |  | ~~~~~~~~~~~~~~~
   |  |     send lcl_bm    | | |  | |  | expected = nxt wnd
   |  |                    v | v  | |  | Clear lcl_bm
   |  |w=expected& All-1 +=+=+=+==+=++ | set lcl_bm(FCN)
   |  |  ~~~~~~~~~~~  +->+    Wait   +<+ send lcl_bm
   |  |    discard    +--|    Next   |
   |  | All-0  +---------+  Window   +--->* ABORT
   |  | ~~~~~  +-------->+========+=++
   |  | snd lcl_bm  All-1 & w=next| |  All-1 & w=nxt
   |  |                & RCS wrong| |  & RCS right
   |  |          ~~~~~~~~~~~~~~~~~| | ~~~~~~~~~~~~~~~~~~
   |  |            set lcl_bm(FCN)| |set lcl_bm(FCN)
   |  |                send lcl_bm| |send lcl_bm
   |  |                           | +----------------------+
   |  |All-1 & w=expected         |                        |
   |  |& RCS wrong                v   +---+ w=expected &   |
   |  |~~~~~~~~~~~~~~~~~~~~  +====+=====+ | RCS wrong      |
   |  |set lcl_bm(FCN)       |          +<+ ~~~~~~~~~~~~~~ |
   |  |send lcl_bm           | Wait End |   set lcl_bm(FCN)|
   |  +--------------------->+          +--->* ABORT       |
   |                         +===+====+=+-+ All-1&RCS wrong|
   |                             |    ^   | ~~~~~~~~~~~~~~~|
   |      w=expected & RCS right |    +---+   send lcl_bm  |
   |      ~~~~~~~~~~~~~~~~~~~~~~ |                         |
   |       set lcl_bm(FCN)       | +-+ Not All-1           |
   |        send lcl_bm          | | | ~~~~~~~~~           |
   |                             | | |  discard            |
   |All-1&w=expected & RCS right | | |                     |
   |~~~~~~~~~~~~~~~~~~~~~~~~~~~~ v | v +----+All-1         |
   |set lcl_bm(FCN)            +=+=+=+=+==+ |~~~~~~~~~     |
   |send lcl_bm                |          +<+Send lcl_bm   |
   +-------------------------->+    END   |                |
                               +==========+<---------------+

          --->* ABORT

          In any state
             on receiving a SCHC ACK REQ
                Send a SCHC ACK for the current window

