Internet-Draft | Corrections and Clarifications to CoAP | September 2024 |
Bormann | Expires 28 March 2025 | [Page] |
RFC 7252 defines the Constrained Application Protocol (CoAP), along with a number of additional specifications, including RFC 7641, RFC 7959, RFC 8132, and RFC 8323. RFC 6690 defines the link format that is used in CoAP self-description documents.¶
Some parts of the specification may be unclear or even contain errors that may lead to misinterpretations that may impair interoperability between different implementations. The present document provides corrections, additions, and clarifications to the RFCs cited; this document thus updates these RFCs. In addition, other clarifications related to the use of CoAP in other specifications, including RFC 7390 and RFC 8075, are also provided.¶
This note is to be removed before publishing as an RFC.¶
Status information for this document may be found at https://datatracker.ietf.org/doc/draft-bormann-core-corr-clar/.¶
Discussion of this document takes place on the core Working Group mailing list (mailto:core@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/core/. Subscribe at https://www.ietf.org/mailman/listinfo/core/.¶
Source for this draft and an issue tracker can be found at https://github.com/core-wg/corrclar.¶
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 https://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 28 March 2025.¶
Copyright (c) 2024 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 Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
[RFC7252] defines the Constrained Application Protocol (CoAP), along with a number of additional specifications, including [RFC7641], [RFC7959], [RFC8132], and [RFC8323]. [RFC6690] defines the link format that is used in CoAP self-description documents.¶
During implementation and interoperability testing of these RFCs, and in their practical use, some ambiguities and common misinterpretations have been identified, as well as a few errors.¶
The present document summarizes identified issues and provides corrections needed for implementations of CoAP to interoperate, i.e., it constitutes an update to the RFCs referenced. This document also provides other clarifications related to common misinterpretations of the specification. References to CoAP should, therefore, also include this document.¶
In addition, some clarifications and corrections are also provided for documents that are related to CoAP, including RFC 7390 and RFC 8075.¶
The present document is an Internet-Draft, which is not intended to be published as an RFC quickly. Instead, it will be maintained as a running document of the CoRE WG, probably for a number of years, until the need for new entries tails off and the document can finally be published as an RFC. (This paragraph to be rephrased when that happens.)¶
The status of this document as a running document of the WG implies a consensus process that is applied in making updates to it. The rest of this subsection provides more details about this consensus process. (This is the intended status; currently, the document is an individual submission only.)¶
(Consensus process TBD, but it will likely be based on an editor's version in a publicly accessible git repository, as well as periodic calls for consensus that lead to a new published Internet-Draft.)¶
This section describes the process that will be used for developing the present document (called "-corr-clar" colloquially).¶
This process might be revised as its execution progresses.¶
(Done as of this a draft): include the present process proposal.
The document can then already be considered for WG adoption.¶
Go through the following available material and revise/create Github issues at ISSUES as needed:¶
Categorize the Github issues at ISSUES as to the topics they relate to, by tagging them.
Completing a first round of this will be a task for a dedicated team.¶
For each issue or set of issues at ISSUES, confirm with the CoRE WG and gather feedback from affected protocol designers/implementors if the issue is best to be:¶
Reshape the -corr-clar document in order to reflect a sequence of pairs (Diagnosis, Therapy), where:¶
Diagnosis is the gist of a set of Github issues to cover, and¶
Therapy is the correction or clarification to address those.¶
Even though at a high-level, the scope should be already clear by looking at the table of contents. That is, at this stage, there is no need to necessarily elaborate the Therapy in detail, but it is necessary to make a reader understand "what we are dealing with and taking which direction".¶
WG document work can then focus on improving the therapy parts, until all points are satisfactorily addressed and documented.¶
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.¶
When a section of this document makes formal corrections, additions or invalidations to text in a referenced RFC, this is clearly summarized. The text from the RFC that is being addressed is given and labeled "INCOMPLETE", "INCORRECT", or "INCORRECT AND INVALIDATED", followed by the correct text labeled "CORRECTED", where applicable. When text is added that does not simply correct text in previous specifications, it is given with the label "FORMAL ADDITION".¶
