Internet-Draft CoRE Resource Directory March 2021
Amsüss, et al. Expires 8 September 2021 [Page]
Workgroup:
CoRE
Internet-Draft:
draft-ietf-core-resource-directory-latest
Published:
Intended Status:
Standards Track
Expires:
Authors:
C. Amsüss, Ed.
Z. Shelby
ARM
M. Koster
SmartThings
C. Bormann
Universitaet Bremen TZI
P. van der Stok
consultant

CoRE Resource Directory

Abstract

In many IoT applications, direct discovery of resources is not practical due to sleeping nodes, or networks where multicast traffic is inefficient. These problems can be solved by employing an entity called a Resource Directory (RD), which contains information about resources held on other servers, allowing lookups to be performed for those resources. The input to an RD is composed of links and the output is composed of links constructed from the information stored in the RD. This document specifies the web interfaces that an RD supports for web servers to discover the RD and to register, maintain, lookup and remove information on resources. Furthermore, new target attributes useful in conjunction with an RD are defined.

Note to Readers

Discussion of this document takes place on the CORE Working Group mailing list (core@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/core/.

Source for this draft and an issue tracker can be found at https://github.com/core-wg/resource-directory.

Status of This Memo

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 8 September 2021.

Table of Contents

1. Introduction

In the work on Constrained RESTful Environments (CoRE), a REST architecture suitable for constrained nodes (e.g. with limited RAM and ROM [RFC7228]) and networks (e.g. 6LoWPAN [RFC4944]) has been established and is used in Internet-of-Things (IoT) or machine-to-machine (M2M) applications such as smart energy and building automation.

The discovery of resources offered by a constrained server is very important in machine-to-machine applications where there are no humans in the loop and static interfaces result in fragility. The discovery of resources provided by an HTTP Web Server is typically called Web Linking [RFC8288]. The use of Web Linking for the description and discovery of resources hosted by constrained web servers is specified by the CoRE Link Format [RFC6690]. However, [RFC6690] only describes how to discover resources from the web server that hosts them by querying /.well-known/core. In many constrained scenarios, direct discovery of resources is not practical due to sleeping nodes, or networks where multicast traffic is inefficient. These problems can be solved by employing an entity called a Resource Directory (RD), which contains information about resources held on other servers, allowing lookups to be performed for those resources.

This document specifies the web interfaces that an RD supports for web servers to discover the RD and to register, maintain, lookup and remove information on resources. Furthermore, new target attributes useful in conjunction with an RD are defined. Although the examples in this document show the use of these interfaces with CoAP [RFC7252], they can be applied in an equivalent manner to HTTP [RFC7230].

2. Terminology

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.

The term "byte" is used in its now customary sense as a synonym for "octet".

This specification requires readers to be familiar with all the terms and concepts that are discussed in [RFC3986], [RFC8288] and [RFC6690]. Readers should also be familiar with the terms and concepts discussed in [RFC7252]. To describe the REST interfaces defined in this specification, the URI Template format is used [RFC6570].

This specification makes use of the following additional terminology:

resolve against

The expression "a URI-reference is resolved against a base URI" is used to describe the process of [RFC3986] Section 5.2. Noteworthy corner cases are that if the URI-reference is a (full) URI and resolved against any base URI, that gives the original full URI, and that resolving an empty URI reference gives the base URI without any fragment identifier.

Resource Directory (RD)

A web entity that stores information about web resources and implements the REST interfaces defined in this specification for discovery, for the creation, maintenance and removal of registrations, and for lookup of the registered resources.

Sector

In the context of an RD, a sector is a logical grouping of endpoints.

The abbreviation "d=" is used for the sector in query parameters for compatibility with deployed implementations.

Endpoint

Endpoint (EP) is a term used to describe a web server or client in [RFC7252]. In the context of this specification an endpoint is used to describe a web server that registers resources to the RD. An endpoint is identified by its endpoint name, which is included during registration, and has a unique name within the associated sector of the registration.

Registration Base URI

The Base URI of a Registration is a URI that typically gives scheme and authority information about an Endpoint. The Registration Base URI is provided at registration time, and is used by the RD to resolve relative references of the registration into URIs.

Target

The target of a link is the destination address (URI) of the link. It is sometimes identified with "href=", or displayed as <target>. Relative targets need resolving with respect to the Base URI (section 5.2 of [RFC3986]).

This use of the term Target is consistent with [RFC8288]'s use of the term.

Context

The context of a link is the source address (URI) of the link, and describes which resource is linked to the target. A link's context is made explicit in serialized links as the "anchor=" attribute.

This use of the term Context is consistent with [RFC8288]'s use of the term.

Directory Resource

A resource in the RD containing registration resources.

Registration Resource

A resource in the RD that contains information about an Endpoint and its links.

Commissioning Tool

Commissioning Tool (CT) is a device that assists during installation events by assigning values to parameters, naming endpoints and groups, or adapting the installation to the needs of the applications.

Registrant-ep

Registrant-ep is the endpoint that is registered into the RD. The registrant-ep can register itself, or a CT registers the registrant-ep.

RDAO

Resource Directory Address Option. A new IPv6 Neighbor Discovery option defined for announcing an RD's address.

3. Architecture and Use Cases

3.1. Principles

The RD is primarily a tool to make discovery operations more efficient than querying /.well-known/core on all connected devices, or across boundaries that would limit those operations.

It provides information about resources hosted by other devices that could otherwise only be obtained by directly querying the /.well-known/core resource on these other devices, either by a unicast request or a multicast request.

Information SHOULD only be stored in the RD if it can be obtained by querying the described device's /.well-known/core resource directly.

Data in the RD can only be provided by the device which hosts those data or a dedicated Commissioning Tool (CT). These CTs act on behalf of endpoints too constrained, or generally unable, to present that information themselves. No other client can modify data in the RD. Changes to the information in the RD do not propagate automatically back to the web servers from where the information originated.

3.2. Architecture

The RD architecture is illustrated in Figure 1. An RD is used as a repository of registrations describing resources hosted on other web servers, also called endpoints (EP). An endpoint is a web server associated with a scheme, IP address and port. A physical node may host one or more endpoints. The RD implements a set of REST interfaces for endpoints to register and maintain RD registrations, and for endpoints to lookup resources from the RD. An RD can be logically segmented by the use of Sectors.

A mechanism to discover an RD using CoRE Link Format [RFC6690] is defined.

Registrations in the RD are soft state and need to be periodically refreshed.

An endpoint uses specific interfaces to register, update and remove a registration. It is also possible for an RD to fetch Web Links from endpoints and add their contents to its registrations.

At the first registration of an endpoint, a "registration resource" is created, the location of which is returned to the registering endpoint. The registering endpoint uses this registration resource to manage the contents of registrations.

A lookup interface for discovering any of the Web Links stored in the RD is provided using the CoRE Link Format.

             Registration         Lookup
              Interface         Interface
  +----+          |                 |
  | EP |----      |                 |
  +----+    ----  |                 |
                --|-    +------+    |
  +----+          | ----|      |    |     +--------+
  | EP | ---------|-----|  RD  |----|-----| Client |
  +----+          | ----|      |    |     +--------+
                --|-    +------+    |
  +----+    ----  |                 |
  | CT |----      |                 |
  +----+

Figure 1: The RD architecture.

A Registrant-EP MAY keep concurrent registrations to more than one RD at the same time if explicitly configured to do so, but that is not expected to be supported by typical EP implementations. Any such registrations are independent of each other. The usual expectation when multiple discovery mechanisms or addresses are configured is that they constitute a fall-back path for a single registration.

3.3. RD Content Model

The Entity-Relationship (ER) models shown in Figure 2 and Figure 3 model the contents of /.well-known/core and the RD respectively, with entity-relationship diagrams [ER]. Entities (rectangles) are used for concepts that exist independently. Attributes (ovals) are used for concepts that exist only in connection with a related entity. Relations (diamonds) give a semantic meaning to the relation between entities. Numbers specify the cardinality of the relations.

Some of the attribute values are URIs. Those values are always full URIs and never relative references in the information model. They can, however, be expressed as relative references in serializations, and often are.

These models provide an abstract view of the information expressed in link-format documents and an RD. They cover the concepts, but not necessarily all details of an RD's operation; they are meant to give an overview, and not be a template for implementations.

                    +----------------------+
                    |   /.well-known/core  |
                    +----------------------+
                               |
                               | 1
                       ////////\\\\\\\
                      <    contains   >
                       \\\\\\\\///////
                               |
                               | 0+
                     +--------------------+
                     |      link          |
                     +--------------------+
                               |
                               |  1   oooooooo
                               +-----o target o
                               |      oooooooo
          oooooooooooo   0+    |
         o    target  o--------+
         o  attribute o        | 0+   oooooo
          oooooooooooo         +-----o rel  o
                               |      oooooo
                               |
                               | 1    ooooooooo
                               +-----o context o
                                      ooooooooo



Figure 2: ER Model of the content of /.well-known/core

The model shown in Figure 2 models the contents of /.well-known/core which contains:

  • a set of links belonging to the hosting web server

The web server is free to choose links it deems appropriate to be exposed in its /.well-known/core. Typically, the links describe resources that are served by the host, but the set can also contain links to resources on other servers (see examples in [RFC6690] page 14). The set does not necessarily contain links to all resources served by the host.

A link has the following attributes (see [RFC8288]):

  • Zero or more link relations: They describe relations between the link context and the link target.

    In link-format serialization, they are expressed as space-separated values in the "rel" attribute, and default to "hosts".

  • A link context URI: It defines the source of the relation, e.g. who "hosts" something.

    In link-format serialization, it is expressed in the "anchor" attribute and defaults to the Origin of the target (practically: the target with its path and later components removed)

  • A link target URI: It defines the destination of the relation (e.g. what is hosted), and is the topic of all target attributes.

    In link-format serialization, it is expressed between angular brackets, and sometimes called the "href".

  • Other target attributes (e.g. resource type (rt), interface (if), or content format (ct)). These provide additional information about the target URI.
                 +--------------+
                 +      RD      +
                 +--------------+
                        | 1
                        |
                        |
                        |
                        |
                   //////\\\\
                  < contains >
                   \\\\\/////
                        |
                     0+ |
 ooooooo     1  +---------------+
o  base o-------|  registration |
 ooooooo        +---------------+
                    |       | 1
                    |       +--------------+
       oooooooo   1 |                      |
      o  href  o----+                 /////\\\\
       oooooooo     |                < contains >
                    |                 \\\\\/////
       oooooooo   1 |                      |
      o   ep   o----+                      | 0+
       oooooooo     |             +------------------+
                    |             |      link        |
       oooooooo 0-1 |             +------------------+
      o    d   o----+                      |
       oooooooo     |                      |  1   oooooooo
                    |                      +-----o target o
       oooooooo   1 |                      |      oooooooo
      o   lt   o----+     ooooooooooo   0+ |
       oooooooo     |    o  target   o-----+
                    |    o attribute o     | 0+   oooooo
    ooooooooooo 0+  |     ooooooooooo      +-----o rel  o
   o  endpoint o----+                      |      oooooo
   o attribute o                           |
    ooooooooooo                            | 1   ooooooooo
                                           +----o context o
                                                 ooooooooo
Figure 3: ER Model of the content of the RD

The model shown in Figure 3 models the contents of the RD which contains in addition to /.well-known/core:

  • 0 to n Registrations of endpoints,

A registration is associated with one endpoint. A registration defines a set of links as defined for /.well-known/core. A Registration has six types of attributes:

  • an endpoint name ("ep", a Unicode string) unique within a sector
  • a Registration Base URI ("base", a URI typically describing the scheme://authority part)
  • a lifetime ("lt"),
  • a registration resource location inside the RD ("href"),
  • optionally a sector ("d", a Unicode string)
  • optional additional endpoint attributes (from Section 9.3)

The cardinality of "base" is currently 1; future documents are invited to extend the RD specification to support multiple values (e.g. [I-D.silverajan-core-coap-protocol-negotiation]). Its value is used as a Base URI when resolving URIs in the links contained in the endpoint.

Links are modelled as they are in Figure 2.

