lpwan Working Group | A. Minaburo |
Internet-Draft | Acklio |
Intended status: Informational | L. Toutain |
Expires: September 11, 2017 | Institut MINES TELECOM ; IMT Atlantique |
March 10, 2017 |
LPWAN Static Context Header Compression (SCHC) for CoAP
draft-ietf-lpwan-coap-static-context-hc-01
This draft discusses the way SCHC header compression can be applied to CoAP headers in an LPWAN flow regarding the generated traffic. CoAP protocol differs from IPv6 and UDP protocols because the CoAP Header has a flexible header due to variable options. Another important difference is the asymmetric format in the header information used in the request and the response packets. This draft shows that the Client and the Server do not uses the same fields and how the SCHC header compression can be used.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on September 11, 2017.
Copyright (c) 2017 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 (http://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.
[I-D.toutain-lpwan-ipv6-static-context-hc] defines a header compression mechanism for LPWAN network based on a static context. Where 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 describes the header fields with some associated Target Values (TV). 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.
This draft discusses the way SCHC can be applied to CoAP headers, how to extend MOs to match a specific element when several fields of the same type are presented in the header. It also introduces the notion of bidirectional or unidirectional (upstream and downstream) fields.
CoAP [RFC7252] 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. CoAP compression is not straightforward. Some differences between IPv6/UDP and CoAP can be highlighted. CoAP differs from IPv6 and UDP protocols in the following
aspects:
A LPWAN node can either be a client or a server and sometimes both. In the client mode, the LPWAN node sends request to a server and expects an answer or acknowledgements. Acknowledgements can be at 2 different levels:
Note that acknowledgement can be optimized and a REST level acknowledgement can be used as a transport level acknowledgement.
CoAP header format defines the following fields:
For a given flow some value options are stable through time. Observe, ETag, If-Match, If-None-Match and Size varies in each message.
The CoAP protocol must not be altered by the compression/decompression phase, but if no semantic is attributed to a value, it may be changed during this phase. For instance, the compression phase may reduce the size of a token but must maintain its unicity. The decompressor will not be able to restore the original value but the behavior will remain the same. If no special semantic is assigned to the token, this will be transparent. If a special semantic is assigned to the token, its compression may not be possible.
This draft refines the rules definition by adding the direction of the message, from the Thing point of view (uplink, downlink or bidirectional). It does not introduce new Machting Operator or new Compression Decompression Function, but add some possibility to check one particular element when several of them are present at the same time.
A rule can contain CoAP and IPv6/UDP entries. In that case, IPv6/UDP entries are tagged bidirectional.
By default, an entry in a rule is bidirectional which means that it can be applied either on the uplink or downlink headers. By specifying the direction, the LC will take into account the specific field only if the direction match.
If the Thing is a client, the URI-Path option is only present on request and not on the response. Therefore, the exact matching principle to select a rule cannot apply.
Some options are marked unidirectional, the value (uplink or downlink) depends of the scenario. A Uri-Path option will be marked uplink if the Thing acts as a client and downlink if the Thing acts as a server. If the Thing acts both as client and server, two different rules will be defined.
The Matching Operator behavior has not changed, but the value must take a position value, if the entry is repeated :
FID TV MO CDF URI-Path foo equal 1 not-sent URI-Path bar equal 2 not-sent
Figure 1: Position entry.
For instance, the rule Figure 1 matches with /foo/bar, but not /bar/foo.
The position is added after the natural argument of the MO, for example MSB (4,3) indicates a most significant bit matching of 4 bits in a field located in position 3.
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 where the delta-T is set to 0.
For the CoMi path /c/X6?k=”eth0” the rule can be set to:
FID TV MO CDF URI-Path c equal 1 not-sent URI-Path ignore 2 value-sent URI-Query k= MSB (16, 1) value-sent
Figure 2: CoMi URI compression
Figure 2 shows the parsing and the compression of the URI. where c is not sent. The second element is sent with the length (i.e. 0x02 X 6) followed by the query option (i.e. 0x08 k=”eth0”).
[[NOTE we don’t process URI with a multiple number of path element ??]].
This section lists the different CoAP header fields and how they can be compressed.
This field is bidirectional.
This field contains always the same value, therefore the TV may be 1, the MO is set to “equal” and the CDF is set to “not-sent”
This field is bidirectional or undirectional.
Several strategies can be applied to this field regarding the values used:
This field is bi-directional.
Several strategies can be applied to this field regarding the values:
This field is unidirectional. The client and the server do not use the same values.
The CoAP code field defines a tricky way to ensure compatibility with HTTP values. Nevertheless only 21 values are defined by [RFC7252] compared to the 255 possible values. So, it could efficiently be coded on 5 bits. The number of code may vary over time, some new codes may be introduced or some applications use a limited number of values.