Where a resolution has not yet been agreed, the resolution is marked PENDING.¶
In this document, a reference to a section in RFC nnnn is written as RFC nnnn-<number>, where <number> is the section number.¶
[RFC7252] uses the term "request URI" repeatedly, but only provides a vague definition in Section 5.7.2 of [RFC7252]. Text should be added to the definitions in Section 1.2 (Terminology) of [RFC7252].¶
The URI that identifies a resource on a server intended to be addressed by a request; constructed from the context of the request combined with Options present in the request using the process defined in Section 6.5 (Composing URIs from Options) of [RFC7252], see Section 5.7.2 (Forward-Proxies) of [RFC7252] for further details. Related to the HTTP concept of "target URI"; see Section 7.1 (Determining the Target Resource) of [RFC9110]; previously called "effective request URI" in Section 5.5 (Effective Request URI) of [RFC7230].¶
PENDING.¶
Note that a similar, but distinct concept is the "base URI", relative to which relative URIs are resolved. This is more complex in CoAP than in HTTP because CoAP can multicast requests (Section 8 of [RFC7252]), expecting unicast responses; these need to be interpreted relative to the unicast source address from which the responses come.¶
Section 8.2 of [RFC7252] has:¶
For the purposes of interpreting the Location-* options and any links embedded in the representation, the request URI (i.e., the base URI relative to which the response is interpreted) is formed by replacing the multicast address in the Host component of the original request URI by the literal IP address of the endpoint actually responding.¶
Similarly, Section 8.2.1 of [RFC7252] (Caching) says:¶
A response received in reply to a GET request to a multicast group MAY be used to satisfy a subsequent request on the related unicast request URI. The unicast request URI is obtained by replacing the authority part of the request URI with the transport-layer source address of the response message.¶
Further discussion of a more generalized response concept can be found in [I-D.bormann-core-responses].¶
In the discussion of [RFC8516], a comment was made that it would be needed to define the point in time relative to which Max-Age is defined. A sender might reference it to the time it actually sends the message containing the option (and paragraph 3 of RFC7252-5.10.5 indeed requests that Max-Age be updated each time a message is retransmitted). The receiver of the message does not have reliable information about the time of sending, though. It may instead reference the Max-Age to the time of reception. This in effect extends the time of Max-Age by the latency of the packet. This extension was deemed acceptable for the purposes of [RFC8516], but may be suboptimal when Max-Age is about the lifetime of a response object.¶
The value is intended to be current at the time of transmission.¶
PENDING.¶
[Err4895] notes (text updated with newer link):¶
The current specification for decomposing a URI into CoAP Options (Section 6.4) is correct; however the text may still be unclear to implementers who may think that the phrase "not including the delimiting slash characters" means simply omitting a trailing slash character in the URI path. This is incorrect. See the discussion outcome in email thread https://mailarchive.ietf.org/arch/msg/core/vqOiUreodGXqWZGeGOTREChCsKY/.¶
[Err4895] proceeds to propose adding another note at the end of Section 6.4 of [RFC7252]. A slightly updated version of the proposed note might be:¶
Also note that a trailing slash character in the <path> component represents a separate, zero-character path segment (see the ABNF in Section 3.3 of [RFC3986]). This is encoded using a Uri-Path Option of zero length. The exception at the start of step 8 means that no such zero-character path segment is added for a trailing slash that immediately follows the authority (host and port).¶
As has been noted in [Err5078], there is no information in [RFC7252] about whether the "ct" target attribute can be present multiply or not.¶
The text does indicate that the value of the attribute MAY be "a space-separated sequence of Content-Format codes, indicating that multiple content-formats are available", but it does not repeat the prohibition of multiple instances that the analogously structured "rt" and "if" attributes in Sections 3.1 and 3.2 of [RFC6690] have.¶
This appears to be an oversight. Published examples in Section 4.1 of [RFC9148] and Section 4.3 of [RFC9176] further illustrate that the space-separated approach is generally accepted to be the one to be used. There is no gain to be had from allowing both variants, and it would be likely to cause interoperability problems.¶
At the 2022-11-23 CoRE WG interim meeting, there was agreement that [Err5078] should be marked "VERIFIED", which was done on 2023-01-18.¶
The Content-Format code attribute MUST NOT appear more than once in a link.¶