Registration Base URIs can contain link-local IP addresses. To be usable across hosts, those cannot be serialized to contain zone identifiers (see [RFC6874] Section 1).

Link-local addresses can only be used on a single link (therefore RD servers cannot announce them when queried on a different link), and lookup clients using them need to keep track of which interface they got them from.

Therefore, it is advisable in many scenarios to use addresses with larger scope if available.

3.5. Use Case: Cellular M2M

Over the last few years, mobile operators around the world have focused on development of M2M solutions in order to expand the business to the new type of users: machines. The machines are connected directly to a mobile network using an appropriate embedded wireless interface (GSM/GPRS, WCDMA, LTE) or via a gateway providing short and wide range wireless interfaces. The ambition in such systems is to build them from reusable components. These speed up development and deployment, and enable shared use of machines across different applications. One crucial component of such systems is the discovery of resources (and thus the endpoints they are hosted on) capable of providing required information at a given time or acting on instructions from the end users.

Imagine a scenario where endpoints installed on vehicles enable tracking of the position of these vehicles for fleet management purposes and allow monitoring of environment parameters. During the boot-up process endpoints register with an RD, which is hosted by the mobile operator or somewhere in the cloud. Periodically, these endpoints update their registration and may modify resources they offer.

When endpoints are not always connected, for example because they enter a sleep mode, a remote server is usually used to provide proxy access to the endpoints. Mobile apps or web applications for environment monitoring contact the RD, look up the endpoints capable of providing information about the environment using an appropriate set of link parameters, obtain information on how to contact them (URLs of the proxy server), and then initiate interaction to obtain information that is finally processed, displayed on the screen and usually stored in a database. Similarly, fleet management systems provide the appropriate link parameters to the RD to look up for EPs deployed on the vehicles the application is responsible for.

3.6. Use Case: Home and Building Automation

Home and commercial building automation systems can benefit from the use of IoT web services. The discovery requirements of these applications are demanding. Home automation usually relies on run-time discovery to commission the system, whereas in building automation a combination of professional commissioning and run-time discovery is used. Both home and building automation involve peer-to-peer interactions between endpoints, and involve battery-powered sleeping devices. Both can use the common RD infrastructure to establish device interactions efficiently, but can pick security policies suitable for their needs.

Two phases can be discerned for a network servicing the system: (1) installation and (2) operation. During the operational phase, the network is connected to the Internet with a Border Router (e.g. a 6LoWPAN Border Router (6LBR), see [RFC6775]) and the nodes connected to the network can use the Internet services that are provided by the Internet Provider or the network administrator. During the installation phase, the network is completely stand-alone, no Border Router is connected, and the network only supports the IP communication between the connected nodes. The installation phase is usually followed by the operational phase. As an RD's operations work without hard dependencies on names or addresses, it can be used for discovery across both phases.

Resources may be shared through data brokers that have no knowledge beforehand of who is going to consume the data. An RD can be used to hold links about resources and services hosted anywhere to make them discoverable by a general class of applications.

For example, environmental and weather sensors that generate data for public consumption may provide data to an intermediary server, or broker. Sensor data are published to the intermediary upon changes or at regular intervals. Descriptions of the sensors that resolve to links to sensor data may be published to an RD. Applications wishing to consume the data can use RD Lookup to discover and resolve links to the desired resources and endpoints. The RD service need not be coupled with the data intermediary service. Mapping of RDs to data intermediaries may be many-to-many.

Metadata in web link formats like [RFC6690] which may be internally stored as triples, or relation/attribute pairs providing metadata about resource links, need to be supported by RDs. External catalogues that are represented in other formats may be converted to common web linking formats for storage and access by RDs. Since it is common practice for these to be encoded in URNs [RFC8141], simple and lossless structural transforms should generally be sufficient to store external metadata in RDs.

The additional features of an RD allow sectors to be defined to enable access to a particular set of resources from particular applications. This provides isolation and protection of sensitive data when needed. Application groups with multicast addresses may be defined to support efficient data transport.

4. RD discovery and other interface-independent components

This and the following sections define the required set of REST interfaces between an RD, endpoints and lookup clients. Although the examples throughout these sections assume the use of CoAP [RFC7252], these REST interfaces can also be realized using HTTP [RFC7230]. The multicast discovery and simple registration operations are exceptions to that, as they rely on mechanisms unavailable in HTTP. In all definitions in these sections, both CoAP response codes (with dot notation) and HTTP response codes (without dot notation) are shown. An RD implementing this specification MUST support the discovery, registration, update, lookup, and removal interfaces.

All operations on the contents of the RD MUST be atomic and idempotent.

For several operations, interface templates are given in list form; those describe the operation participants, request codes, URIs, content formats and outcomes. Sections of those templates contain normative content about Interaction, Method, URI Template and URI Template Variables as well as the details of the Success condition. The additional sections on options like Content-Format and on Failure codes give typical cases that an implementation of the RD should deal with. Those serve to illustrate the typical responses to readers who are not yet familiar with all the details of CoAP based interfaces; they do not limit what a server may respond under atypical circumstances.

REST clients (registrant-EPs and CTs during registration and maintenance, lookup clients, RD servers during simple registrations) must be prepared to receive any unsuccessful code and act upon it according to its definition, options and/or payload to the best of their capabilities, falling back to failing the operation if recovery is not possible. In particular, they SHOULD retry the request upon 5.03 (Service Unavailable; 503 in HTTP) according to the Max-Age (Retry-After in HTTP) option, and SHOULD fall back to link-format when receiving 4.15 (Unsupported Content-Format; 415 in HTTP).

An RD MAY make the information submitted to it available to further directories (subject to security policies on link confidentiality), if it can ensure that a loop does not form. The protocol used between directories to ensure loop-free operation is outside the scope of this document.

4.1. Finding a Resource Directory

A (re-)starting device may want to find one or more RDs before it can discover their URIs. Dependent on the operational conditions, one or more of the techniques below apply.

The device may be pre-configured to exercise specific mechanisms for finding the RD:

  1. It may be configured with a specific IP address for the RD. That IP address may also be an anycast address, allowing the network to forward RD requests to an RD that is topologically close; each target network environment in which some of these preconfigured nodes are to be brought up is then configured with a route for this anycast address that leads to an appropriate RD. (Instead of using an anycast address, a multicast address can also be preconfigured. The RD servers then need to configure one of their interfaces with this multicast address.)
  2. It may be configured with a DNS name for the RD and use DNS to return the IP address of the RD; it can find a DNS server to perform the lookup using the usual mechanisms for finding DNS servers.
  3. It may be configured to use a service discovery mechanism such as DNS-SD, as outlined in Section 4.1.2.

For cases where the device is not specifically configured with a way to find an RD, the network may want to provide a suitable default.

  1. The IPv6 Neighbor Discovery option RDAO Section 4.1.1 can do that.
  2. When DHCP is in use, this could be provided via a DHCP option (no such option is defined at the time of writing).

Finally, if neither the device nor the network offers any specific configuration, the device may want to employ heuristics to find a suitable RD.

The present specification does not fully define these heuristics, but suggests a number of candidates:

  1. In a 6LoWPAN, just assume the Border Router (6LBR) can act as an RD (using the ABRO option to find that [RFC6775]). Confirmation can be obtained by sending a unicast to coap://[6LBR]/.well-known/core?rt=core.rd*.
  2. In a network that supports multicast well, discovering the RD using a multicast query for /.well-known/core as specified in CoRE Link Format [RFC6690]: Sending a Multicast GET to coap://[MCD1]/.well-known/core?rt=core.rd*. RDs within the multicast scope will answer the query.

When answering a multicast request directed at a link-local group, the RD may want to respond from a routable address; this makes it easier for registrants to use one of their own routable addresses for registration. When [RFC6724] is used for source address selection, this can be achieved by applying the changes of its Section 10.4, picking public addresses in its Section 5 Rule 7, and superseding rule 8 with preferring the source address's precedence.

As some of the RD addresses obtained by the methods listed here are just (more or less educated) guesses, endpoints MUST make use of any error messages to very strictly rate-limit requests to candidate IP addresses that don't work out. For example, an ICMP Destination Unreachable message (and, in particular, the port unreachable code for this message) may indicate the lack of a CoAP server on the candidate host, or a CoAP error response code such as 4.05 "Method Not Allowed" may indicate unwillingness of a CoAP server to act as a directory server.

The following RD discovery mechanisms are recommended:

  • In managed networks with border routers that need stand-alone operation, the RDAO option is recommended (e.g. operational phase described in Section 3.6).
  • In managed networks without border router (no Internet services available), the use of a preconfigured anycast address is recommended (e.g. installation phase described in Section 3.6).
  • In networks managed using DNS-SD, the use of DNS-SD for discovery as described in Section 4.1.2 is recommended.

The use of multicast discovery in mesh networks is NOT RECOMMENDED.

4.1.1. Resource Directory Address Option (RDAO)

The Resource Directory Address Option (RDAO) carries information about the address of the RD in RAs (Router Advertisements) of IPv6 Neighbor Discovery (ND), similar to how RDNSS options [RFC8106] are sent. This information is needed when endpoints cannot discover the RD with a link-local or realm-local scope multicast address, for instance because the endpoint and the RD are separated by a Border Router (6LBR). In many circumstances the availability of DHCP cannot be guaranteed either during commissioning of the network. The presence and the use of the RD is essential during commissioning.

It is possible to send multiple RDAO options in one message, indicating as many RD addresses.

The RDAO format is:

0                   1                   2                   3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|     Type      |  Length = 3   |          Reserved             |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                        Valid Lifetime                         |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                                                               |
+                                                               +
|                                                               |
+                          RD Address                           +
|                                                               |
+                                                               +
|                                                               |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Fields:

Type:                   TBD38

Length:                 8-bit unsigned integer.  The length of
                        the option in units of 8 bytes.
                        Always 3.

Reserved:               This field is unused.  It MUST be
                        initialized to zero by the sender and
                        MUST be ignored by the receiver.

Valid Lifetime:         32-bit unsigned integer.  The length of
                        time in seconds (relative to
                        the time the packet is received) that
                        this RD address is valid.
                        A value of all zero bits (0x0) indicates
                        that this RD address
                        is not valid anymore.

RD Address:             IPv6 address of the RD.
Figure 4: Resource Directory Address Option

4.1.2. Using DNS-SD to discover a Resource Directory

An RD can advertise its presence in DNS-SD [RFC6763] using the service name _core-rd._udp (for CoAP), _core-rd-dtls._udp (for CoAP over DTLS), _core-rd._tcp (for CoAP over TCP) or _core-rd-tls._tcp (for CoAP over TLS) defined in this document. (For the WebSocket transports of CoAP, no service is defined as DNS-SD is typically unavailable in environments where CoAP over WebSockets is used).

The selection of the service indicates the protocol used, and the SRV record points the client to a host name and port to use as a starting point for the URI discovery steps of Section 4.3.

This section is a simplified concrete application of the more generic mechanism specified in [I-D.ietf-core-rd-dns-sd].

4.2. Payload Content Formats

RDs implementing this specification MUST support the application/link-format content format (ct=40).

RDs implementing this specification MAY support additional content formats.

Any additional content format supported by an RD implementing this specification SHOULD be able to express all the information expressible in link-format. It MAY be able to express information that is inexpressible in link-format, but those expressions SHOULD be avoided where possible.

4.3. URI Discovery

Before an endpoint can make use of an RD, it must first know the RD's address and port, and the URI path information for its REST APIs. This section defines discovery of the RD and its URIs using the well-known interface of the CoRE Link Format [RFC6690] after having discovered a host as described in Section 4.1.

Discovery of the RD registration URI is performed by sending either a multicast or unicast GET request to /.well-known/core and including a Resource Type (rt) parameter [RFC6690] with the value "core.rd" in the query string. Likewise, a Resource Type parameter value of "core.rd-lookup*" is used to discover the URIs for RD Lookup operations, core.rd* is used to discover all URIs for RD operations. Upon success, the response will contain a payload with a link format entry for each RD function discovered, indicating the URI of the RD function returned and the corresponding Resource Type. When performing multicast discovery, the multicast IP address used will depend on the scope required and the multicast capabilities of the network (see Section 9.5).