+------+------------------------------+-----------+ | Code | Description | Mapping | +------+------------------------------+-----------+ | 0.00 | | 0x00 | | 0.01 | GET | 0x01 | | 0.02 | POST | 0x02 | | 0.03 | PUT | 0x03 | | 0.04 | DELETE | 0x04 | | 0.05 | FETCH | 0x05 | | 0.06 | PATCH | 0x06 | | 0.07 | iPATCH | 0x07 | | 2.01 | Created | 0x08 | | 2.02 | Deleted | 0x09 | | 2.03 | Valid | 0x0A | | 2.04 | Changed | 0x0B | | 2.05 | Content | 0x0C | | 4.00 | Bad Request | 0x0D | | 4.01 | Unauthorized | 0x0E | | 4.02 | Bad Option | 0x0F | | 4.03 | Forbidden | 0x10 | | 4.04 | Not Found | 0x11 | | 4.05 | Method Not Allowed | 0x12 | | 4.06 | Not Acceptable | 0x13 | | 4.12 | Precondition Failed | 0x14 | | 4.13 | Request Entity Too Large | 0x15 | | 4.15 | Unsupported Content-Format | 0x16 | | 5.00 | Internal Server Error | 0x17 | | 5.01 | Not Implemented | 0x18 | | 5.02 | Bad Gateway | 0x19 | | 5.03 | Service Unavailable | 0x1A | | 5.04 | Gateway Timeout | 0x1B | | 5.05 | Proxying Not Supported | 0x1C | +------+------------------------------+-----------+
Figure 3: Example of CoAP code mapping
Figure 3 gives a possible mapping, it can be changed to add new codes or reduced if some values are never used by both ends.
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.
This field is bidirectional.
Message ID is used for two purposes:
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. In that case TV is not set, MO is set to ignore and CDF is set to “not-sent”. The decompressor can generate a number.
[[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. In that case, the TV is set to 0x0000, MO is set to “MSB(l)” and CDF is set to “LSB(16-l)”, where “l” is the size of the compressed header.
Otherwise if no compression is needed the TV is not set, MO is set to ignore and CDF is set to “not-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.
Otherwise the TV is not set, the MO is set to ignore and CDF is set to “value-sent”.
The decompression may know the length of the token field from the token length field.
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 the value is known by both sides, the TV contains that value and MO is set to “equal” and the CDF is set to “not-sent”.
Otherwise the TV is not set, MO is set to “ignore” and CDF is set to “value-sent”
A mapping list can also be used to reduce the size.
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.
if the accept value must be sent, the TV contains that value, MO is set to “ignore x” where “x” is the accept option’s position and CDF is set to value-sent. Since the value length is not known, it must be sent as a CoAP TLV with delta-T set to 0.
Otherwise the TV is not set, MO is set to “equal x” where x is the accept option’s position and CDF is set to “not-sent”
[[note: it could be more liberal and do not provide the same order after decompression]]
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, the TV is set with this duration, the MO is set to “equal” and the CDF is set to “not-sent”.
Otherwise the TV is not set, the MO is set to “ignore” and the CDF is set to “value-sent”. Since the value length is not known, it must be sent as a CoAP TLV with delta-T set to 0.
[[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 server and are never found in server response.
For each option, if the value is known by both ends, the TV is set with this value, the MO is set to “equal” and the CDF is set to “not-sent”.
Otherwise the TV is not set, the MO is set to “ignore” and the CDF is set to “value-sent”. Since the value length is not known, it must be sent as a CoAP TLV with delta-T set to 0.
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.
Path and Query option may have different formats. Section 3.2 gives some examples.
If the URI path as well as the query is composed of 2 or more elements, then the position must be set in the MO parameters.
If a Path or Query element is stable over the time, then TV is set with its value, MO is set to “equal x” where x is the position in the Path or the Query and CDF is set to “not-sent”.
Otherwise if the value varies over time, TV is not set, MO is set to “ignore x” where x is the position in the Path or in the Query and CDF is set to “value-sent”. Since the value length is not known, it must be sent as a CoAP TLV with deltaT set to 0.
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 Figure 4.
FID TV MO CDF URI-Path {"/foo/bar":1, match-mapping 1 mapping-sent "/bar/foo":2} URI-Path ignore 3 value-sent URI-Path stable equal 4 not-sent
Figure 4: complex path example
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?]]
[RFC7641] 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.
[RFC7967] 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 Figure 5.
rule id 1 +-------------+------+---------+-------------+-----+----------------+ | Field |TV |MO |CDF |dir | Sent | +=============+======+=========+=============+=====+================+ |CoAP version | 01 |equal |not-sent |bi | | |CoAP Type | |ignore |value-sent |bi |TT | |CoAP TKL | 0 |equal |not-sent |bi | | |CoAP Code | ML1 |match-map|matching-sent|bi | CC CCC | |CoAP MID | 0000 |MSB(7 ) |LSB(9) |bi | M-ID | |CoAP Uri-Path| path |equal 1 |not-sent |down | | +-------------+------+---------+-------------+-----+----------------+
Figure 5: CoAP Context to compress header without token
The version and Token Length fields are elided. Code has shrunk to 5 bits using the matching list (as the one given Figure 3: 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.
