The Constrained RESTful Application Language (CoRAL)

1.1. Data and Interaction Model

The data model is similar to the Resource Description Framework (RDF) [W3C.REC-rdf11-concepts-20140225] model, with provisions to enable form based interaction and to express data from Web Linking ([RFC8288]) based models such as [RFC6690]'s Link Format.¶

The interaction model derives from the processing model of HTML [W3C.REC-html52-20171214] and specifies how an automated software agent can change the application state by navigating between resources following links and performing operations on resources submitting forms.¶

1.2. Notational Conventions

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.¶

Terms defined in this document appear in cursive where they are introduced (rendered in plain text as the new term surrounded by underscores).¶

2.1. Browsing Context

Borrowing from HTML 5 [W3C.REC-html52-20171214], each such agent maintains a browsing context in which the representations of Web resources are processed. (In HTML, the browsing context typically corresponds to a tab or window in a Web browser.)¶

At any time, one representation in a browsing context is designated the active representation.¶

2.2. Documents

A resource representation in one of the CoRAL serialization formats is called a CoRAL document. The URI that was used to retrieve such a document is called the document's retrieval context. This URI is also considered the base URI for relative URI references in the document.¶

A CoRAL document consists of a list of zero or more statements that can express links or (in a composition of statements) forms. CoRAL serialization formats may contain additional elements for efficiency or convenience, such as an embedded base URI that takes precedence over the document's base URI, or to concisely represent compound statements (e.g., to express forms).¶

2.3. Data model

The basic CoRAL information model is similar to the Resource Description Framework (RDF) [W3C.REC-rdf11-concepts-20140225] information model: Data is expressed as an (unordered) set of triples (also called statements), consisting of a subject, a predicate and an object. The predicate is always a URI, the subject is a URI or a blank node, and the object is either a URI, a blank node or a litreal. All URIs here are limited to the syntax-based normalized form of [RFC3986] Section 6.2.2.¶

Blank nodes are unnamed entities. Literals are CBOR objects.¶

These triples form a directed multigraph with the subject and object being source and destination, and the predicate a description on the edge. That graph is equivalent to the data.¶

To form a set and a graph, we define an equivalence relation: URIs are only equal to URIs and if they are identical byte-wise. A blank node is only equal to itself. A literal is equal to a different literal if its value is equal to the other literal's value in the CBOR generic data model.¶

Triples are equivalent to each other if their subject, predicate and object are pair-wise equivalent.¶

The CoRAL structured information model is a sequence of "passings" of the basic model's edges, starting at a node identifying the document (its retrieval context, typically URI from which it was obtained) where¶

• each edge is passed at least one time in total,¶

• each edge is passed at most one time after each passing that ends in its start point (with the obvious exception that edges from the retrieval context can be passed once from the start), and¶

• between a passing of an edge from A to B and a later passing from B to C, passings can only be along edges that can be reached from B along the graph, until B is the end of a different passing.¶

For better understanding, think of the structured information model as a sort of tree spanning from the retrieval context, with the oddity that when a node is reached along two different edges (which a normal tree doesn't do), it is up to the builder of the tree whether to describe anything children of the entered node on one parent or on the other parent, on both, or to describe some children at the first and others at a later occasion.¶

Exceeding the RDF-like model, this represents CoRAL's focus on the discovery of possivble future application states over the description of a graph of resources.¶

</sensors>;ct=40;title="Sensor Index",
</sensors/temp>;rt="temperature-c";if="sensor",
</sensors/light>;rt="light-lux";if="sensor",
<http://www.example.com/sensors/t123>;anchor="/sensors/temp";rel="describedby",
</t>;anchor="/sensors/temp";rel="alternate"

[
  [2, simple(10) / item 10 for rel:hosts /, cri"/sensors", [
    [2, 6(2) / item 20 for linkformat:ct /, 40],
    [2, simple(15) / item 15 for linkformat:title /, "Sensor Index"]
  ]],
  [2, simple(10) / item 10 for rel:hosts /, cri"/sensors/temp", [
    [2, 6(1) / item 18 for linkformat:if /, 6(200) / cri"http:∕∕TBD∕...∕temperature-c" /],
    [2, 6(-2) / item 19 for linkformat:rt /, 6(250) / cri"http:∕∕TBD∕...∕sensor" /],
    [2, simple(12) / item 12 for rel:describedby /, cri"http://www.example.com/sensors/t123"],
    [2, simple(11) / item 11 for rel:alternate /, cri"/t"]
  ]],
  [2, 10 / item10 for rel:hosts /, cri"/sensors/light", [
    [2, 6(1) / item 18 for linkformat:if /, 6(-201)],
    [2, 6(-2) / item 19 for linkformat:rt /, 6(250)]
  ]]
]