An RD MAY provide hints about the content-formats it supports in the links it exposes or registers, using the "ct" target attribute, as shown in the example below. Clients MAY use these hints to select alternate content-formats for interaction with the RD.

HTTP does not support multicast and consequently only unicast discovery can be supported at the using the HTTP /.well-known/core resource.

RDs implementing this specification MUST support query filtering for the rt parameter as defined in [RFC6690].

While the link targets in this discovery step are often expressed in path-absolute form, this is not a requirement. Clients of the RD SHOULD therefore accept URIs of all schemes they support, both as URIs and relative references, and not limit the set of discovered URIs to those hosted at the address used for URI discovery.

With security policies where the client requires the RD to be authorized to act as an RD, that authorization may be limited to resources on which the authorized RD advertises the adequate resource types. Clients that have obtained links they can not rely on yet can repeat the URI discovery step at the /.well-known/core resource of the indicated host to obtain the resource type information from an authorized source.

The URI Discovery operation can yield multiple URIs of a given resource type. The client of the RD can use any of the discovered addresses initially.

The discovery request interface is specified as follows (this is exactly the Well-Known Interface of [RFC6690] Section 4, with the additional requirement that the server MUST support query filtering):

Interaction:

EP, CT or Client -> RD

Method:

GET

URI Template:

/.well-known/core{?rt}

URI Template Variables:
rt :=

Resource Type. SHOULD contain one of the values "core.rd", "core.rd-lookup*", "core.rd-lookup-res", "core.rd-lookup-ep", or "core.rd*"

Accept:

absent, application/link-format or any other media type representing web links

The following response is expected on this interface:

Success:

2.05 "Content" or 200 "OK" with an application/link-format or other web link payload containing one or more matching entries for the RD resource.

The following example shows an endpoint discovering an RD using this interface, thus learning that the directory resource location, in this example, is /rd, and that the content-format delivered by the server hosting the resource is application/link-format (ct=40). Note that it is up to the RD to choose its RD locations.

Req: GET coap://[MCD1]/.well-known/core?rt=core.rd*

Res: 2.05 Content
Payload:
</rd>;rt=core.rd;ct=40,
</rd-lookup/ep>;rt=core.rd-lookup-ep;ct=40,
</rd-lookup/res>;rt=core.rd-lookup-res;ct=40
Figure 5: Example discovery exchange

The following example shows the way of indicating that a client may request alternate content-formats. The Content-Format code attribute "ct" MAY include a space-separated sequence of Content-Format codes as specified in Section 7.2.1 of [RFC7252], indicating that multiple content-formats are available. The example below shows the required Content-Format 40 (application/link-format) indicated as well as a CBOR and JSON representation from [I-D.ietf-core-links-json] (which have no numeric values assigned yet, so they are shown as TBD64 and TBD504 as in that draft). The RD resource locations /rd, and /rd-lookup are example values. The server in this example also indicates that it is capable of providing observation on resource lookups.

Req: GET coap://[MCD1]/.well-known/core?rt=core.rd*

Res: 2.05 Content
Payload:
</rd>;rt=core.rd;ct="40 65225",
</rd-lookup/res>;rt=core.rd-lookup-res;ct="40 TBD64 TBD504";obs,
</rd-lookup/ep>;rt=core.rd-lookup-ep;ct="40 TBD64 TBD504"
Figure 6: Example discovery exchange indicating additional content-formats

For maintenance, management and debugging, it can be useful to identify the components that constitute the RD server. The identification can be used to find client-server incompatibilities, supported features, required updates and other aspects. The Well-Known interface described in Section 4 of [RFC6690] can be used to find such data.

It would typically be stored in an implementation information link (as described in [I-D.bormann-t2trg-rel-impl]):

Req: GET /.well-known/core?rel=impl-info

Res: 2.05 Content
Payload:
<http://software.example.com/shiny-resource-directory/1.0beta1>;
    rel=impl-info
Figure 7: Example exchange of obtaining implementation information, using the relation type currently proposed in the work-in-progress document

Note that depending on the particular server's architecture, such a link could be anchored at the RD server's root (as in this example), or at individual RD components. The latter is to be expected when different applications are run on the same server.

5. Registration

After discovering the location of an RD, a registrant-ep or CT MAY register the resources of the registrant-ep using the registration interface. This interface accepts a POST from an endpoint containing the list of resources to be added to the directory as the message payload in the CoRE Link Format [RFC6690] or other representations of web links, along with query parameters indicating the name of the endpoint, and optionally the sector, lifetime and base URI of the registration. It is expected that other specifications will define further parameters (see Section 9.3). The RD then creates a new registration resource in the RD and returns its location. The receiving endpoint MUST use that location when refreshing registrations using this interface. Registration resources in the RD are kept active for the period indicated by the lifetime parameter. The creating endpoint is responsible for refreshing the registration resource within this period using either the registration or update interface. The registration interface MUST be implemented to be idempotent, so that registering twice with the same endpoint parameters ep and d (sector) does not create multiple registration resources.

The following rules apply for a registration request targeting a given (ep, d) value pair:

The posted link-format document can (and typically does) contain relative references both in its link targets and in its anchors, or contain empty anchors. The RD server needs to resolve these references in order to faithfully represent them in lookups. They are resolved against the base URI of the registration, which is provided either explicitly in the base parameter or constructed implicitly from the requester's URI as constructed from its network address and scheme.

For media types to which Appendix C applies (i.e. documents in application/link-format), request bodies MUST be expressed in Limited Link Format.

The registration request interface is specified as follows:

Interaction:

EP or CT -> RD

Method:

POST

URI Template:

{+rd}{?ep,d,lt,base,extra-attrs*}

URI Template Variables:
rd :=

RD registration URI (mandatory). This is the location of the RD, as obtained from discovery.

ep :=

Endpoint name (mostly mandatory). The endpoint name is an identifier that MUST be unique within a sector.

As the endpoint name is a Unicode string, it is encoded in UTF-8 (and possibly pct-encoded) during variable expansion (see [RFC6570] Section 3.2.1). The endpoint name MUST NOT contain any character in the inclusive ranges 0-31 or 127-159.

The maximum length of this parameter is 63 UTF-8 encoded bytes.

If the RD is configured to recognize the endpoint to be authorized to use exactly one endpoint name, the RD assigns that name. In that case, giving the endpoint name becomes optional for the client; if the client gives any other endpoint name, it is not authorized to perform the registration.

d :=

Sector (optional). The sector to which this endpoint belongs. When this parameter is not present, the RD MAY associate the endpoint with a configured default sector (possibly based on the endpoint's authorization) or leave it empty.

The sector is encoded like the ep parameter, and is limited to 63 UTF-8 encoded bytes as well.

lt :=

Lifetime (optional). Lifetime of the registration in seconds. Range of 1-4294967295. If no lifetime is included in the initial registration, a default value of 90000 (25 hours) SHOULD be assumed.

base :=

Base URI (optional). This parameter sets the base URI of the registration, under which the relative links in the payload are to be interpreted. The specified URI typically does not have a path component of its own, and MUST be suitable as a base URI to resolve any relative references given in the registration. The parameter is therefore usually of the shape "scheme://authority" for HTTP and CoAP URIs. The URI SHOULD NOT have a query or fragment component as any non-empty relative part in a reference would remove those parts from the resulting URI.

In the absence of this parameter the scheme of the protocol, source address and source port of the registration request are assumed. The Base URI is consecutively constructed by concatenating the used protocol's scheme with the characters "://", the requester's source address as an address literal and ":" followed by its port (if it was not the protocol's default one) in analogy to [RFC7252] Section 6.5.

This parameter is mandatory when the directory is filled by a third party such as an commissioning tool.

If the registrant-ep uses an ephemeral port to register with, it MUST include the base parameter in the registration to provide a valid network path.

A registrant that cannot be reached by potential lookup clients at the address it registers from (e.g. because it is behind some form of Network Address Translation (NAT)) MUST provide a reachable base address with its registration.

If the Base URI contains a link-local IP literal, it MUST NOT contain a Zone Identifier, and MUST be local to the link on which the registration request is received.

Endpoints that register with a base that contains a path component cannot efficiently express their registrations in Limited Link Format (Appendix C). Those applications should use different representations of links to which Appendix C is not applicable (e.g. [I-D.hartke-t2trg-coral]).

extra-attrs :=

Additional registration attributes (optional). The endpoint can pass any parameter registered at Section 9.3 to the directory. If the RD is aware of the parameter's specified semantics, it processes it accordingly. Otherwise, it MUST store the unknown key and its value(s) as an endpoint attribute for further lookup.

Content-Format:

application/link-format or any other indicated media type representing web links

The following response is expected on this interface:

Success:

2.01 "Created" or 201 "Created". The Location-Path option or Location header field MUST be included in the response. This location MUST be a stable identifier generated by the RD as it is used for all subsequent operations on this registration resource. The registration resource location thus returned is for the purpose of updating the lifetime of the registration and for maintaining the content of the registered links, including updating and deleting links.

A registration with an already registered ep and d value pair responds with the same success code and location as the original registration; the set of links registered with the endpoint is replaced with the links from the payload.

The location MUST NOT have a query or fragment component, as that could conflict with query parameters during the Registration Update operation. Therefore, the Location-Query option MUST NOT be present in a successful response.

If the registration fails, including request timeouts, or if delays from Service Unavailable responses with Max-Age or Retry-After accumulate to exceed the registrant's configured timeouts, it SHOULD pick another registration URI from the "URI Discovery" step and if there is only one or the list is exhausted, pick other choices from the "Finding a Resource Directory" step. Care has to be taken to consider the freshness of results obtained earlier, e.g. of the result of a /.well-known/core response, the lifetime of an RDAO option and of DNS responses. Any rate limits and persistent errors from the "Finding a Resource Directory" step must be considered for the whole registration time, not only for a single operation.

The following example shows a registrant-ep with the name "node1" registering two resources to an RD using this interface. The location "/rd" is an example RD location discovered in a request similar to Figure 5.

Req: POST coap://rd.example.com/rd?ep=node1
Content-Format: 40
Payload:
</sensors/temp>;rt=temperature-c;if=sensor,
<http://www.example.com/sensors/temp>;
  anchor="/sensors/temp";rel=describedby

Res: 2.01 Created
Location-Path: /rd/4521
Figure 8: Example registration payload

An RD may optionally support HTTP. Here is an example of almost the same registration operation above, when done using HTTP.

Req:
POST /rd?ep=node1&base=http://[2001:db8:1::1] HTTP/1.1
Host: rd.example.com
Content-Type: application/link-format

</sensors/temp>;rt=temperature-c;if=sensor,
<http://www.example.com/sensors/temp>;
  anchor="/sensors/temp";rel=describedby

Res:
HTTP/1.1 201 Created
Location: /rd/4521
Figure 9: Example registration payload as expressed using HTTP

5.1. Simple Registration

Not all endpoints hosting resources are expected to know how to upload links to an RD as described in Section 5. Instead, simple endpoints can implement the Simple Registration approach described in this section. An RD implementing this specification MUST implement Simple Registration. However, there may be security reasons why this form of directory discovery would be disabled.

This approach requires that the registrant-ep makes available the hosted resources that it wants to be discovered, as links on its /.well-known/core interface as specified in [RFC6690]. The links in that document are subject to the same limitations as the payload of a registration (with respect to Appendix C).

  • The registrant-ep finds one or more addresses of the directory server as described in Section 4.1.
  • The registrant-ep sends (and regularly refreshes with) a POST request to the /.well-known/rd URI of the directory server of choice. The body of the POST request is empty, and triggers the resource directory server to perform GET requests at the requesting registrant-ep's /.well-known/core to obtain the link-format payload to register.

    The registrant-ep includes the same registration parameters in the POST request as it would with a regular registration per Section 5. The registration base URI of the registration is taken from the registrant-ep's network address (as is default with regular registrations).

    Example request from registrant-EP to RD (unanswered until the next step):

Req: POST /.well-known/rd?lt=6000&ep=node1
(No payload)
Figure 10: First half example exchange of a simple registration
  • The RD queries the registrant-ep's discovery resource to determine the success of the operation. It SHOULD keep a cache of the discovery resource and not query it again as long as it is fresh.