         Figure 42: Receiver State Machine for the ACK-Always Mode

                     +=======+
                     |       |
                     | INIT  |
                     |       |       FCN!=0 & more frags
                     +======++       ~~~~~~~~~~~~~~~~~~~~~~
        Frag RuleID trigger |   +--+ Send cur_W + frag(FCN);
        ~~~~~~~~~~~~~~~~~~~ |   |  | FCN--;
     cur_W=0; FCN=max_value;|   |  | set [cur_W, cur_Bmp]
       clear [cur_W, Bmp_n];|   |  v
             clear rcv_Bmp  |  ++==+==========+       **BACK_TO_SEND
                            +->+              |   cur_W==rcv_W &
         **BACK_TO_SEND        |     SEND     |   [cur_W,Bmp_n]==rcv_Bmp
   +-------------------------->+              |   & more frags
   |  +----------------------->+              |   ~~~~~~~~~~~~
   |  |                        ++==+==========+   cur_W++;
   |  |      FCN==0 & more frags|  |last frag     clear [cur_W, Bmp_n]
   |  |  ~~~~~~~~~~~~~~~~~~~~~~~|  |~~~~~~~~~
   |  |        set cur_Bmp;     |  |set [cur_W, Bmp_n];
   |  |send cur_W + frag(All-0);|  |send cur_W + frag(All-1)+RCS;
   |  |        set Retrans_Timer|  |set Retrans_Timer
   |  |                         |  | +---------------------------------+
   |  |                         |  | |cur_W ==                         |
   |  |Retrans_Timer expires &  |  | |   rcv_W & [cur_W,Bmp_n]!=rcv_Bmp|
   |  |more Frags               |  | |  ~~~~~~~~~~~~~~~~~~~            |
   |  |~~~~~~~~~~~~~~~~~~~~     |  | |  Attempts++; W=cur_W            |
   |  |stop Retrans_Timer;      |  | | +--------+           rcv_W==Wn &|
   |  |[cur_W,Bmp_n]==cur_Bmp;  v  v | |        v   [Wn,Bmp_n]!=rcv_Bmp|
   |  |cur_W++            +=====+==+=+=+==+   +=+=========+ ~~~~~~~~~~~|
   |  +-------------------+               |   | Resend    | Attempts++;|
   +----------------------+   Wait x ACK  |   | Missing   |       W=Wn |
   +--------------------->+               |   | Frags(W)  +<-----------+
   |         rcv_W==Wn &+-+               |   +======+====+
   | [Wn,Bmp_n]!=rcv_Bmp| ++=+===+===+==+=+          |
   |      ~~~~~~~~~~~~~~|  ^ |   |   |  ^            |
   |        send (cur_W,+--+ |   |   |  +------------+
   |        ALL-0-empty)     |   |   |     all missing frag sent(W)
   |                         |   |   |     ~~~~~~~~~~~~~~~~~
   |  Retrans_Timer expires &|   |   |     set Retrans_Timer
   |            No more Frags|   |   |
   |           ~~~~~~~~~~~~~~|   |   |
   |      stop Retrans_Timer;|   |   |
   |(re)send frag(All-1)+RCS |   |   |
   +-------------------------+   |   |
                    cur_W==rcv_W&|   |
          [cur_W,Bmp_n]==rcv_Bmp&|   | Attempts > MAX_ACK_REQUESTS
     No more Frags & RCS flag==OK|   | ~~~~~~~~~~
               ~~~~~~~~~~~~~~~~~~|   | send Abort
    +=========+stop Retrans_Timer|   |  +===========+
    |   END   +<-----------------+   +->+   ERROR   |
    +=========+                         +===========+

         Figure 43: Sender State Machine for the ACK-on-Error Mode

   This is an example only.  It is not normative.  The specification in
   Section 8.4.3.1 allows for sequences of operations different from the
   one shown here.

                    +=======+        New frag RuleID received
                    |       |        ~~~~~~~~~~~~~
                    | INIT  +-------+cur_W=0;clear([cur_W,Bmp_n]);
                    +=======+       |sync=0
                                    |
       Not All* & rcv_W==cur_W+---+ | +--+
         ~~~~~~~~~~~~~~~~~~~~ |   | | | (E)
         set[cur_W,Bmp_n(FCN)]|   v v v  |
                             ++===+=+=+==+=+
      +----------------------+             +--+ All-0&Full[cur_W,Bmp_n]
      |           ABORT *<---+  Rcv Window |  | ~~~~~~~~~~
      |  +-------------------+             +<-+ cur_W++;set Inact_timer;
      |  |                +->+=+=+=+=+=+===+    clear [cur_W,Bmp_n]
      |  | All-0 empty(Wn)|    | | | ^ ^
      |  | ~~~~~~~~~~~~~~ +----+ | | | |rcv_W==cur_W & sync==0;
      |  | sendACK([Wn,Bmp_n])   | | | |& Full([cur_W,Bmp_n])
      |  |                       | | | |& All* || last_miss_frag
      |  |                       | | | |~~~~~~~~~~~~~~~~~~~~~~
      |  |    All* & rcv_W==cur_W|(C)| |sendACK([cur_W,Bmp_n]);
      |  |              & sync==0| | | |cur_W++; clear([cur_W,Bmp_n])
      |  |&no_full([cur_W,Bmp_n])| |(E)|
      |  |      ~~~~~~~~~~~~~~~~ | | | |              +========+
      |  | sendACK([cur_W,Bmp_n])| | | |              | Error/ |
      |  |                       | | | |   +----+     | Abort  |
      |  |                       v v | |   |    |     +===+====+
      |  |                   +===+=+=+=+===+=+ (D)        ^
      |  |                +--+    Wait x     |  |         |
      |  | All-0 empty(Wn)+->| Missing Frags |<-+         |
      |  | ~~~~~~~~~~~~~~    +=============+=+            |
      |  | sendACK([Wn,Bmp_n])             +--------------+
      |  |                                       *ABORT
      v  v
     (A)(B)
                                      (D) All* || last_miss_frag
       (C) All* & sync>0                  & rcv_W!=cur_W & sync>0
           ~~~~~~~~~~~~                   & Full([rcv_W,Bmp_n])
           Wn=oldest[not full(W)];        ~~~~~~~~~~~~~~~~~~~~
           sendACK([Wn,Bmp_n])            Wn=oldest[not full(W)];
                                          sendACK([Wn,Bmp_n]);sync--