 Link: </TheBook/chapter2>;
         rel="previous"; title*=UTF-8'de'letztes%20Kapitel,

[
  [2, 6(...) / rel:previous /, cri"/TheBook/chapter2", [
    [2, simple(15) / item 15 for linkformat:title /, 38(["de", "letztes Kapitel"])]
  ]]
]

2.4. Serialization Format

The primary serialization format is a compact, binary encoding of links and forms in Concise Binary Object Representation (CBOR) [RFC8949]. This format is intended for environments with constraints on power, memory, and processing resources [RFC7228] and shares many similarities with the message format of CoAP: In place of verbose strings, small numeric identifiers are used to encode link relation types and operation types. Uniform Resource Identifiers (URIs) [RFC3986] are expressed as Constrained Resource Identifier (CRI) references [I-D.ietf-core-href] and thus pre-parsed for easy use with CoAP. As a result, link serializations in CoRAL are often much more compact and easier to process than equivalent serializations in CoRE Link Format [RFC6690].¶

For easy representation of CoRAL documents in text, CBOR diagnostic notation is used. Along with indentation and comments, the notation introduced in [I-D.bormann-cbor-edn-literals] is used to represent CRIs. This format is not expected to be sent over the network.¶

[ To be discussed: For even better readability, the RDF Turtle [W3C.REC-turtle-20140225] format can be used when only the basic information model content is to be conveyed. When used like this, the conversion according to the RDF appendix is implied. ]¶

2.5. Links

Any statement "links" a resource with a second resource or literal, and is thus also referred to as a link.¶

In [RFC8288] terminology, a CoRAL link's subject is the link context, the predicate is the link relation type, and the object is the link target.¶

However, a link in CoRAL does not have target attributes. Instead, a link may have a list of zero or more nested elements. These enable both the description of resource metadata and the chaining of links, which is done in [RFC8288] by setting the anchor of one link to the target of another.¶

• A link can be viewed as a statement of the form "{link context} has a {link relation type} resource at {link target}" where the link target may be further described by nested elements.¶

A link relation type identifies the semantics of a link. In HTML and in [RFC8288], link relation types are typically denoted by an IANA-registered name, such as stylesheet or type. In CoRAL, all link relation types are, in contrast, denoted by a Universal Resource Identifier (URI) [RFC3986], such as <http://www.iana.org/assignments/relation/stylesheet> or <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>. This allows for the decentralized creation of new link relation types without the risk of collisions when they come from different organizations or domains of knowledge. URIs can also lead to documentation, schema, and other information about a link relation type. In CoRAL documents, these URIs are only used as identity tokens, though, and are compared with Simple String Comparison as specified in Section 6.2.1 of [RFC3986].¶

If the link target is a URI and the URI scheme indicates a Web transfer protocol like HTTP or CoAP, an agent can dereference the URI and navigate the browsing context to its target resource; this is called following the link. An anonymous resource is a resource that is identified by neither a URI nor a literal representation. The agent can still follow the link, but can not dereference it and is limited in its next steps by the outgoing links that are expressed in the current document.¶

A link can occur as a top-level element in a document or as a nested element within a link. When a link occurs as a top-level element, the link context implicitly is the document's retrieval context. When a link occurs nested within a link, the link context of the nested link is the link target of the enclosing link.¶

There are no restrictions on the cardinality of links; there can be multiple links to and from a particular target, and multiple links of the same or different types between a given link context and target. However, the nesting nature of the data model constrains the description of resource relations to a tree: Relations between linked resources can only be described by further nesting links.¶

2.6. Forms

A form provides instructions to an agent for performing an operation on a resource on the Web. A form has a form context, an operation type, a request method, and a submission target. Additionally, a form may be accompanied by a list of zero or more form fields.¶

In the basic information model, the form is identified with an anonymous node. The form context and operation type are the subject and predicate of an incoming link, respectively; request method and submission target of an outgoing link. Form fields are additional links from that form.¶