    Example request from the RD to the registrant-EP:

Req: GET /.well-known/core
Accept: 40

Res: 2.05 Content
Content-Format: 40
Payload:
</sen/temp>
Figure 11: Example exchange of the RD querying the simple endpoint

With this response, the RD would answer the previous step's request:

The sequence of fetching the registration content before sending a successful response was chosen to make responses reliable, and the point about caching was chosen to still allow very constrained registrants. Registrants MUST be able to serve a GET request to /.well-known/core after having requested registration. Constrained devices MAY regard the initial request as temporarily failed when they need RAM occupied by their own request to serve the RD's GET, and retry later when the RD already has a cached representation of their discovery resources. Then, the RD can reply immediately and the registrant can receive the response.

The simple registration request interface is specified as follows:

Interaction:

EP -> RD

Method:

POST

URI Template:

/.well-known/rd{?ep,d,lt,extra-attrs*}

URI Template Variables are as they are for registration in Section 5. The base attribute is not accepted to keep the registration interface simple; that rules out registration over CoAP-over-TCP or HTTP that would need to specify one. For some time during this document's development, the URI template /.well-known/core{?ep,...} has been in use instead.

The following response is expected on this interface:

Success:

2.04 "Changed".

For the second interaction triggered by the above, the registrant-ep takes the role of server and the RD the role of client. (Note that this is exactly the Well-Known Interface of [RFC6690] Section 4):

Interaction:

RD -> EP

Method:

GET

URI Template:

/.well-known/core

The following response is expected on this interface:

Success:

2.05 "Content".

When the RD uses any authorization credentials to access the endpoint's discovery resource, or when it is deployed in a location where third parties might reach it but not the endpoint, it SHOULD verify that the apparent registrant-ep intends to register with the given registration parameters before revealing the obtained discovery information to lookup clients. An easy way to do that is to verify the simple registration request's sender address using the Echo option as described in [I-D.ietf-core-echo-request-tag] Section 2.4.

The RD MUST delete registrations created by simple registration after the expiration of their lifetime. Additional operations on the registration resource cannot be executed because no registration location is returned.

5.2. Third-party registration

For some applications, even Simple Registration may be too taxing for some very constrained devices, in particular if the security requirements become too onerous.

In a controlled environment (e.g. building control), the RD can be filled by a third party device, called a Commissioning Tool (CT). The commissioning tool can fill the RD from a database or other means. For that purpose scheme, IP address and port of the URI of the registered device is the value of the "base" parameter of the registration described in Section 5.

It should be noted that the value of the "base" parameter applies to all the links of the registration and has consequences for the anchor value of the individual links as exemplified in Appendix B. An eventual (currently non-existing) "base" attribute of the link is not affected by the value of "base" parameter in the registration.

5.3. Operations on the Registration Resource

This section describes how the registering endpoint can maintain the registrations that it created. The registering endpoint can be the registrant-ep or the CT. The registrations are resources of the RD.

An endpoint should not use this interface for registrations that it did not create. This is usually enforced by security policies, which in general require equivalent credentials for creation of and operations on a registration.

After the initial registration, the registering endpoint retains the returned location of the registration resource for further operations, including refreshing the registration in order to extend the lifetime and "keep-alive" the registration. When the lifetime of the registration has expired, the RD SHOULD NOT respond to discovery queries concerning this endpoint. The RD SHOULD continue to provide access to the registration resource after a registration time-out occurs in order to enable the registering endpoint to eventually refresh the registration. The RD MAY eventually remove the registration resource for the purpose of garbage collection. If the registration resource is removed, the corresponding endpoint will need to be re-registered.

The registration resource may also be used cancel the registration using DELETE, and to perform further operations beyond the scope of this specification.

Operations on the registration resource are sensitive to reordering; Section 5.3.4 describes how order is restored.

The operations on the registration resource are described below.

5.3.1. Registration Update

The update interface is used by the registering endpoint to refresh or update its registration with an RD. To use the interface, the registering endpoint sends a POST request to the registration resource returned by the initial registration operation.

An update MAY update registration parameters like lifetime, base URI or others. Parameters that are not being changed should not be included in an update. Adding parameters that have not changed increases the size of the message but does not have any other implications. Parameters are included as query parameters in an update operation as in Section 5.

A registration update resets the timeout of the registration to the (possibly updated) lifetime of the registration, independent of whether a lt parameter was given.

If the base URI of the registration is changed in an update, relative references submitted in the original registration or later updates are resolved anew against the new base.

The registration update operation only describes the use of POST with an empty payload. Future standards might describe the semantics of using content formats and payloads with the POST method to update the links of a registration (see Section 5.3.3).

The update registration request interface is specified as follows:

Interaction:

EP or CT -> RD

Method:

POST

URI Template:

{+location}{?lt,base,extra-attrs*}

URI Template Variables:
location :=

This is the Location returned by the RD as a result of a successful earlier registration.

lt :=

Lifetime (optional). Lifetime of the registration in seconds. Range of 1-4294967295. If no lifetime is included, the previous last lifetime set on a previous update or the original registration (falling back to 90000) SHOULD be used.

base :=

Base URI (optional). This parameter updates the Base URI established in the original registration to a new value, and is subject to the same restrictions as in the registration.

If the parameter is set in an update, it is stored by the RD as the new Base URI under which to interpret the relative links present in the payload of the original registration.

If the parameter is not set in the request but was set before, the previous Base URI value is kept unmodified.

If the parameter is not set in the request and was not set before either, the source address and source port of the update request are stored as the Base URI.

extra-attrs :=

Additional registration attributes (optional). As with the registration, the RD processes them if it knows their semantics. Otherwise, unknown attributes are stored as endpoint attributes, overriding any previously stored endpoint attributes of the same key.

Note that this default behavior does not allow removing an endpoint attribute in an update. For attributes whose functionality depends on the endpoints' ability to remove them in an update, it can make sense to define a value whose presence is equivalent to the absence of a value. As an alternative, an extension can define different updating rules for their attributes. That necessitates either discovery of whether the RD is aware of that extension, or tolerating the default behavior.

Content-Format:

none (no payload)

The following responses are expected on this interface:

Success:

2.04 "Changed" or 204 "No Content" if the update was successfully processed.

Failure:

4.04 "Not Found" or 404 "Not Found". Registration does not exist (e.g. may have been removed).

If the registration update fails in any way, including "Not Found" and request timeouts, or if the time indicated in a Service Unavailable Max-Age/Retry-After exceeds the remaining lifetime, the registering endpoint SHOULD attempt registration again.

The following example shows how the registering endpoint resets the timeout on its registration resource at an RD using this interface with the example location value: /rd/4521.

Req: POST /rd/4521

Res: 2.04 Changed
Figure 13: Example update of a registration

The following example shows the registering endpoint updating its registration resource at an RD using this interface with the example location value: /rd/4521. The initial registration by the registering endpoint set the following values:

  • endpoint name (ep)=endpoint1
  • lifetime (lt)=500
  • Base URI (base)=coap://local-proxy-old.example.com
  • payload of Figure 8

The initial state of the RD is reflected in the following request:

Req: GET /rd-lookup/res?ep=endpoint1

Res: 2.05 Content
Payload:
<coap://local-proxy-old.example.com/sensors/temp>;
    rt=temperature-c;if=sensor,
<http://www.example.com/sensors/temp>;
    anchor="coap://local-proxy-old.example.com/sensors/temp";
    rel=describedby
Figure 14: Example lookup before a change to the base address

The following example shows the registering endpoint changing the Base URI to coaps://new.example.com:5684:

Req: POST /rd/4521?base=coaps://new.example.com

Res: 2.04 Changed
Figure 15: Example registration update that changes the base address

The consecutive query returns:

Req: GET /rd-lookup/res?ep=endpoint1

Res: 2.05 Content
Payload:
<coaps://new.example.com/sensors/temp>;
    rt=temperature-c;if=sensor,
<http://www.example.com/sensors/temp>;
    anchor="coaps://new.example.com/sensors/temp";
    rel=describedby
Figure 16: Example lookup after a change to the base address

5.3.2. Registration Removal

Although RD registrations have soft state and will eventually timeout after their lifetime, the registering endpoint SHOULD explicitly remove an entry from the RD if it knows it will no longer be available (for example on shut-down). This is accomplished using a removal interface on the RD by performing a DELETE on the endpoint resource.

The removal request interface is specified as follows:

Interaction:

EP or CT -> RD

Method:

DELETE

URI Template:

{+location}

URI Template Variables:
location :=

This is the Location returned by the RD as a result of a successful earlier registration.

The following responses are expected on this interface:

Success:

2.02 "Deleted" or 204 "No Content" upon successful deletion

Failure:

4.04 "Not Found" or 404 "Not Found". Registration does not exist (e.g. may already have been removed).

The following examples shows successful removal of the endpoint from the RD with example location value /rd/4521.

Req: DELETE /rd/4521

Res: 2.02 Deleted
Figure 17: Example of a registration removal

5.3.4. Request freshness

Some security mechanisms usable with an RD allow out of order request processing, or do not even mandate replay protection at all. The RD needs to ensure that operations on the registration resource are executed in an order that does not distort the client's intentions.

This ordering of operations is expressed in terms of freshness as defined in [I-D.ietf-core-echo-request-tag]. Requests that alter a resource's state need to be fresh relative to the latest request that altered that state in a conflicting way.

An RD SHOULD determine a request's freshness, and MUST use the Echo option if it requires request freshness and can not determine the it in any other way. An endpoint MUST support the use of the Echo option. (One reason why an RD would not require freshness is when no relevant registration properties are covered by is security policies.)

5.3.4.1. Efficient use of Echo by an RD

To keep latency and traffic added by the freshness requirements to a minimum, RDs should avoid naive (sufficient but inefficient) freshness criteria.

Some simple mechanisms the RD can employ are:

  • State counter. The RD can keep a monotonous counter that increments whenever a registration changes. For every registration resource, it stores the post-increment value of that resource's last change. Requests altering them need to have at least that value encoded in their Echo option, and are otherwise rejected with a 4.01 Unauthorized and the current counter value as the Echo value. If other applications on the same server use Echo as well, that encoding may include a prefix indicating that it pertains to the RD's counter.

    The value associated with a resource needs to be kept across the removal of registrations if the same registration resource is to be reused.

    The counter can be reset (and the values of removed resources forgotten) when all previous security associations are reset.

    This is the "Persistent Counter" method of [I-D.ietf-core-echo-request-tag] Appendix A.

  • Preemptive Echo values. The current state counter can be sent in an Echo option not only when requests are rejected with 4.01 Unauthorized, but also with successful responses. Thus, clients can be provided with Echo values sufficient for their next request on a regular basis.

    While endpoints may discard received Echo values at leisure between requests, they are encouraged to retain these values for the next request to avoid additional round trips.

  • If the RD can ensure that only one security association has modifying access to any registration at any given time, and that security association provides order on the requests, that order is sufficient to show request freshness.
5.3.4.2. Examples of Echo usage

Figure 18 shows the interactions of an endpoint that has forgotten the server's latest Echo value and temporarily reduces its registration lifetime:

Req: POST /rd/4521?lt=7200

Res: 4.01 Unauthorized
Echo: 0x0123

(EP tries again immediately)

Req: POST /rd/4521?lt=7200
Echo: 0x0123

Res: 2.04 Changed
Echo: 0x0124

(Later the EP regains its confidence in its long-term reachability)

Req: POST /rd/4521?lt=90000
Echo: 0x0124

Res: 2.04 Changed
Echo: 0x0247
Figure 18: Example update of a registration

The other examples do not show Echo options for simplicity, and because they lack the context for any example values to have meaning.

6. RD Lookup

To discover the resources registered with the RD, a lookup interface must be provided. This lookup interface is defined as a default, and it is assumed that RDs may also support lookups to return resource descriptions in alternative formats (e.g. JSON or CBOR link format [I-D.ietf-core-links-json]) or using more advanced interfaces (e.g. supporting context or semantic based lookup) on different resources that are discovered independently.

