LPWAN Static Context Header Compression (SCHC) for CoAP

CoAP is an implementation of the REST architecture for constrained devices. Gateway between CoAP and HTTP can be easily built since both protocols uses the same address space (URL), caching mechanisms and methods. Nevertheless, if limited, the size of a CoAP header may be too large for LPWAN constraints and some compression may be needed to reduce the header size. defines a header compression mechanism for LPWAN network based on a static context. The context is said static since the element values composing the context are not learned during the packet exchanges but are previously defined. The context(s) is(are) known by both ends before transmission. A context is composed of a set of rules (contexts) that are referenced by Rule IDs (identifiers). A rule contains an ordered list of the header fields containing à field ID (FID) and its position when repeated, a direction indicator (DI) (upstream, downstream and bidirectional) and some associated Target Values (TV) which are expected in the message header. A Matching Operator (MO) is associated to each header field description. The rule is selected if all the MOs fit the TVs. In that case, a Compression Decompression Function (CDF) associated to each field defines the link between the compressed and decompressed value for each of the header fields.

CoAP differs from IPv6 and UDP protocols on the following aspects: IPv6 and UDP are symmetrical protocols. The same fields are found in the request and in the response, only location in the header may vary (e.g. source and destination fields). A CoAP request is different from an response. For example, the URI-path option is mandatory in the request and is not found in the response, request may contain an Accept option and the response a Content-format option. Even when a field is “symmetric” (i.e. found in both directions) the values carried are different. For instance the Type field will contain a CON value in the request and a ACK or RST value in the response. Exploiting the asymmetry in compression will allow to send no bit in the compressed request and a single bit in the answer. Same behavior can be applied to the CoAP Code field (O.OX code are present in the request and Y.ZZ in the answer). CoAP also obeys to the client/server paradigm and the compression rate can be different if the request is issued from a LPWAN node or from an non LPWAN device. For instance a Thing (ES) aware of LPWAN constraints can generate a 1 byte token, but a regular CoAP client will certainly send a larger token to the Thing. SCHC compression will not modify the values to offer a better compression rate. Nevertheless a proxy placed before the compressor may change some field values to offer a better compression rate and maintain the necessary context for interoperability with existing CoAP implementations. In IPv6 and UDP header fields have a fixed size. In CoAP, Token size may vary from 0 to 8 bytes, length is given by a field in the header. More systematically, the CoAP options are described using the Type-Length-Value. When applying SCHC header compression. By sending compressed field information following the rule order, SCHC offers a serialization/deserialization mechanism. Since a field exists to indicate the token length there is no ambiguity. For options, the rule indicates also the expected options found the int CoAP header. Therefore only the length is needed to recognise an option. The length will be send using the same CoAP encoding (size less than 12 are directly sent, higher values uses the escape mechanisms defined by ). Delta Type is omitted, the value will be recovered by the decompressor. This reduce the option length of 4, 12 or 20 bits regarding the orignial size of the delta type encoding in the option. In CoAP headers a field can be duplicated several times, for instances, elements of an URI (path or queries) or accepted formats. The position defined in a rule, associated to a Field ID, can be used to identify the proper element.

This section discusses of the compression of the different CoAP header fields. These are just examples. The compression should take into account the nature of the traffic and not only the field values. Next chapter will define some compression rules for some common exchanges.

This field is bidirectional and can be elided during the SCHC compression, since it always contains the same value. It appears only in first position.

This field can be managed bidirectionally or unidirectionally.Several strategies can be applied to this field regarding the values used: If the ES is a client or a Server and non confirmable message are used, the transmission of the Type field can be avoided: Pos is always 1, DI can either be “uplink” if the ES is a CoAP client or “downlink” if the ES is a CoAP server, or “bidirectional” TV is set to the value, MO is set to “equal” CDF is set to “not-sent”.

If the ES is either a client or a Server and confirmable message are used, the DI can be used to elide the type on the request and compress it to 1 bit on the response. The example above shows the rule for a ES acting as a client, directions need to be reversed for a ES acting as a server.

Otherwise if the ES is acting simultaneously as a client and a server and the rule handle these two traffics, Type field must be sent uncompressed.