Sections 9.1.1 and 9.1.2 of [RFC7252] provide details about using CoAP over DTLS connections; in particular they restrict both message-id matching and request/response matching to within a single combination of DTLS session/DTLS epoch.¶
At the time, this was a decision to be very conservative based on the wide variety of security implications a new DTLS epoch might have (which also were not widely understood, e.g., for a re-authentication). The normally short time between a request and a response made this rather strict boxing appear acceptable.¶
However, with the Observe functionality [RFC7641], it is quite likely that significant time elapses between a request and the arrival of a notification that is sent back as a response, causing a need for the latter to use a different box from the original request.¶
Also, additions to CoAP such as CoAP over connection-oriented transports [RFC8323] and OSCORE [RFC8613] create similar matching boxes that also do not fit certain likely use cases of these additions (e.g., with short-lived TCP connections as discussed in Section 4.3 of [RFC9006]).¶
The match boxing semantics of the current protocol are clearly defined, but can be unsatisfactory given the current requirements.¶
This calls for careful design choices and enhancements when developing extensions for CoAP or protocols and methods applicable to CoAP, such as in the cases overviewed in the following Section 2.5.2 and Section 2.5.3.¶
PENDING:¶
The security protocol Object Security for Constrained RESTful Environments (OSCORE) defined in [RFC8613] provides end-to-end security for CoAP messages at the application level, by using CBOR Object Signing and Encryption (COSE) [RFC9052]. In order to protect their communications, two peers need to have already established an OSCORE Security Context.¶
Appendix B.2 of [RFC8613] provides an example for a key update procedure, which two OSCORE peers can run for establishing a new shared OSCORE Security Context that replaces their old one. The recent key update protocol KUDOS [I-D.ietf-core-oscore-key-update] specifies how two OSCORE peers can establish a new shared OSCORE Security Context that replaces their old one, with a number of advantages compared to the method defined in Appendix B.2 of [RFC8613].¶
The security protocol Group Object Security for Constrained RESTful Environments (Group OSCORE) defined in [I-D.ietf-core-oscore-groupcomm] builds on OSCORE and protects group communication for CoAP [I-D.ietf-core-groupcomm-bis]. The management of the group keying material is entrusted to a Group Manager (see Section 3 of [I-D.ietf-core-oscore-groupcomm]), which provides the latest group keying material to a new group member upon its group joining, and can update the group keying material by performing a group rekeying.¶
The following discusses how OSCORE, KUDOS, and Group OSCORE position themselves with respect to the match boxing, the transport used underlying CoAP, and the renewal of the keying material.¶
The security processing of (Group) OSCORE is agnostic of the value assumed by the CoAP Token and Message ID. Also, (Group) OSCORE can be seamlessly used in the presence of (cross-)proxies that will change the value of the CoAP Token and Message ID on the different communication legs. This does not affect the security processing at the (Group) OSCORE endpoints.¶
Before any security processing is performed, the only use that (Group) OSCORE makes of the Token value is on the CoAP client upon receiving a response, in order to retrieve the right Security Context to use for decrypting and verifying the response.¶
Even in case the Token value in a CoAP response is manipulated to induce a Request-Response matching at the client, there is no risk for the client to successfully decrypt the response instead of failing as expected. This is because, per Section 12.3 of [RFC8613], the OSCORE Master Secret of each OSCORE Security Context is required not only to be secret, but also to have a good amount of randomness.¶
Building on that, an HKDF is used to derive the actual encryption keys from the Master Secret and, optionally, from an additional Master Salt. Furthermore, for each OSCORE Security Context, the quartet (Master Secret, Master Salt, ID Context, Sender ID) must be unique. As per Section 3.3 of [RFC8613], this is a hard requirement and guarantees unique (key, nonce) pairs for the AEAD algorithm used.¶
In Group OSCORE, the Security Context extends that of OSCORE, and the same as above holds (see Sections 2, 2.2, and 13.11 of [I-D.ietf-core-oscore-groupcomm]).¶
Finally, (Group) OSCORE performs a separate secure match boxing under its own control, by cryptographically binding each protected request with all the corresponding protected responses. This is achieved by means of the COSE external_aad involved during the security processing of protected messages (see Section 5.4 of [RFC8613] and Section 4.4 of [I-D.ietf-core-oscore-groupcomm]).¶
The security protocol (Group) OSCORE does not have any requirement on binding the Security Context in use to specific addressing information used by the transport protocol underlying CoAP. What occurs below (Group) OSCORE with transport-specific addressing information is transparent to (Group) OSCORE, but it needs to work well enough to ensure that received data is delivered to (Group) OSCORE for security processing.¶