• A form can be viewed as an instruction of the form "To perform an {operation type} operation on {form context}, make a {request method} request to {submission target}" where the request may be further described by form fields.¶

An operation type identifies the semantics of the operation. Operation types are denoted (like link relation types) by a URI.¶

Form contexts and submission targets are both denoted by a URI. The form context is the resource on which the operation is ultimately performed. To perform the operation, an agent needs to construct a request with the specified method as the request method and the specified submission target as the request URI. Usually, the submission target is the same resource as the form context, but may be a different resource. Constructing and sending the request is called submitting the form.¶

A form can occur as a top-level element in a document or as a nested element within a link. When a form occurs as a top-level element, the form context implicitly is the document's retrieval context. When a form occurs nested within a link, the form context is the link target of the enclosing link.¶

2.7. Form Fields

Form fields can be used to provide more detailed instructions to agents for constructing the request when submitting a form. For example, a form field could instruct an agent to include a certain payload or header field in the request. A payload could, for instance, be described by form fields providing acceptable media types, a reference to schema information, or a number of individual data items that the agents needs to supply. Form fields can be specific to the Web transfer protocol that is used for submitting the form.¶

A form field is a pair of a form field type and a form field value. Additionally, a form field may have a list of zero or more nested elements that further describe the form field value.¶

A form field type identifies the semantics of the form field. Form field types are predicates and thus URIs. Form field values are URIs, blank nodes or literals.¶

2.8. Navigation

An agent begins the interaction with an application by performing a GET request on an entry point URI. The entry point URI is the only URI that the agent is expected to know beforehand. From then on, the agent is expected to make all requests by following links and submitting forms that are provided in the responses resulting from the requests. The entry point URI could be obtained through some discovery process or manual configuration.¶

If dereferencing the entry point URI yields a CoRAL document (or any other representation that implements the CoRAL data and interaction model), the agent makes this document the active representation in the browsing context and proceeds as follows:¶

1. The first step for the agent is to decide what to do next, i.e., which type of link to follow or form to submit, based on the link relation types and operation types it understands.¶

   An agent may follow a link without understanding the link relation type, e.g., for the sake of pre-fetching or building a search index. However, an agent MUST NOT submit a form without understanding the operation type.¶

2. The agent then finds the link(s) or form(s) with the respective type in the active representation. This may yield one or more candidates, from which the agent will have to select the most appropriate one. The set of candidates can be empty, for example, when an application state transition is not supported or not allowed.¶

3. The agent selects one of the candidates based on the metadata associated them (in the form of form fields and nested elements) and their order of appearance in the document. Examples for relevant metadata could include the indication of a media type for the target resource representation, the URI scheme of a target resource, or the request method of an operation.¶

4. The agent obtains the request URI from the link target or submission target. Fragment identifiers are not part of the request URI and MUST be separated from the rest of the URI prior to the next step.¶

5. The agent constructs a new request with the request URI. If the agent is following a link, then the request method MUST be GET. If the agent is submitting a form, then the request method MUST be the one supplied by the form.¶

   The agent SHOULD set HTTP header fields and CoAP request options according to the metadata (e.g., set the HTTP Accept header field or the CoAP Accept option when a media type for the target resource is provided). Depending on the operation type of a form, the agent may also have to include a request payload that matches the specifications of some form fields.¶

6. The agent sends the request and receives the response.¶

7. If a fragment identifier was separated from the request URI, the agent selects the fragment indicated by the fragment identifier within the received representation according to the semantics of its media type.¶

8. The agent updates the browsing context by making the (selected fragment of the) received representation the active representation.¶

9. Finally, the agent processes the representation according to the semantics of its media type. If the representation is a CoRAL document (or any other representation that implements the CoRAL data and interaction model), the agent again has the choice of what to do next. Go to step 1.¶

2.9. History Traversal

A browsing context has a session history, which lists the resource representations that the agent has processed, is processing, or will process.¶

A session history consists of session history entries. The number of session history entries may be limited and dependent on the agent. An agent with severe constraints on memory size might only have enough memory for the most recent entry.¶

An entry in the session history consists of a resource representation and the representation's retrieval context. New entries are added to the session history as the agent navigates from resource to resource, discarding entries that are no longer used.¶