This field is bi-directional. Several strategies can be applied to this field regarding the values: no token or a wellknown length, the transmission can be avoided. A special care must be taken, if CON messages are acknowledged with an empty ACK message. In that case the token is not always present.

If the length is changing from one message to an other, the Token Length field must be sent. If the Token length can be limited, then only the least significant bits have to be sent. The example below allows values between 0 and 3.

otherwise the field value has to be sent.

This field is bidirectional, but compression can be enhanced using DI. The CoAP Code field defines a tricky way to ensure compatibility with HTTP values. Nevertheless only 21 values are defined by compared to the 255 possible values.

gives a possible mapping, it can be changed to add new codes or reduced if some values are never used by both ends. It could efficiently be coded on 5 bits. Even if the number of code can be increase with other RFC, implementations may use a limited number of values, which can help to reduce the number of bits sent on the LPWAN. The number of code may vary over time, some new codes may be introduced or some applications use a limited number of values. The client and the server do not use the same values. This asymmetry can be exploited to reduce the size sent on the LPWAN. The field can be treated differently in upstream than in downstream. If the Thing is a client an entry can be set on the uplink message with a code matching for 0.0X values and another for downlink values for Y.ZZ codes. It is the opposite if the thing is a server. If the ES always sends or receives requests with the same method, the Code field can be elided. The entry below shows a rule for a client sending only GET request.

If the client may send different methods, a matching-list can be applied. For table , 3 bits are necessary, but it could be less if fewer methods are used. Example below gives an example where the ES is a server and receives only GET and POST requests.

The same approach can be applied to responses.

This field is bidirectional. Message ID is used for two purposes: To acknowledge a CON message with an ACK. To avoid duplicate messages. In LPWAN, since a message can be received by several radio gateway, some LPWAN technologies include a sequence number in L2 to avoid duplicate frames. Therefore if the message does not need to be acknowledged (NON or RST message), the Message ID field can be avoided.

The decompressor must generate a value. [[Note; check id this field is not used by OSCOAP .]] To optimize information sent on the LPWAN, shorter values may be used during the exchange, but Message ID values generated a common CoAP implementation will not take into account this limitation. Before the compression, a proxy may be needed to reduce the size.

Otherwise if no compression is possible, the field has to be sent

This field is bi-directional. Token is used to identify transactions and varies from one transaction to another. Therefore, it is usually necessary to send the value of the token field on the LPWAN network. The optimization will occur by using small values. Common CoAP implementations may generate large tokens, even if shorter tokens could be used regarding the LPWAN characteristics. A proxy may be needed to reduce the size of the token before compression. The size of the compress token sent is known by a combination of the Token Length field and the rule entry. For instance, with the entry below:

The uncompressed token is 2 bytes long, but the compressed size will be 4 bits.

This field is unidirectional and must not be set to bidirectional in a rule entry. It is used only by the server to inform the client about of the payload type and is never found in client requests. If single value is expected by the client, the TV contains that value and MO is set to “equal” and the CDF is set to “not-sent”. The examples below describe the rules for an ES acting as a server.

If several possible value are expected by the client, a matching-list can be used.

Otherwise the value can be sent.The value-sent CDF in the compressor do not send the option type and the decompressor reconstruct it regarding the position in the rule.

This field is unidirectional and must not be set to bidirectional in a rule entry. It is used only by the client to inform of the possible payload type and is never found in server response. The number of accept options is not limited and can vary regarding the usage. To be selected a rule must contain the exact number about accept options with their positions. Since the order in which the Accept value are sent, the position order can be modified. The rule below

will be selected only if two accept options are in the CoAP header if this order. The rule below:

will accept a-only CoAP messages with 2 accept options, but the order will not influence the rule selection. The decompression will reconstruct the header regarding the rule order. Otherwise a matching-list can be applied to the different values, in that case the order is important to recover the appropriate value and the position must be clearly indicate.

Finally, the option can be explicitly sent.