RD Lookup allows lookups for endpoints and resources using attributes defined in this document and for use with the CoRE Link Format. The result of a lookup request is the list of links (if any) corresponding to the type of lookup. Thus, an endpoint lookup MUST return a list of endpoints and a resource lookup MUST return a list of links to resources.

The lookup type is selected by a URI endpoint, which is indicated by a Resource Type as per Table 1 below:

Table 1: Lookup Types
Lookup Type Resource Type Mandatory
Resource core.rd-lookup-res Mandatory
Endpoint core.rd-lookup-ep Mandatory

6.1. Resource lookup

Resource lookup results in links that are semantically equivalent to the links submitted to the RD by the registrant. The links and link parameters returned by the lookup are equal to the originally submitted ones, except that the target reference is fully resolved, and that the anchor reference is fully resolved if it is present in the lookup result at all.

Links that did not have an anchor attribute in the registration are returned without an anchor attribute. Links of which href or anchor was submitted as a (full) URI are returned with the respective attribute unmodified.

The above rules allow the client to interpret the response as links without any further knowledge of the storage conventions of the RD. The RD MAY replace the registration base URIs with a configured intermediate proxy, e.g. in the case of an HTTP lookup interface for CoAP endpoints.

If the base URI of a registration contains a link-local address, the RD MUST NOT show its links unless the lookup was made from the link on which the registered endpoint can be reached. The RD MUST NOT include zone identifiers in the resolved URIs.

6.2. Lookup filtering

Using the Accept Option, the requester can control whether the returned list is returned in CoRE Link Format (application/link-format, default) or in alternate content-formats (e.g. from [I-D.ietf-core-links-json]).

Multiple search criteria MAY be included in a lookup. All included criteria MUST match for a link to be returned. The RD MUST support matching with multiple search criteria.

A link matches a search criterion if it has an attribute of the same name and the same value, allowing for a trailing "*" wildcard operator as in Section 4.1 of [RFC6690]. Attributes that are defined as relation-types (in the link-format ABNF) match if the search value matches any of their values (see Section 4.1 of [RFC6690]; e.g. ?if=tag:example.net,2020:sensor matches ;if="example.regname tag:example.net,2020:sensor";). A resource link also matches a search criterion if its endpoint would match the criterion, and vice versa, an endpoint link matches a search criterion if any of its resource links matches it.

Note that href is a valid search criterion and matches target references. Like all search criteria, on a resource lookup it can match the target reference of the resource link itself, but also the registration resource of the endpoint that registered it. Queries for resource link targets MUST be in URI form (i.e. not relative references) and are matched against a resolved link target. Queries for endpoints SHOULD be expressed in path-absolute form if possible and MUST be expressed in URI form otherwise; the RD SHOULD recognize either. The anchor attribute is usable for resource lookups, and, if queried, MUST be in URI form as well.

Additional query parameters "page" and "count" are used to obtain lookup results in specified increments using pagination, where count specifies how many links to return and page specifies which subset of links organized in sequential pages, each containing 'count' links, starting with link zero and page zero. Thus, specifying count of 10 and page of 0 will return the first 10 links in the result set (links 0-9). Count = 10 and page = 1 will return the next 'page' containing links 10-19, and so on. Unlike block-wise transfer of a compelte result set, these parameters ensure that each chunk of results can be interpreted on its own. This simplifies the processing, but can result in duplicate or missed items when coinciding with changes from the registration interface.

Endpoints that are interested in a lookup result repeatedly or continuously can use mechanisms like ETag caching, resource observation ([RFC7641]), or any future mechanism that might allow more efficient observations of collections. These are advertised, detected and used according to their own specifications and can be used with the lookup interface as with any other resource.

When resource observation is used, every time the set of matching links changes, or the content of a matching link changes, the RD sends a notification with the matching link set. The notification contains the successful current response to the given request, especially with respect to representing zero matching links (see "Success" item below).

The lookup interface is specified as follows:

Interaction:

Client -> RD

Method:

GET

URI Template:

{+type-lookup-location}{?page,count,search*}

URI Template Variables:
type-lookup-location :=

RD Lookup URI for a given lookup type (mandatory). The address is discovered as described in Section 4.3.

search :=

Search criteria for limiting the number of results (optional).

The search criteria are an associative array, expressed in a form-style query as per the URI template (see [RFC6570] Sections 2.4.2 and 3.2.8)

page :=

Page (optional). Parameter cannot be used without the count parameter. Results are returned from result set in pages that contain 'count' links starting from index (page * count). Page numbering starts with zero.

count :=

Count (optional). Number of results is limited to this parameter value. If the page parameter is also present, the response MUST only include 'count' links starting with the (page * count) link in the result set from the query. If the count parameter is not present, then the response MUST return all matching links in the result set. Link numbering starts with zero.

Accept:

absent, application/link-format or any other indicated media type representing web links

The following responses codes are defined for this interface:

Success:

2.05 "Content" or 200 "OK" with an application/link-format or other web link payload containing matching entries for the lookup.

The payload can contain zero links (which is an empty payload in [RFC6690] link format, but could also be [] in JSON based formats), indicating that no entities matched the request.

6.3. Resource lookup examples

The examples in this section assume the existence of CoAP hosts with a default CoAP port 61616. HTTP hosts are possible and do not change the nature of the examples.

The following example shows a client performing a resource lookup with the example resource look-up locations discovered in Figure 5:

Req: GET /rd-lookup/res?rt=tag:example.org,2020:temperature

Res: 2.05 Content
Payload:
<coap://[2001:db8:3::123]:61616/temp>;
    rt="tag:example.org,2020:temperature"
Figure 19: Example a resource lookup

A client that wants to be notified of new resources as they show up can use observation:

Req: GET /rd-lookup/res?rt=tag:example.org,2020:light
Observe: 0

Res: 2.05 Content
Observe: 23
Payload: empty

(at a later point in time)

Res: 2.05 Content
Observe: 24
Payload:
<coap://[2001:db8:3::124]/west>;rt="tag:example.org,2020:light",
<coap://[2001:db8:3::124]/south>;rt="tag:example.org,2020:light",
<coap://[2001:db8:3::124]/east>;rt="tag:example.org,2020:light"
Figure 20: Example an observing resource lookup

The following example shows a client performing a paginated resource lookup

Req: GET /rd-lookup/res?page=0&count=5

Res: 2.05 Content
Payload:
<coap://[2001:db8:3::123]:61616/res/0>;ct=60,
<coap://[2001:db8:3::123]:61616/res/1>;ct=60,
<coap://[2001:db8:3::123]:61616/res/2>;ct=60,
<coap://[2001:db8:3::123]:61616/res/3>;ct=60,
<coap://[2001:db8:3::123]:61616/res/4>;ct=60

Req: GET /rd-lookup/res?page=1&count=5

Res: 2.05 Content
Payload:
<coap://[2001:db8:3::123]:61616/res/5>;ct=60,
<coap://[2001:db8:3::123]:61616/res/6>;ct=60,
<coap://[2001:db8:3::123]:61616/res/7>;ct=60,
<coap://[2001:db8:3::123]:61616/res/8>;ct=60,
<coap://[2001:db8:3::123]:61616/res/9>;ct=60
Figure 21: Examples of paginated resource lookup

The following example shows a client performing a lookup of all resources of all endpoints of a given endpoint type. It assumes that two endpoints (with endpoint names sensor1 and sensor2) have previously registered with their respective addresses coap://sensor1.example.com and coap://sensor2.example.com, and posted the very payload of the 6th response of section 5 of [RFC6690].

It demonstrates how absolute link targets stay unmodified, while relative ones are resolved:

Req: GET /rd-lookup/res?et=tag:example.com,2020:platform

Res: 2.05 Content
Payload:
<coap://sensor1.example.com/sensors>;ct=40;title="Sensor Index",
<coap://sensor1.example.com/sensors/temp>;rt=temperature-c;if=sensor,
<coap://sensor1.example.com/sensors/light>;rt=light-lux;if=sensor,
<http://www.example.com/sensors/t123>;rel=describedby;
    anchor="coap://sensor1.example.com/sensors/temp",
<coap://sensor1.example.com/t>;rel=alternate;
    anchor="coap://sensor1.example.com/sensors/temp",
<coap://sensor2.example.com/sensors>;ct=40;title="Sensor Index",
<coap://sensor2.example.com/sensors/temp>;rt=temperature-c;if=sensor,
<coap://sensor2.example.com/sensors/light>;rt=light-lux;if=sensor,
<http://www.example.com/sensors/t123>;rel=describedby;
    anchor="coap://sensor2.example.com/sensors/temp",
<coap://sensor2.example.com/t>;rel=alternate;
    anchor="coap://sensor2.example.com/sensors/temp"
Figure 22: Example of resource lookup from multiple endpoints

6.4. Endpoint lookup

The endpoint lookup returns links to and information about registration resources, which themselves can only be manipulated by the registering endpoint.

Endpoint registration resources are annotated with their endpoint names (ep), sectors (d, if present) and registration base URI (base; reports the registrant-ep's address if no explicit base was given) as well as a constant resource type (rt="core.rd-ep"); the lifetime (lt) is not reported. Additional endpoint attributes are added as target attributes to their endpoint link unless their specification says otherwise.

Links to endpoints SHOULD be presented in path-absolute form or, if required, as (full) URIs. (This ensures that the output conforms to Limited Link Format as described in Appendix C.)

Base addresses that contain link-local addresses MUST NOT include zone identifiers, and such registrations MUST NOT be shown unless the lookup was made from the same link from which the registration was made.

While Endpoint Lookup does expose the registration resources, the RD does not need to make them accessible to clients. Clients SHOULD NOT attempt to dereference or manipulate them.

An RD can report registrations in lookup whose URI scheme and authority differ from the lookup resource's. Lookup clients MUST be prepared to see arbitrary URIs as registration resources in the results and treat them as opaque identifiers; the precise semantics of such links are left to future specifications.

The following example shows a client performing an endpoint lookup limited to endpoints of endpoint type tag:example.com,2020:platform:

Req: GET /rd-lookup/ep?et=tag:example.com,2020:platform

Res: 2.05 Content
Payload:
</rd/1234>;base="coap://[2001:db8:3::127]:61616";ep=node5;
    et="tag:example.com,2020:platform";ct=40;rt=core.rd-ep,
</rd/4521>;base="coap://[2001:db8:3::129]:61616";ep=node7;
    et="tag:example.com,2020:platform";ct=40;d=floor-3;
    rt=core.rd-ep
Figure 23: Examples of endpoint lookup

7. Security policies

The security policies that are applicable to an RD strongly depend on the application, and are not set out normatively here.

This section provides a list of aspects that applications should consider when describing their use of the RD, without claiming to cover all cases. It is using terminology of [I-D.ietf-ace-oauth-authz], in which the RD acts as the Resource Server (RS), and both registrant-eps and lookup clients act as Clients (C) with support from an Authorization Server (AS), without the intention of ruling out other (e.g. certificate / public-key infrastructure (PKI) based) schemes.

Any, all or none of the below can apply to an application. Which are relevant depends on its protection objectives.

Security policies are set by configuration of the RD, or by choice of the implementation. Lookup clients (and, where relevant, endpoints) can only trust an RD to uphold them if it is authenticated, and authorized to serve as an RD according to the application's requirements.

7.1. Endpoint name

Whenever an RD needs to provide trustworthy results to clients doing endpoint lookup, or resource lookup with filtering on the endpoint name, the RD must ensure that the registrant is authorized to use the given endpoint name. This applies both to registration and later to operations on the registration resource. It is immaterial whether the client is the registrant-ep itself or a CT is doing the registration: The RD cannot tell the difference, and CTs may use authorization credentials authorizing only operations on that particular endpoint name, or a wider range of endpoint names.

It is up to the concrete security policy to describe how endpoint name and sector are transported when certificates are used. For example, it may describe how SubjectAltName dNSName entries are mapped to endpoint and domain names.

7.1.1. Random endpoint names

Conversely, in applications where the RD does not check the endpoint name, the authorized registering endpoint can generate a random number (or string) that identifies the endpoint. The RD should then remember unique properties of the registrant, associate them with the registration for as long as its registration resource is active (which may be longer than the registration's lifetime), and require the same properties for operations on the registration resource.