Consistent with the above, (Group) OSCORE does not interfere with any low-layer, transport specific information. Instead, it entrusts data to a Request-Response exchange mechanism that can rely on different means to enforce the Request-Response matching at the transport level (e.g., the 5-tuple, the CoAP Message ID, a file handle). Also, (Group) OSCORE does not alter the fact that a CoAP response needs to come from where the corresponding CoAP request was sent, which simply follows from using transports where that is a requirement.¶
Furthermore, two peers can seamlessly use (Group) OSCORE also in the presence of cross-proxies that change transport across different communication legs. This does not affect the security processing at the (Group) OSCORE endpoints.¶
Practically, (Group) OSCORE relies on the underlying CoAP implementation for obtaining received CoAP messages on which to perform the expected security processing.¶
Upon receiving a protected message, the recipient endpoint retrieves the OSCORE Security Context to use for decryption based on key identifier information specified in the CoAP OSCORE Option of protected requests, and on the value of the CoAP Token of protected responses.¶
In OSCORE, the key identifier information in request messages is typically limited to a "kid", with a value the OSCORE Sender ID associated with the message sender. In case Sender IDs are not unique among different OSCORE Security Contexts stored by the same peer, it is possible to disambiguate by additionally using a "kid context" identifying the OSCORE Security Context as a whole. Instead, response messages are not required to convey key identifier information, as the client can rely on the Token conveyed in the response for retrieving the Security Context to use (see above).¶
In Group OSCORE, the key identifier information in request messages always includes also a "kid context", whose value is used as identifier of the OSCORE group associated with the Security Context to use for security processing of the exchanged message. Response messages are also required to convey a "kid" as key identifier information (i.e., the OSCORE Sender ID associated with the message sender), if the corresponding request was protected with the group mode of Group OSCORE (see Section 8 of [I-D.ietf-core-oscore-groupcomm]) .¶
Some particular uses of (Group) OSCORE enable to build OSCORE-based tunneling [I-D.ietf-core-oscore-capable-proxies]. In such a case, a CoAP server might have to enforce that some OSCORE Security Contexts are not just looked up by a "kid" and similar key identifier information from the CoAP OSCORE Option in the incoming request to decrypt. Such a lookup should also rely on the alleged client's address, or on an alternative identifier of the tunnel from which the request came from.¶
Updating an OSCORE Security Context does not change or interfere with the values of the Token or Message ID used in the exchanged CoAP messages. However, if long-term exchanges are involved (e.g., CoAP Observations [RFC7641]), one has to be careful to ensure that updating the Security Context does not impair the security properties of (Group) OSCORE or result in other security vulnerabilities.¶
The following provides more details about key update, separately for OSCORE, KUDOS, and Group OSCORE.¶
OSCORE: [RFC8613] tacitly assumes that two peers terminate any ongoing CoAP Observation that they still have ongoing, upon installing a new OSCORE Security Context, irrespective of the method used to perform the key update.¶
On these premises, a belated response protected with the old OSCORE Security Context will fail decryption, as that Security Context is not available anymore on the receiving client.¶
KUDOS: The key update protocol KUDOS allows the two OSCORE peers to negotiate about preserving their ongoing CoAP Observations across the performed key update. If and only if both peers agree to do that during an execution of KUDOS, their Observations will remain active after installing the new OSCORE Security Context, which the two peers will use from then on to protect their exchanged Observe notifications.¶
Furthermore, irrespective of the method used to perform a key update, Section 3 of [I-D.ietf-core-oscore-key-update] updates the security protocol OSCORE [RFC8613] in order to prevent security issues that can arise from misbinding a request and a response, when those are protected with two different OSCORE Security Contexts.¶
Such an update to the OSCORE protocol requires a server to include its own Sender Sequence Number as Partial IV of an outgoing response, when protecting it with a Security Context different from the one used to protect the corresponding request. An exception safely applies to the response messages that are sent when running the key update procedure defined in Appendix B.2 of [RFC8613].¶