                                ABORT-->* Uplink Only &
                                          Inact_Timer expires
       (E) Not All* & rcv_W!=cur_W        || Attempts > MAX_ACK_REQUESTS
           ~~~~~~~~~~~~~~~~~~~~           ~~~~~~~~~~~~~~~~~~~~~
           sync++; cur_W=rcv_W;           send Abort
           set[cur_W,Bmp_n(FCN)]

     (A)(B)
      |  |
      |  | All-1 & rcv_W==cur_W & RCS!=OK        All-0 empty(Wn)
      |  | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~     +-+  ~~~~~~~~~~
      |  | sendACK([cur_W,Bmp_n],C=0)       | v  sendACK([Wn,Bmp_n])
      |  |                      +===========+=++
      |  +--------------------->+   Wait End   +-+
      |                         +=====+=+====+=+ | All-1
      |     rcv_W==cur_W & RCS==OK    | |    ^   | & rcv_W==cur_W
      |     ~~~~~~~~~~~~~~~~~~~~~~    | |    +---+ & RCS!=OK
      |  sendACK([cur_W,Bmp_n],C=1)   | |          ~~~~~~~~~~~~~~~~~~~
      |                               | | sendACK([cur_W,Bmp_n],C=0);
      |                               | |          Attempts++
      |All-1 & Full([cur_W,Bmp_n])    | |
      |& RCS==OK & sync==0            | +-->* ABORT
      |~~~~~~~~~~~~~~~~~~~            v
      |sendACK([cur_W,Bmp_n],C=1)   +=+=========+
      +---------------------------->+    END    |
                                    +===========+

        Figure 44: Receiver State Machine for the ACK-on-Error Mode

Appendix D.  SCHC Parameters

   This section lists the information that needs to be provided in the
   LPWAN technology-specific documents.

   *  Most common uses cases, deployment scenarios.

   *  Mapping of the SCHC architectural elements onto the LPWAN
      architecture.

   *  Assessment of LPWAN integrity checking.

   *  Various potential channel conditions for the technology and the
      corresponding recommended use of SCHC C/D and SCHC F/R.

   This section lists the parameters that need to be defined in the
   Profile.

   *  RuleID numbering scheme, fixed-size or variable-size RuleIDs,
      number of Rules, the way the RuleID is transmitted.

   *  maximum packet size that should ever be reconstructed by SCHC
      decompression (MAX_PACKET_SIZE).  See Section 12.

   *  Padding: size of the L2 Word (for most LPWAN technologies, this
      would be a byte; for some technologies, a bit).

   *  Decision to use SCHC fragmentation mechanism or not.  If yes, the
      document must describe:

      -  reliability mode(s) used, in which cases (e.g., based on link
         channel condition).

      -  RuleID values assigned to each mode in use.

      -  presence and number of bits for DTag (T) for each RuleID value,
         lifetime of DTag at the receiver.

      -  support for interleaved packet transmission, to what extent.

      -  WINDOW_SIZE, for modes that use windows.

      -  number of bits for W (M) for each RuleID value, for modes that
         use windows.

      -  number of bits for FCN (N) for each RuleID value, meaning of
         the FCN values.

      -  what makes an All-0 SCHC Fragment and a SCHC ACK REQ
         distinguishable (see Section 8.3.1.1).

      -  what makes an All-1 SCHC Fragment and a SCHC Sender-Abort
         distinguishable (see Section 8.3.1.2).

      -  for RuleIDs that use ACK-on-Error mode: when the last tile of a
         SCHC Packet is to be sent in a Regular SCHC Fragment, alone in
         an All-1 SCHC Fragment or with any of these two methods.

      -  for RuleIDs that use ACK-on-Error mode: if the penultimate tile
         of a SCHC Packet is of the regular size only or if it can also
         be one L2 Word shorter.