Registrants that are prepared to pick a different identifier when their initial attempt (or attempts, in the unlikely case of two subsequent collisions) at registration is unauthorized should pick an identifier at least twice as long as the expected number of registrants; registrants without such a recovery options should pick significantly longer endpoint names (e.g. using UUID URNs [RFC4122]).

7.2. Entered resources

When lookup clients expect that certain types of links can only originate from certain endpoints, then the RD needs to apply filtering to the links an endpoint may register.

For example, if clients use an RD to find a server that provides firmware updates, then any registrant that wants to register (or update) links to firmware sources will need to provide suitable credentials to do so, independently of its endpoint name.

Note that the impact of having undesirable links in the RD depends on the application: if the client requires the firmware server to present credentials as a firmware server, a fraudulent link's impact is limited to the client revealing its intention to obtain updates and slowing down the client until it finds a legitimate firmware server; if the client accepts any credentials from the server as long as they fit the provided URI, the impact is larger.

An RD may also require that links are only registered if the registrant is authorized to publish information about the anchor (or even target) of the link. One way to do this is to demand that the registrant present the same credentials as a client that they'd need to present if contacted as a server at the resources' URI, which may include using the address and port that are part of the URI. Such a restriction places severe practical limitations on the links that can be registered.

As above, the impact of undesirable links depends on the extent to which the lookup client relies on the RD. To avoid the limitations, RD applications should consider prescribing that lookup clients only use the discovered information as hints, and describe which pieces of information need to be verified because they impact the application's security. A straightforward way to verify such information is to request it again from an authorized server, typically the one that hosts the target resource. That similar to what happens in Section 4.3 when the URI discovery step is repeated.

7.4. Segmentation

Within a single RD, different security policies can apply.

One example of this are multi-tenant deployments separated by the sector (d) parameter. Some sectors might apply limitations on the endpoint names available, while others use a random identifier approach to endpoint names and place limits on the entered links based on their attributes instead.

Care must be taken in such setups to determine the applicable access control measures to each operation. One easy way to do that is to mandate the use of the sector parameter on all operations, as no credentials are suitable for operations across sector borders anyway.

7.5. First-Come-First-Remembered: A default policy

The First-Come-First-Remembered policy is provided both as a reference example for a security policy definition, and as a policy that implementations may choose to use as default policy in absence of other configuration. It is designed to enable efficient discovery operations even in ad-hoc settings.

Under this policy, the RD accepts registrations for any endpoint name that is not assigned to an active registration resource, and only accepts registration updates from the same endpoint. The policy is minimal in that towards lookup clients it does not make any of the claims of Section 7.2 and Section 7.3, and its claims on Section 7.1 are limited to the lifetime of that endpoint's registration. It does, however, guarantee towards any endpoint that for the duration of its registration, its links will be discoverable on the RD.

When a registration or operation is attempted, the RD MUST determine the client's subject name or public key:

  • If the client's credentials indicate any subject name that is certified by any authority which the RD recognizes (which may be the system's trust anchor store), all such subject names are stored. With CWT or JWT based credentials (as common with ACE), the Subject (sub) claim is stored as a single name, if it exists. With X.509 certificates, the Common Name (CN) and the complete list of SubjectAltName entries are stored. In both cases, the authority that certified the claim is stored along with the subject, as the latter may only be locally unique.
  • Otherwise, if the client proves possession of a private key, the matching public key is stored. This applies both to raw public keys and to the public keys indicated in certificates that failed the above authority check.
  • If neither is present, a reference to the security session itself is stored. With (D)TLS, that is the connection itself, or the session resumption information if available. With OSCORE, that is the security context.

As part of the registration operation, that information is stored along with the registration resource.

The RD MUST accept all registrations whose registration resource is not already active, as long as they are made using a security layer supported by the RD.

Any operation on a registration resource, including registrations that lead to an existing registration resource, MUST be rejected by the RD unless all the stored information is found in the new request's credentials.

Note that even though subject names are compared in this policy, they are never directly compared to endpoint names, and an endpoint can not expect to "own" any particular endpoint name outside of an active registration -- even if a certificate says so. It is an accepted shortcoming of this approach that the endpoint has no indication of whether the RD remembers it by its subject name or public key; recognition by subject happens on a best-effort base (given the RD may not recognize any authority). Clients MUST be prepared to pick a different endpoint name when rejected by the RD initially or after a change in their credentials; picking an endpoint name as per Section 7.1.1 is an easy option for that.

For this policy to be usable without configuration, clients should not set a sector name in their registrations. An RD can set a default sector name for registrations accepted under this policy, which is useful especially in a segmented setup where different policies apply to different sectors. The configuration of such a behavior, as well as any other configuration applicable to such an RD (i.e. the set of recognized authorities) is out of scope for this document.

8. Security Considerations

The security considerations as described in Section 5 of [RFC8288] and Section 6 of [RFC6690] apply. The /.well-known/core resource may be protected e.g. using DTLS when hosted on a CoAP server as described in [RFC7252].

Access that is limited or affects sensitive data SHOULD be protected, e.g. using (D)TLS or OSCORE ([RFC8613]; which aspects of the RD this affects depends on the security policies of the application (see Section 7).

8.1. Discovery

Most steps in discovery of the RD, and possibly its resources, are not covered by CoAP's security mechanisms. This will not endanger the security properties of the registrations and lookup itself (where the client requires authorization of the RD if it expects any security properties of the operation), but may leak the client's intention to third parties, and allow them to slow down the process.

To mitigate that, clients can retain the RD's address, use secure discovery options like configured addresses, and send queries for RDs in a very general form (?rt=core.rd* rather than ?rt=core.rd-lookup-ep).

8.2. Endpoint Identification and Authentication

An Endpoint (name, sector) pair is unique within the set of endpoints registered by the RD. An Endpoint MUST NOT be identified by its protocol, port or IP address as these may change over the lifetime of an Endpoint.

Every operation performed by an Endpoint on an RD SHOULD be mutually authenticated using Pre-Shared Key, Raw Public Key or Certificate based security.

Consider the following threat: two devices A and B are registered at a single server. Both devices have unique, per-device credentials for use with DTLS to make sure that only parties with authorization to access A or B can do so.

Now, imagine that a malicious device A wants to sabotage the device B. It uses its credentials during the DTLS exchange. Then, it specifies the endpoint name of device B as the name of its own endpoint in device A. If the server does not check whether the identifier provided in the DTLS handshake matches the identifier used at the CoAP layer then it may be inclined to use the endpoint name for looking up what information to provision to the malicious device.

Endpoint authorization needs to be checked on registration and registration resource operations independently of whether there are configured requirements on the credentials for a given endpoint name (and sector; Section 7.1) or whether arbitrary names are accepted (Section 7.1.1).

Simple registration could be used to circumvent address-based access control: An attacker would send a simple registration request with the victim's address as source address, and later look up the victim's /.well-known/core content in the RD. Mitigation for this is recommended in Section 5.1.

The registration resource path is visible to any client that is allowed endpoint lookup, and can be extracted by resource lookup clients as well. The same goes for registration attributes that are shown as target attributes or lookup attributes. The RD needs to consider this in the choice of registration resource paths, and administrators or endpoint in their choice of attributes.

8.3. Access Control

Access control SHOULD be performed separately for the RD registration and Lookup API paths, as different endpoints may be authorized to register with an RD from those authorized to lookup endpoints from the RD. Such access control SHOULD be performed in as fine-grained a level as possible. For example access control for lookups could be performed either at the sector, endpoint or resource level.

The precise access controls necessary (and the consequences of failure to enforce them) depend on the protection objectives of the application and the security policies (Section 7) derived from them.

8.4. Denial of Service Attacks

Services that run over UDP unprotected are vulnerable to unknowingly amplify and distribute a DoS attack as UDP does not require return routability check. Since RD lookup responses can be significantly larger than requests, RDs are prone to this.

[RFC7252] describes this at length in its Section 11.3, including some mitigation by using small block sizes in responses. The upcoming [I-D.ietf-core-echo-request-tag] updates that by describing a source address verification mechanism using the Echo option.

[ If this document is published together with or after I-D.ietf-core-echo-request-tag, the above paragraph is replaced with the following:

[RFC7252] describes this at length in its Section 11.3, and [I-D.ietf-core-echo-request-tag] (which updates the former) recommends using the Echo option to verify the request's source address.

]

8.5. Skipping freshness checks

When RD based applications are built in which request freshness checks are not performed, these concerns need to be balanced:

  • When alterations to registration attributes are reordered, an attacker may create any combination of attributes ever set, with the attack difficulty determined by the security layer's replay properties.

    For example, if Figure 18 were conducted without freshness assurances, an attacker could later reset the lifetime back to 7200. Thus, the device is made unreachable to lookup clients.

  • When registration updates without query parameters (which just serve to restart the lifetime) can be reordered, an attacker can use intercepted messages to give the appearance of the device being alive to the RD.

    This is unacceptable when when the RD's security policy promises reachability of endpoints (e.g. when disappearing devices would trigger further investigation), but may be acceptable with other policies.

9. IANA Considerations

9.1. Resource Types

IANA is asked to enter the following values into the Resource Type (rt=) Link Target Attribute Values sub-registry of the Constrained Restful Environments (CoRE) Parameters registry defined in [RFC6690]:

Table 2
Value Description Reference
core.rd Directory resource of an RD RFCTHIS Section 4.3
core.rd-lookup-res Resource lookup of an RD RFCTHIS Section 4.3
core.rd-lookup-ep Endpoint lookup of an RD RFCTHIS Section 4.3
core.rd-ep Endpoint resource of an RD RFCTHIS Section 6

9.2. IPv6 ND Resource Directory Address Option

This document registers one new ND option type under the sub-registry "IPv6 Neighbor Discovery Option Formats" of the "Internet Control Message Protocol version 6 (ICMPv6) Parameters" registry:

  • Resource Directory Address Option (TBD38)

[ The RFC editor is asked to replace TBD38 with the assigned number in the document; the value 38 is suggested. ]

9.3. RD Parameter Registry

This specification defines a new sub-registry for registration and lookup parameters called "RD Parameters" under "CoRE Parameters". Although this specification defines a basic set of parameters, it is expected that other standards that make use of this interface will define new ones.

Each entry in the registry must include

  • the human readable name of the parameter,
  • the short name as used in query parameters or target attributes,
  • indication of whether it can be passed as a query parameter at registration of endpoints, as a query parameter in lookups, or be expressed as a target attribute,
  • syntax and validity requirements if any,
  • a description,
  • and a link to reference documentation.

The query parameter MUST be both a valid URI query key [RFC3986] and a token as used in [RFC8288].

The description must give details on whether the parameter can be updated, and how it is to be processed in lookups.

The mechanisms around new RD parameters should be designed in such a way that they tolerate RD implementations that are unaware of the parameter and expose any parameter passed at registration or updates on in endpoint lookups. (For example, if a parameter used at registration were to be confidential, the registering endpoint should be instructed to only set that parameter if the RD advertises support for keeping it confidential at the discovery step.)

Initial entries in this sub-registry are as follows:

Table 3: RD Parameters
Full name Short Validity Use Description
Endpoint Name ep Unicode* RLA Name of the endpoint
Lifetime lt 1-4294967295 R Lifetime of the registration in seconds
Sector d Unicode* RLA Sector to which this endpoint belongs
Registration Base URI base URI RLA The scheme, address and port and path at which this server is available
Page page Integer L Used for pagination
Count count Integer L Used for pagination
Endpoint Type et Section 9.3.1 RLA Semantic type of the endpoint (see Section 9.4)

(Short: Short name used in query parameters or target attributes. Validity: Unicode* = 63 Bytes of UTF-8 encoded Unicode, with no control characters as per Section 5. Use: R = used at registration, L = used at lookup, A = expressed in target attribute.)

The descriptions for the options defined in this document are only summarized here. To which registrations they apply and when they are to be shown is described in the respective sections of this document. All their reference documentation entries point to this document.