Figure 6 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.
End System LPWA LC | | | rule id=1 |<---------------------- |<--------------------| +-+-+--+----+--------+ <-------------------- | TTCC CCCM MMMM MMMM| |1|0| 4|0.01| 0x0034 | +-+-+--+----+--------+ | 0000 0010 0011 0100| | 0xb4 p a t | |1|0| 1|0.01| 0x0034 | | | | h | | 0xb4 p a t | | | +------+ | h | | | +------+ | | | | | | ----------------------->| 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 +-+-+--+----+--------+
Figure 6: Compression with global addresses
The message can be further optimized by setting some fields unidirectional, as described in Figure 7. 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 Figure 3)
rule id 1 +-------------+------+---------+-------------+---+----------------+ | Field |TV |MO |CDF |dir| Sent | +=============+======+=========+=============+===+================+ |CoAP version | 01 |equal |not-sent |bi | | |CoAP Type | CON |equal |not-sent |dw | | |CoAP Type | ACK |equal |not-sent |up | | |CoAP TKL | 0 |equal |not-sent |bi | | |CoAP Code | ML2 |match-map|matching-sent|dw |CCCC C | |CoAP Code | ML3 |match-map|matching-sent|up |CCCC C | |CoAP MID | 0000 |MSB(5) |LSB(11) |bi | M-ID | |CoAP Uri-Path| path |equal 1 |not-sent |dw | | +-------------+------+---------+-------------+---+----------------+ ML1 = {CON : 0, ACK:1} ML2 = {POST:0, 2.04:1, 0.00:3}
Figure 7: CoAP Context to compress header without token
In that example, the Thing is using CoMi and sends queries for 2 SID.
CON MID=0x0012 | | POST | | Accept X | | /c/k=AS |------------------------>| | | | | |<------------------------| ACK MID=0x0012 | | 0.00 | | | | |<------------------------| CON | | MID=0X0034 | | Content-Format X ACK MID=0x0034 |------------------------>| 0.00
rule id 3 +-------------+------+---------+-------------+---+----------------+ | Field |TV |MO |CDF |dir| Sent | +=============+======+=========+=============+===+================+ |CoAP version | 01 |equal |not-sent |bi | | |CoAP Type | CON |equal |not-sent |up | | |CoAP Type | ACK |equal |not-sent |dw | | |CoAP TKL | 1 |equal |not-sent |bi | | |CoAP Code | POST |equal |not-sent |up | | |CoAP Code | 0.00 |equal |not-sent |dw | | |CoAP MID | 0000 |MSB(8) |LSB(8) |bi |MMMMMMMM | |CoAP Token | |ignore |send-value |up |TTTTTTTT | |CoAP Uri-Path| /c |equal 1 |not-sent |dw | | |CoAP Uri-query ML4 |equal 1 |not-sent |dw |P | |CoAP Content | X |equal |not-sent |up | | +-------------+------+---------+-------------+---+----------------+ rule id 4 +-------------+------+---------+-------------+---+----------------+ | Field |TV |MO |CDF |dir| Sent | +=============+======+=========+=============+===+================+ |CoAP version | 01 |equal |not-sent |bi | | |CoAP Type | CON |equal |not-sent |dw | | |CoAP Type | ACK |equal |not-sent |up | | |CoAP TKL | 1 |equal |not-sent |bi | | |CoAP Code | 2.05 |equal |not-sent |dw | | |CoAP Code | 0.00 |equal |not-sent |up | | |CoAP MID | 0000 |MSB(8) |LSB(8) |bi |MMMMMMMM | |CoAP Token | |ignore |send-value |dw |TTTTTTTT | |COAP Accept | X |equal |not-sent |dw | | +-------------+------+---------+-------------+---+----------------+ alternative rule: rule id 4 +-------------+------+---------+-------------+---+----------------+ | Field |TV |MO |CDF |dir| Sent | +=============+======+=========+=============+===+================+ |CoAP version | 01 |equal |not-sent |bi | | |CoAP Type | ML1 |equal |match-sent(1)|bi |t | |CoAP TKL | 1 |equal |not-sent |bi | | |CoAP Code | ML2 |equal |match-sent(1)|up | cc | |CoAP Code | ML3 |equal |match-sent(2)|dw | cc | |CoAP MID | 0000 |MSB(8) |LSB(8) |bi |MMMMMMMM | |CoAP Token | |ignore |send-value |dw |TTTTTTTT | |CoAP Uri-Path| /c |equal 1 |not-sent |dw | | |CoAP Uri-query ML4 |equal 1 |not-sent |dw |P | |CoAP Content | X |equal |not-sent |up | | |COAP Accept | x |equal |not-sent |dw | | +-------------+------+---------+-------------+---+----------------+ ML1 {CON:0, ACK:1} ML2 {POST:0, 0.00: 1} ML3 {2.05:0, 0.00:1} ML4 {NULL:0, k=AS:1, K=AZE:2}