      -  for RuleIDs that use ACK-on-Error mode: times at which the
         sender must listen for SCHC ACKs.

      -  size of RCS and algorithm for its computation, for each RuleID,
         if different from the default CRC32.  Byte fill-up with zeroes
         or other mechanism, to be specified.  Support for UDP checksum
         elision.

      -  Retransmission Timer duration for each RuleID value, if
         applicable to the SCHC F/R mode.

      -  Inactivity Timer duration for each RuleID value, if applicable
         to the SCHC F/R mode.

      -  MAX_ACK_REQUESTS value for each RuleID value, if applicable to
         the SCHC F/R mode.

   *  if L2 Word is wider than a bit and SCHC fragmentation is used,
      value of the padding bits (0 or 1).

   A Profile may define a delay to be added after each SCHC message
   transmission for compliance with local regulations or other
   constraints imposed by the applications.

   *  In some LPWAN technologies, as part of energy-saving techniques,
      Downlink transmission is only possible immediately after an Uplink
      transmission.  In order to avoid potentially high delay in the
      Downlink transmission of a fragmented SCHC Packet, the SCHC
      Fragment receiver may perform an Uplink transmission as soon as
      possible after reception of a SCHC Fragment that is not the last
      one.  Such Uplink transmission may be triggered by the L2 (e.g.,
      an L2 ACK sent in response to a SCHC Fragment encapsulated in a L2
      PDU that requires an L2 ACK) or it may be triggered from an upper
      layer.  See Appendix F.

   *  the following parameters need to be addressed in documents other
      than this one but not necessarily in the LPWAN technology-specific
      documents:

      -  The way the Contexts are provisioned.

      -  The way the Rules are generated.

Appendix E.  Supporting Multiple Window Sizes for Fragmentation

   For ACK-Always or ACK-on-Error, implementers may opt to support a
   single window size or multiple window sizes.  The latter, when
   feasible, may provide performance optimizations.  For example, a
   large WINDOW_SIZE should be used for packets that need to be split
   into a large number of tiles.  However, when the number of tiles
   required to carry a packet is low, a smaller WINDOW_SIZE and, thus, a
   shorter Bitmap, may be sufficient to provide reception status on all
   tiles.  If multiple window sizes are supported, the RuleID signals
   what WINDOW_SIZE is in use for a specific packet transmission.

Appendix F.  ACK-Always and ACK-on-Error on Quasi-Bidirectional Links

   The ACK-Always and ACK-on-Error modes of SCHC F/R are bidirectional
   protocols: they require a feedback path from the reassembler to the
   fragmenter.

   Some LPWAN technologies provide quasi-bidirectional connectivity,
   whereby a Downlink transmission from the Network Infrastructure can
   only take place right after an Uplink transmission by the Dev.

   When using SCHC F/R to send fragmented SCHC Packets Downlink over
   these quasi-bidirectional links, the following situation may arise:
   if an Uplink SCHC ACK is lost, the SCHC ACK REQ message by the sender
   could be stuck indefinitely in the Downlink queue at the Network
   Infrastructure, waiting for a transmission opportunity.

   There are many ways by which this deadlock can be avoided.  The Dev
   application might be sending recurring Uplink messages such as keep-
   alive, or the Dev application stack might be sending other recurring
   Uplink messages as part of its operation.  However, these are out of
   the control of this generic SCHC specification.

   In order to cope with quasi-bidirectional links, a SCHC-over-foo
   specification may want to amend the SCHC F/R specification to add a
   timer-based retransmission of the SCHC ACK.  Below is an example of
   the suggested behavior for ACK-Always mode.  Because it is an
   example, [RFC2119] language is deliberately not used here.

   For Downlink transmission of a fragmented SCHC Packet in ACK-Always
   mode, the SCHC Fragment receiver may support timer-based SCHC ACK
   retransmission.  In this mechanism, the SCHC Fragment receiver
   initializes and starts a timer (the UplinkACK Timer) after the
   transmission of a SCHC ACK, except when the SCHC ACK is sent in
   response to the last SCHC Fragment of a packet (All-1 fragment).  In
   the latter case, the SCHC Fragment receiver does not start a timer
   after transmission of the SCHC ACK.