The IANA policy for future additions to the sub-registry is "Expert Review" as described in [RFC8126]. The evaluation should consider formal criteria, duplication of functionality (Is the new entry redundant with an existing one?), topical suitability (E.g. is the described property actually a property of the endpoint and not a property of a particular resource, in which case it should go into the payload of the registration and need not be registered?), and the potential for conflict with commonly used target attributes (For example, if could be used as a parameter for conditional registration if it were not to be used in lookup or attributes, but would make a bad parameter for lookup, because a resource lookup with an if query parameter could ambiguously filter by the registered endpoint property or the [RFC6690] target attribute).

9.3.1. Full description of the "Endpoint Type" RD Parameter

An endpoint registering at an RD can describe itself with endpoint types, similar to how resources are described with Resource Types in [RFC6690]. An endpoint type is expressed as a string, which can be either a URI or one of the values defined in the Endpoint Type sub-registry. Endpoint types can be passed in the et query parameter as part of extra-attrs at the Registration step, are shown on endpoint lookups using the et target attribute, and can be filtered for using et as a search criterion in resource and endpoint lookup. Multiple endpoint types are given as separate query parameters or link attributes.

Note that Endpoint Type differs from Resource Type in that it uses multiple attributes rather than space separated values. As a result, RDs implementing this specification automatically support correct filtering in the lookup interfaces from the rules for unknown endpoint attributes.

9.4. "Endpoint Type" (et=) RD Parameter values

This specification establishes a new sub-registry under "CoRE Parameters" called '"Endpoint Type" (et=) RD Parameter values'. The registry properties (required policy, requirements, template) are identical to those of the Resource Type parameters in [RFC6690], in short:

The review policy is IETF Review for values starting with "core", and Specification Required for others.

The requirements to be enforced are:

  • The values MUST be related to the purpose described in Section 9.3.1.
  • The registered values MUST conform to the ABNF reg-rel-type definition of [RFC6690] and MUST NOT be a URI.
  • It is recommended to use the period "." character for segmentation.

The registry initially contains one value:

  • "core.rd-group": An application group as described in Appendix A.

9.5. Multicast Address Registration

IANA is asked to assign the following multicast addresses for use by CoAP nodes:

IPv4 -- "all CoRE Resource Directories" address MCD2 (suggestion: 224.0.1.189), from the "IPv4 Multicast Address Space Registry". As the address is used for discovery that may span beyond a single network, it has come from the Internetwork Control Block (224.0.1.x) [RFC5771].

IPv6 -- "all CoRE Resource Directories" address MCD1 (suggestions FF0X::FE), from the "IPv6 Multicast Address Space Registry", in the "Variable Scope Multicast Addresses" space (RFC 3307). Note that there is a distinct multicast address for each scope that interested CoAP nodes should listen to; CoAP needs the Link-Local and Site-Local scopes only.

[ The RFC editor is asked to replace MCD1 and MCD2 with the assigned addresses throughout the document. ]

9.6. Well-Known URIs

IANA is asked to permanently register the URI suffix "rd" in the "Well-Known URIs" registry. The change controller is the IETF, this document is the reference.

9.7. Service Names and Transport Protocol Port Number Registry

IANA is asked to enter four new items into the Service Names and Transport Protocol Port Number Registry:

  • Service name: "core-rd", Protocol: "udp", Description: "Resource Directory accessed using CoAP"
  • Service name "core-rd-dtls", Protocol: "udp", Description: "Resource Directory accessed using CoAP over DTLS"
  • Service name: "core-rd", Protocol: "tcp", Description: "Resource Directory accessed using CoAP over TCP"
  • Service name "core-rd-tls", Protocol: "tcp", Description: "Resource Directory accessed using CoAP over TLS"

All in common have this document as their reference.

10. Examples

Two examples are presented: a Lighting Installation example in Section 10.1 and a LwM2M example in Section 10.2.

10.1. Lighting Installation

This example shows a simplified lighting installation which makes use of the RD with a CoAP interface to facilitate the installation and start-up of the application code in the lights and sensors. In particular, the example leads to the definition of a group and the enabling of the corresponding multicast address as described in Appendix A. No conclusions must be drawn on the realization of actual installation or naming procedures, because the example only "emphasizes" some of the issues that may influence the use of the RD and does not pretend to be normative.

10.1.1. Installation Characteristics

The example assumes that the installation is managed. That means that a Commissioning Tool (CT) is used to authorize the addition of nodes, name them, and name their services. The CT can be connected to the installation in many ways: the CT can be part of the installation network, connected by WiFi to the installation network, or connected via GPRS link, or other method.

It is assumed that there are two naming authorities for the installation: (1) the network manager that is responsible for the correct operation of the network and the connected interfaces, and (2) the lighting manager that is responsible for the correct functioning of networked lights and sensors. The result is the existence of two naming schemes coming from the two managing entities.

The example installation consists of one presence sensor, and two luminaries, luminary1 and luminary2, each with their own wireless interface. Each luminary contains three lamps: left, right and middle. Each luminary is accessible through one endpoint. For each lamp a resource exists to modify the settings of a lamp in a luminary. The purpose of the installation is that the presence sensor notifies the presence of persons to a group of lamps. The group of lamps consists of: middle and left lamps of luminary1 and right lamp of luminary2.

Before commissioning by the lighting manager, the network is installed and access to the interfaces is proven to work by the network manager.

At the moment of installation, the network under installation is not necessarily connected to the DNS infrastructure. Therefore, SLAAC IPv6 addresses are assigned to CT, RD, luminaries and the sensor. The addresses shown in Table 4 below stand in for these in the following examples.

Table 4: Addresses used in the examples
Name IPv6 address
luminary1 2001:db8:4::1
luminary2 2001:db8:4::2
Presence sensor 2001:db8:4::3
RD 2001:db8:4::ff

In Section 10.1.2 the use of RD during installation is presented.

10.1.2. RD entries

It is assumed that access to the DNS infrastructure is not always possible during installation. Therefore, the SLAAC addresses are used in this section.

For discovery, the resource types (rt) of the devices are important. The lamps in the luminaries have rt=tag:example.com,2020:light, and the presence sensor has rt=tag:example.com,2020:p-sensor. The endpoints have names which are relevant to the light installation manager. In this case luminary1, luminary2, and the presence sensor are located in room 2-4-015, where luminary1 is located at the window and luminary2 and the presence sensor are located at the door. The endpoint names reflect this physical location. The middle, left and right lamps are accessed via path /light/middle, /light/left, and /light/right respectively. The identifiers relevant to the RD are shown in Table 5 below:

Table 5: RD identifiers
Name endpoint resource path resource type
luminary1 lm_R2-4-015_wndw /light/left tag:example.com,2020:light
luminary1 lm_R2-4-015_wndw /light/middle tag:example.com,2020:light
luminary1 lm_R2-4-015_wndw /light/right tag:example.com,2020:light
luminary2 lm_R2-4-015_door /light/left tag:example.com,2020:light
luminary2 lm_R2-4-015_door /light/middle tag:example.com,2020:light
luminary2 lm_R2-4-015_door /light/right tag:example.com,2020:light
Presence sensor ps_R2-4-015_door /ps tag:example.com,2020:p-sensor

It is assumed that the CT has performed RD discovery and has received a response like the one in the Section 4.3 example.

The CT inserts the endpoints of the luminaries and the sensor in the RD using the registration base URI parameter (base) to specify the interface address:

Req: POST coap://[2001:db8:4::ff]/rd
  ?ep=lm_R2-4-015_wndw&base=coap://[2001:db8:4::1]&d=R2-4-015
Payload:
</light/left>;rt="tag:example.com,2020:light",
</light/middle>;rt="tag:example.com,2020:light",
</light/right>;rt="tag:example.com,2020:light"

Res: 2.01 Created
Location-Path: /rd/4521

Req: POST coap://[2001:db8:4::ff]/rd
  ?ep=lm_R2-4-015_door&base=coap://[2001:db8:4::2]&d=R2-4-015
Payload:
</light/left>;rt="tag:example.com,2020:light",
</light/middle>;rt="tag:example.com,2020:light",
</light/right>;rt="tag:example.com,2020:light"

Res: 2.01 Created
Location-Path: /rd/4522

Req: POST coap://[2001:db8:4::ff]/rd
  ?ep=ps_R2-4-015_door&base=coap://[2001:db8:4::3]&d=R2-4-015
Payload:
</ps>;rt="tag:example.com,2020:p-sensor"

Res: 2.01 Created
Location-Path: /rd/4523
Figure 24: Example of registrations a CT enters into an RD

The sector name d=R2-4-015 has been added for an efficient lookup because filtering on "ep" name is more awkward. The same sector name is communicated to the two luminaries and the presence sensor by the CT.

The group is specified in the RD. The base parameter is set to the site-local multicast address allocated to the group. In the POST in the example below, the resources supported by all group members are published.

Req: POST coap://[2001:db8:4::ff]/rd
  ?ep=grp_R2-4-015&et=core.rd-group&base=coap://[ff05::1]
Payload:
</light/left>;rt="tag:example.com,2020:light",
</light/middle>;rt="tag:example.com,2020:light",
</light/right>;rt="tag:example.com,2020:light"

Res: 2.01 Created
Location-Path: /rd/501
Figure 25: Example of a multicast group a CT enters into an RD

After the filling of the RD by the CT, the application in the luminaries can learn to which groups they belong, and enable their interface for the multicast address.

The luminary, knowing its sector and being configured to join any group containing lights, searches for candidate groups and joins them:

Req: GET coap://[2001:db8:4::ff]/rd-lookup/ep
  ?d=R2-4-015&et=core.rd-group&rt=light

Res: 2.05 Content
Payload:
</rd/501>;ep=grp_R2-4-015;et=core.rd-group;
          base="coap://[ff05::1]";rt=core.rd-ep
Figure 26: Example of a lookup exchange to find suitable multicast addresses

From the returned base parameter value, the luminary learns the multicast address of the multicast group.

The presence sensor can learn the presence of groups that support resources with rt=tag:example.com,2020:light in its own sector by sending the same request, as used by the luminary. The presence sensor learns the multicast address to use for sending messages to the luminaries.

10.2. OMA Lightweight M2M (LwM2M)

OMA LwM2M is a profile for device services based on CoAP, providing interfaces and operations for device management and device service enablement.

An LwM2M server is an instance of an LwM2M middleware service layer, containing an RD ([LwM2M] page 36f).

That RD only implements the registration interface, and no lookup is implemented. Instead, the LwM2M server provides access to the registered resources, in a similar way to a reverse proxy.

The location of the LwM2M Server and RD URI path is provided by the LwM2M Bootstrap process, so no dynamic discovery of the RD is used. LwM2M Servers and endpoints are not required to implement the /.well-known/core resource.

11. Acknowledgments

Oscar Novo, Srdjan Krco, Szymon Sasin, Kerry Lynn, Esko Dijk, Anders Brandt, Matthieu Vial, Jim Schaad, Mohit Sethi, Hauke Petersen, Hannes Tschofenig, Sampo Ukkola, Linyi Tian, Jan Newmarch, Matthias Kovatsch, Jaime Jimenez and Ted Lemon have provided helpful comments, discussions and ideas to improve and shape this document. Zach would also like to thank his colleagues from the EU FP7 SENSEI project, where many of the RD concepts were originally developed.