Group OSCORE: The Group Manager can distribute new group keying material to the members of an OSCORE group, by performing a group rekeying. When receiving updated group keying material from the Group Manager, either upon joining the group or by participating in a group rekeying, a group member uses that material to install a new, commonly shared Group OSCORE Security Context, which replaces the old one (if any is stored).¶
Also, Group OSCORE makes it possible for group members to safely preserve their ongoing active requests (e.g., CoAP Observations), also across the establishment of new Group OSCORE Security Contexts. This is achieved by virtue of how the Group Manager assigns and maintains the identifiers of OSCORE groups (see Section 3.2.1.1 of [I-D.ietf-core-oscore-groupcomm]).¶
Furthermore, analogous to the update that [I-D.ietf-core-oscore-key-update] makes on the OSCORE protocol with respect to protecting responses, Group OSCORE prevents security issues that can arise from misbinding a request and a response, when those are protected with two different Group OSCORE Security Contexts.¶
In the same way specified in Section 3 of [I-D.ietf-core-oscore-key-update] for OSCORE, Group OSCORE requires a server to include its own Sender Sequence Number as Partial IV of an outgoing response, when protecting it with a Security Context different from the one used to protect the corresponding request (see Sections 8.3 and 9.5 of [I-D.ietf-core-oscore-groupcomm]).¶
Enhancements may be called for optimizations such as Eclipse/Californium EndpointContextMatcher [CF-MATCHER] might not work properly unless both sides of the communication agree on the extent of the matching boxes. Mechanisms to indicate capabilities and choices selected may need to be defined; also, a way to define the progression of matching boxes needs to be defined that is compatible with the security properties of the underlying protocols. This may require new efforts, to be initiated based on some formative contributions.¶
PENDING.¶
Relevant starting points for retrieving existing discussion of this issue include:¶
https://mailarchive.ietf.org/arch/msg/core/biDJ8n4w0kBQATzyh9xHlKnGy1o/
("DTLS and Epochs")¶
https://mailarchive.ietf.org/arch/msg/core/TsyyKIHQ1FJtAvAo0ODNEt2Ileo/
("Connection ID")¶
https://github.com/core-wg/corrclar/issues/9
("Clarify/revise language around epochs in section 9.1.1 #9")¶
Section 12.3 of [RFC7252] established the CoAP Content-Formats Registry, which maps a combination of an Internet Media Type with an HTTP Content Coding, collectively called a Content-Format, to a concise numeric identifier for that Content-Format. The "Media Type" is more than a Media-Type-Name (see [RFC9193] for an extensive discussion), i.e., it may contain parameters beyond the mere combination of a type-name and a subtype-name registered in [IANA.media-types], as per [RFC6838], conventionally identified by the two names separated by a slash. This construct is often called a Content Type to reduce the confusion with a Media-Type-Name (e.g., in Section 8.3 of [RFC9110], which then however also opts to use the term Media Type for the same information set).¶
The second column of the Content-Format registry is the Content Coding, which is defined in Section 8.4.1 of [RFC9110]. For historical reasons, the HTTP header field that actually carries the content coding is called Content-Encoding; this often leads to the misnaming of Content Coding as "content encoding".¶
As has been noted in [Err4954], the text in Section 12.3 of [RFC7252] incorrectly uses these terms in the context of content types and content coding:¶
The field that describes the Content Type is called "Media Type".
This can lead to the misunderstanding that this column just carries
a Media-Type-Name (such as "text/plain
"), while it actually also
can carry media type parameters (as in "text/plain; charset=UTF-8
").¶
The field that describes the Content Coding uses the incorrect name "Content Encoding".¶
The VERIFIED Errata Report [Err4954] corrects the usage of "Content-Encoding" in the text and changes the name of the first column of the Content-Format registry to "Content Type" and the name of the second field to "Content Coding". This change has been carried out by IANA.¶
[Err4954] also has some notes on what would be valid or invalid Content-Format registrations. These are not repeated here; they are however quite useful as a reference when preparing additional Content-Format registrations.¶
This document makes no new requests to IANA.¶
Individual clarifications may contain IANA considerations; as for example in Section 2.6.¶
This document provides a number of corrections and clarifications to existing RFCs, but it does not make any changes with regard to the security aspects of the protocol. As a consequence, the security considerations of the referenced RFCs apply without additions.¶
(To be changed when that is no longer true; probably the security considerations will then be on the individual clarifications.)¶
The present document is modeled after RFC 4815 and the Internet-Drafts of the ROHC WG that led to it. Many thanks to the co-chairs of the ROHC WG and WG members that made this a worthwhile and successful experiment at the time.¶