   If, after transmission of a SCHC ACK that is not an All-1 fragment,
   and before expiration of the corresponding UplinkACK timer, the SCHC
   Fragment receiver receives a SCHC Fragment that belongs to the
   current window (e.g., a missing SCHC Fragment from the current
   window) or to the next window, the UplinkACK timer for the SCHC ACK
   is stopped.  However, if the UplinkACK timer expires, the SCHC ACK is
   resent and the UplinkACK timer is reinitialized and restarted.

   The default initial value for the UplinkACK Timer, as well as the
   maximum number of retries for a specific SCHC ACK, denoted
   MAX_ACK_REQUESTS, is to be defined in a Profile.  The initial value
   of the UplinkACK timer is expected to be greater than that of the
   Retransmission timer, in order to make sure that a (buffered) SCHC
   Fragment to be retransmitted finds an opportunity for that
   transmission.  One exception to this recommendation is the special
   case of the All-1 SCHC Fragment transmission.

   When the SCHC Fragment sender transmits the All-1 SCHC Fragment, it
   starts its Retransmission Timer with a large timeout value (e.g.,
   several times that of the initial UplinkACK Timer).  If a SCHC ACK is
   received before expiration of this timer, the SCHC Fragment sender
   retransmits any lost SCHC Fragments as reported by the SCHC ACK, or
   if the SCHC ACK confirms successful reception of all SCHC Fragments
   of the last window, the transmission of the fragmented SCHC Packet is
   considered complete.  If the timer expires, and no SCHC ACK has been
   received since the start of the timer, the SCHC Fragment sender
   assumes that the All-1 SCHC Fragment has been successfully received
   (and possibly, the last SCHC ACK has been lost: this mechanism
   assumes that the Retransmission Timer for the All-1 SCHC Fragment is
   long enough to allow several SCHC ACK retries if the All-1 SCHC
   Fragment has not been received by the SCHC Fragment receiver, and it
   also assumes that it is unlikely that several ACKs become all lost).

Acknowledgements

   Thanks to (in alphabetical order) Sergio Aguilar Romero, David Black,
   Carsten Bormann, Deborah Brungard, Brian Carpenter, Philippe Clavier,
   Alissa Cooper, Roman Danyliw, Daniel Ducuara Beltran, Diego Dujovne,
   Eduardo Ingles Sanchez, Rahul Jadhav, Benjamin Kaduk, Arunprabhu
   Kandasamy, Suresh Krishnan, Mirja Kuehlewind, Barry Leiba, Sergio
   Lopez Bernal, Antoni Markovski, Alexey Melnikov, Georgios
   Papadopoulos, Alexander Pelov, Charles Perkins, Edgar Ramos, Alvaro
   Retana, Adam Roach, Shoichi Sakane, Joseph Salowey, Pascal Thubert,
   and Eric Vyncke for useful design considerations, reviews and
   comments.

   Carles Gomez has been funded in part by the Spanish Government
   (Ministerio de Educacion, Cultura y Deporte) through the Jose
   Castillejo grant CAS15/00336 and by the ERDF and the Spanish
   Government through project TEC2016-79988-P.  Part of his contribution
   to this work has been carried out during his stay as a visiting
   scholar at the Computer Laboratory of the University of Cambridge.

Authors' Addresses

   Ana Minaburo
   Acklio
   1137A avenue des Champs Blancs
   35510 Cesson-Sevigne Cedex
   France

   Email: ana@ackl.io

   Laurent Toutain
   IMT Atlantique
   2 rue de la Chataigneraie
   CS 17607
   35576 Cesson-Sevigne Cedex
   France

   Email: Laurent.Toutain@imt-atlantique.fr

   Carles Gomez
   Universitat Politecnica de Catalunya
   C/Esteve Terradas, 7
   08860 Castelldefels
   Spain

   Email: carlesgo@entel.upc.edu

   Dominique Barthel
   Orange Labs
   28 chemin du Vieux Chene
   38243 Meylan
   France

   Email: dominique.barthel@orange.com

   Juan Carlos Zuniga
   SIGFOX
   425 rue Jean Rostand
   31670 Labege
   France

   Email: JuanCarlos.Zuniga@sigfox.com