12. Changelog

changes from -27 to -28

changes from -26 to -27

changes from -25 to -26

changes from -24 to -25

changes from -23 to -24

changes from -22 to -23

changes from -21 to -22

changes from -20 to -21

(Processing comments during WGLC)

changes from -19 to -20

(Processing comments from the WG chair review)

changes from -18 to -19

changes from -17 to -18

changes from -16 to -17

(Note that -17 is published as a direct follow-up to -16, containing a single change to be discussed at IETF103)

changes from -15 to -16

changes from -14 to -15

changes from -13 to -14

changes from -12 to -13

changes from -11 to -12

changes from -09 to -10

changes from -08 to -09

changes from -07 to -08

changes from -06 to -07

Changes from -05 to -06

changes from -04 to -05

Changes from -03 to -04:

Changes from -02 to -03:

Changes from -01 to -02:

Changes from -00 to -01:

Changes from -05 to WG Document -00:

Changes from -04 to -05:

Changes from -03 to -04:

Changes from -02 to -03:

Changes from -01 to -02:

13. References

13.1. Normative References

[I-D.ietf-core-echo-request-tag]
Amsüss, C., Mattsson, J. P., and G. Selander, "CoAP: Echo, Request-Tag, and Token Processing", Work in Progress, Internet-Draft, draft-ietf-core-echo-request-tag-12, , <https://www.ietf.org/archive/id/draft-ietf-core-echo-request-tag-12.txt>.
[RFC2119]
Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, , <https://www.rfc-editor.org/info/rfc2119>.
[RFC3986]
Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform Resource Identifier (URI): Generic Syntax", STD 66, RFC 3986, DOI 10.17487/RFC3986, , <https://www.rfc-editor.org/info/rfc3986>.
[RFC6570]
Gregorio, J., Fielding, R., Hadley, M., Nottingham, M., and D. Orchard, "URI Template", RFC 6570, DOI 10.17487/RFC6570, , <https://www.rfc-editor.org/info/rfc6570>.
[RFC6690]
Shelby, Z., "Constrained RESTful Environments (CoRE) Link Format", RFC 6690, DOI 10.17487/RFC6690, , <https://www.rfc-editor.org/info/rfc6690>.
[RFC6763]
Cheshire, S. and M. Krochmal, "DNS-Based Service Discovery", RFC 6763, DOI 10.17487/RFC6763, , <https://www.rfc-editor.org/info/rfc6763>.
[RFC7230]
Fielding, R., Ed. and J. Reschke, Ed., "Hypertext Transfer Protocol (HTTP/1.1): Message Syntax and Routing", RFC 7230, DOI 10.17487/RFC7230, , <https://www.rfc-editor.org/info/rfc7230>.
[RFC7252]
Shelby, Z., Hartke, K., and C. Bormann, "The Constrained Application Protocol (CoAP)", RFC 7252, DOI 10.17487/RFC7252, , <https://www.rfc-editor.org/info/rfc7252>.
[RFC8126]
Cotton, M., Leiba, B., and T. Narten, "Guidelines for Writing an IANA Considerations Section in RFCs", BCP 26, RFC 8126, DOI 10.17487/RFC8126, , <https://www.rfc-editor.org/info/rfc8126>.
[RFC8174]
Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, , <https://www.rfc-editor.org/info/rfc8174>.
[RFC8288]
Nottingham, M., "Web Linking", RFC 8288, DOI 10.17487/RFC8288, , <https://www.rfc-editor.org/info/rfc8288>.

13.2. Informative References

[ER]
Chen, P., "The entity-relationship model—toward a unified view of data", ACM Transactions on Database Systems Vol. 1, pp. 9-36, DOI 10.1145/320434.320440, , <https://doi.org/10.1145/320434.320440>.
[I-D.bormann-t2trg-rel-impl]
Bormann, C., "impl-info: A link relation type for disclosing implementation information", Work in Progress, Internet-Draft, draft-bormann-t2trg-rel-impl-02, , <https://www.ietf.org/archive/id/draft-bormann-t2trg-rel-impl-02.txt>.
[I-D.hartke-t2trg-coral]
Hartke, K., "The Constrained RESTful Application Language (CoRAL)", Work in Progress, Internet-Draft, draft-hartke-t2trg-coral-09, , <https://www.ietf.org/archive/id/draft-hartke-t2trg-coral-09.txt>.
[I-D.ietf-ace-oauth-authz]
Seitz, L., Selander, G., Wahlstroem, E., Erdtman, S., and H. Tschofenig, "Authentication and Authorization for Constrained Environments (ACE) using the OAuth 2.0 Framework (ACE-OAuth)", Work in Progress, Internet-Draft, draft-ietf-ace-oauth-authz-37, , <https://www.ietf.org/archive/id/draft-ietf-ace-oauth-authz-37.txt>.
LI, K., Rahman, A., and C. Bormann, "Representing Constrained RESTful Environments (CoRE) Link Format in JSON and CBOR", Work in Progress, Internet-Draft, draft-ietf-core-links-json-10, , <https://www.ietf.org/archive/id/draft-ietf-core-links-json-10.txt>.
[I-D.ietf-core-rd-dns-sd]
Stok, P. V. D., Koster, M., and C. Amsüss, "CoRE Resource Directory: DNS-SD mapping", Work in Progress, Internet-Draft, draft-ietf-core-rd-dns-sd-05, , <https://www.ietf.org/archive/id/draft-ietf-core-rd-dns-sd-05.txt>.
[I-D.silverajan-core-coap-protocol-negotiation]
Silverajan, B. and M. Ocak, "CoAP Protocol Negotiation", Work in Progress, Internet-Draft, draft-silverajan-core-coap-protocol-negotiation-09, , <https://www.ietf.org/archive/id/draft-silverajan-core-coap-protocol-negotiation-09.txt>.
[LwM2M]
Open Mobile Alliance, "Lightweight Machine to Machine Technical Specification: Transport Bindings (Candidate Version 1.1)", , <https://openmobilealliance.org/RELEASE/LightweightM2M/V1_1-20180612-C/OMA-TS-LightweightM2M_Transport-V1_1-20180612-C.pdf>.
[RFC3306]
Haberman, B. and D. Thaler, "Unicast-Prefix-based IPv6 Multicast Addresses", RFC 3306, DOI 10.17487/RFC3306, , <https://www.rfc-editor.org/info/rfc3306>.
[RFC3849]
Huston, G., Lord, A., and P. Smith, "IPv6 Address Prefix Reserved for Documentation", RFC 3849, DOI 10.17487/RFC3849, , <https://www.rfc-editor.org/info/rfc3849>.
[RFC4122]
Leach, P., Mealling, M., and R. Salz, "A Universally Unique IDentifier (UUID) URN Namespace", RFC 4122, DOI 10.17487/RFC4122, , <https://www.rfc-editor.org/info/rfc4122>.
[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, , <https://www.rfc-editor.org/info/rfc4944>.
[RFC5771]
Cotton, M., Vegoda, L., and D. Meyer, "IANA Guidelines for IPv4 Multicast Address Assignments", BCP 51, RFC 5771, DOI 10.17487/RFC5771, , <https://www.rfc-editor.org/info/rfc5771>.
[RFC6724]
Thaler, D., Ed., Draves, R., Matsumoto, A., and T. Chown, "Default Address Selection for Internet Protocol Version 6 (IPv6)", RFC 6724, DOI 10.17487/RFC6724, , <https://www.rfc-editor.org/info/rfc6724>.
[RFC6775]
Shelby, Z., Ed., Chakrabarti, S., Nordmark, E., and C. Bormann, "Neighbor Discovery Optimization for IPv6 over Low-Power Wireless Personal Area Networks (6LoWPANs)", RFC 6775, DOI 10.17487/RFC6775, , <https://www.rfc-editor.org/info/rfc6775>.
[RFC6874]
Carpenter, B., Cheshire, S., and R. Hinden, "Representing IPv6 Zone Identifiers in Address Literals and Uniform Resource Identifiers", RFC 6874, DOI 10.17487/RFC6874, , <https://www.rfc-editor.org/info/rfc6874>.
[RFC7228]
Bormann, C., Ersue, M., and A. Keranen, "Terminology for Constrained-Node Networks", RFC 7228, DOI 10.17487/RFC7228, , <https://www.rfc-editor.org/info/rfc7228>.
[RFC7641]
Hartke, K., "Observing Resources in the Constrained Application Protocol (CoAP)", RFC 7641, DOI 10.17487/RFC7641, , <https://www.rfc-editor.org/info/rfc7641>.
[RFC8106]
Jeong, J., Park, S., Beloeil, L., and S. Madanapalli, "IPv6 Router Advertisement Options for DNS Configuration", RFC 8106, DOI 10.17487/RFC8106, , <https://www.rfc-editor.org/info/rfc8106>.
[RFC8132]
van der Stok, P., Bormann, C., and A. Sehgal, "PATCH and FETCH Methods for the Constrained Application Protocol (CoAP)", RFC 8132, DOI 10.17487/RFC8132, , <https://www.rfc-editor.org/info/rfc8132>.
[RFC8141]
Saint-Andre, P. and J. Klensin, "Uniform Resource Names (URNs)", RFC 8141, DOI 10.17487/RFC8141, , <https://www.rfc-editor.org/info/rfc8141>.
[RFC8613]
Selander, G., Mattsson, J., Palombini, F., and L. Seitz, "Object Security for Constrained RESTful Environments (OSCORE)", RFC 8613, DOI 10.17487/RFC8613, , <https://www.rfc-editor.org/info/rfc8613>.

Appendix A. Groups Registration and Lookup

The RD-Groups usage pattern allows announcing application groups inside an RD.

Groups are represented by endpoint registrations. Their base address is a multicast address, and they SHOULD be entered with the endpoint type core.rd-group. The endpoint name can also be referred to as a group name in this context.

The registration is inserted into the RD by a Commissioning Tool, which might also be known as a group manager here. It performs third party registration and registration updates.

The links it registers SHOULD be available on all members that join the group. Depending on the application, members that lack some resource MAY be permissible if requests to them fail gracefully.

The following example shows a CT registering a group with the name "lights" which provides two resources. The directory resource path /rd is an example RD location discovered in a request similar to Figure 5. The group address in the example is constructed from [RFC3849]'s reserved 2001:db8:: prefix as a unicast-prefix based site-local address (see [RFC3306].

Req: POST coap://rd.example.com/rd?ep=lights&et=core.rd-group
                           &base=coap://[ff35:30:2001:db8:f1::8000:1]
Content-Format: 40
Payload:
</light>;rt="tag:example.com,2020:light";
     if="tag:example.net,2020:actuator",
</color-temperature>;if="tag:example.net,2020:parameter";u=K

Res: 2.01 Created
Location-Path: /rd/12
Figure 27: Example registration of a group

In this example, the group manager can easily permit devices that have no writable color-temperature to join, as they would still respond to brightness changing commands. Had the group instead contained a single resource that sets brightness and color temperature atomically, endpoints would need to support both properties.

The resources of a group can be looked up like any other resource, and the group registrations (along with any additional registration parameters) can be looked up using the endpoint lookup interface.

The following example shows a client performing an endpoint lookup for all groups.

Req: GET /rd-lookup/ep?et=core.rd-group

Res: 2.05 Content
Payload:
</rd/12>;ep=lights&et=core.rd-group;
         base="coap://[ff35:30:2001:f1:db8::8000:1]";rt=core.rd-ep
Figure 28: Example lookup of groups

The following example shows a client performing a lookup of all resources of all endpoints (groups) with et=core.rd-group.

Req: GET /rd-lookup/res?et=core.rd-group

Res: 2.05 Content
Payload:
<coap://[ff35:30:2001:db8:f1::8000:1]/light>;
     rt="tag:example.com,2020:light";
     if="tag:example.net,2020:actuator",
<coap://[ff35:30:2001:db8:f1::8000:1]/color-temperature>;
     if="tag:example.net,2020:parameter";u=K,
Figure 29: Example lookup of resources inside groups

The CoRE Link Format as described in [RFC6690] has been interpreted differently by implementers, and a strict implementation rules out some use cases of an RD (e.g. base values with path components in combination with absent anchors).

This appendix describes a subset of link format documents called Limited Link Format. The one rule herein is not very limiting in practice -- all examples in RFC6690, and all deployments the authors are aware of already stick to them -- but ease the implementation of RD servers.

It is applicable to representations in the application/link-format media type, and any other media types that inherit [RFC6690] Section 2.1.

A link format representation is in Limited Link format if, for each link in it, the following applies:

All URI references either follow the URI or the path-absolute ABNF rule of RFC3986 (i.e. target and anchor each either start with a scheme or with a single slash).

Authors' Addresses

Christian Amsüss (editor)
Hollandstr. 12/4
1020
Austria
Zach Shelby
ARM
150 Rose Orchard
San Jose, 95134
United States of America
Michael Koster
SmartThings
665 Clyde Avenue
Mountain View, 94043
United States of America
Carsten Bormann
Universitaet Bremen TZI
Postfach 330440
D-28359 Bremen
Germany
Peter van der Stok
consultant