This field is unidirectional and must not be set to bidirectional in a rule entry. It is used only by the server to inform of the caching duration and is never found in client requests. If the duration is known by both ends, value can be elided on the LPWAN. A matching list can be used if some wellknown values are defined. Otherwise the option length and value can be sent on the LPWAN. [[note: we can reduce (or create a new option) the unit to minute, second is small for LPWAN ]]

This fields are unidirectional and must not be set to bidirectional in a rule entry. They are used only by the client to access to a specific resource and are never found in server response. The Matching Operator behavior has not changed, but the value must take a position value, if the entry is repeated :

For instance, the rule matches with /foo/bar, but not /bar/foo. When the length is not clearly indicated in the rule, the value length must be sent with the field data, which means for CoAP to send directly the CoAP option with length and value. For instance for a CoMi path /c/X6?k=”eth0” the rule can be set to:

shows the parsing and the compression of the URI. where c is not sent. The second element is sent with the length (i.e. 0x2 X 6) followed by the query option (i.e. 0x05 “eth0”). A Mapping list can be used to reduce size of variable Paths or Queries. In that case, to optimize the compression, several elements can be regrouped into a single entry. Numbering of elements do not change, MO comparison is set with the first element of the matching.

For instance, the following Path /foo/bar/variable/stable can leads to the rule defined .

These fields are unidirectional and must not be set to bidirectional in a rule entry. They are used only by the client to access to a specific resource and are never found in server response. If the field value must be sent, TV is not set, MO is set to “ignore” and CDF is set to “value-sent. A mapping can also be used. Otherwise the TV is set to the value, MO is set to “equal” and CDF is set to “not-sent”

These fields are unidirectional. These fields values cannot be stored in a rule entry. They must always be sent with the request. [[Can include OSCOAP Object security in that category ]]

Block option should be avoided in LPWAN. The minimum size of 16 bytes can be incompatible with some LPWAN technologies. [[Note: do we recommand LPWAN fragmentation since the smallest value of 16 is too big?]]

defines the Observe option. The TV is not set, MO is set to “ignore” and the CDF is set to “value-sent”. SCHC does not limit the maximum size for this option (3 bytes). To reduce the transmission size either the Thing implementation should limit the value increase or a proxy can be used limit the increase. Since RST message may be sent to inform a server that the client do not require Observe response, a rule must allow the transmission of this message.

defines an No-Response option limiting the responses made by a server to a request. If the value is not by both ends, then TV is set to this value, MO is set to “equal” and CDF is set to “not-sent”. Otherwise, if the value is changing over time, TV is not set, MO is set to “ignore” and CDF to “value-sent”. A matching list can also be used to reduce the size.

In this first scenario, the LPWAN compressor receives from outside client a POST message, which is immediately acknowledged by the Thing. For this simple scenario, the rules are described .

The version and Token Length fields are elided. Code has shrunk to 5 bits using the matching list (as the one given : 0.01 is value 0x01 and 2.05 is value 0x0c) Message-ID has shrunk to 9 bits to preserve alignment on byte boundary. The most significant bit must be set to 0 through a CoAP proxy. Uri-Path contains a single element indicated in the matching operator. shows the time diagram of the exchange. A LPWAN Application Server sends a CON message. Compression reduces the header sending only the Type, a mapped code and the least 9 significant bits of Message ID. The receiver decompresses the header. . The CON message is a request, therefore the LC process to a dynamic mapping. When the ES receives the ACK message, this will not initiate locally a message ID mapping since it is a response. The LC receives the ACK and uncompressed it to restore the original value. Dynamic Mapping context lifetime follows the same rules as message ID duration.

| rule id=1 | +-+-+--+----+--------+ |-------------------->| |1|2| 0|2.05| 0x0034 | | TTCC CCCM MMMM MMMM|------------------------> +-+-+--+----+--------+ | 1001 1000 0011 0100| +-+-+--+----+--------+ | | |1|2| 0|2.05| 0x0034 | v v +-+-+--+----+--------+ ]]> The message can be further optimized by setting some fields unidirectional, as described in . Note that Type is no more sent in the compressed format, Compressed Code size in not changed in that example (8 values are needed to code all the requests and 21 to code all the responses in the matching list )

In that example, the Thing is using CoMi and sends queries for 2 SID.

| | | | | |<------------------------| ACK MID=0x0012 | | 0.00 | | | | |<------------------------| CON | | MID=0X0034 | | Content-Format X ACK MID=0x0034 |------------------------>| 0.00 ]]>