An agent can decide to navigate a browsing context (in addition to following links and submitting forms) by traversing the session history. For example, when an agent receives a response with a representation that does not contain any further links or forms, it can navigate back to a resource representation it has visited earlier and make that the active representation.¶

Traversing the history SHOULD take advantage of caches to avoid new requests. An agent may reissue a safe request (e.g., a GET) when it does not have a fresh representation in its cache. An agent MUST NOT reissue an unsafe request (e.g., a PUT or POST) unless it actually intends to perform that operation again.¶

2.10. Designing interactions in an Open World

CoRAL can be used to build both open world systems ("if something is not said, it may or may not be true") and closed world systems ("if something is not said, it is not true").¶

In constrained environments (and the web in general), partial representations are often used for efficiency. For example, a device can query another for particular statements using a yet to be defined FETCH version of CoRAL. It is expected that some tools (e.g., server or agent libraries) require the application to be tolerant of unprocessed statements. Furthermore, it can be easier to evolve applications and their packing dictionaries if loss of statements leads to graceful degradation.¶

Therefore, it is convenient to build applications on open world assumptions. Such applications can only use statements that add possibilities, and none that limit interactions. Any limitations need to be encoded in statements the agent necesarily has to perform an action in the first place, and can then be relaxed in additional statements.¶

For example, an application built with open-world assumptions can not create a form that allows feeding gremlins, and in an additional statement (e.g., a form field) forbid after midnight. Instead, the application needs to describe a limited-feeding form, which can only be used if any of the attached conditions is met; the condition "before midnight" can then be expressed in an additional statement.¶

3.1. Data Structure

The data structure of a document in the binary format is made up of three kinds of elements: links, forms (as short hands for the statements they are constructed of), and (as an extension to the CoRAL data model) directives. Directives provide a way to encode URI references with a common base more efficiently.¶

3.2. Dictionary Compression

A document in the binary format MAY reference values from an external dictionary using Packed CBOR [I-D.ietf-cbor-packed]. This helps to reduce representation size and processing cost.¶

Dictionary references can be used subject to [ yet to be defined ] profiling.¶

Implementers should note that Packed CBOR is not designed to be uncompressed, but to be used in a compressed form. In particular, constrained devices may operate without even knowing what a given dictionary entry expands to (as long as they know its meaning) .¶

3.3. Export Interface

The definition of documents, links, and forms in the CoRAL binary format can be reused in other CBOR-based protocols. Specifications using CDDL should reference the following rules for this purpose:¶

• CoRAL-Document = document
  CoRAL-Link = link
  CoRAL-Form = form
  

  ¶

For each embedded document, link, and form, the CBOR-based protocol needs to specify the document retrieval context, link context, and form context, respectively.¶

4.1. Submitting Documents

By default, a CoRAL document is a representation that captures the current state of a resource. The meaning of a CoRAL document changes when it is submitted in a request. Depending on the request method, the CoRAL document can capture the intended state of a resource (PUT) or be subject to application-specific processing (POST).¶

4.2. Returning Documents

In a response, the meaning of a CoRAL document changes depending on the request method and the response status code. For example, a CoRAL document in a successful response to a GET represents the current state of the target resource, whereas a CoRAL document in a successful response to a POST might represent either the processing result or the new resource state. A CoRAL document in an error response represents the error condition, usually describing the error state and what next steps are suggested for resolving it.¶

5.1. Specifying CoRAL-based Applications

CoRAL-based applications naturally implement the Web architecture [W3C.REC-webarch-20041215] and thus are centered around orthogonal specifications for identification, interaction, and representation:¶

• Resources are identified by URIs or represented by literal values.¶

• Interactions are based on the hypermedia interaction model of the Web and the methods provided by the Web transfer protocol. The semantics of possible interactions are identified by link relation types and operation types.¶

• Representations are CoRAL documents encoded in the binary format defined in Section 3. Depending on the application, additional representation formats may be used.¶

5.2. Minting Vocabulary

New link relation types, operation types, and form field types can be minted by defining a URI that uniquely identifies the item. Although the URI may point to a resource that contains a definition of the semantics, clients SHOULD NOT automatically access that resource to avoid overburdening its server. The URI SHOULD be under the control of the person or party defining it, or be delegated to them.¶

To avoid interoperability problems, it is RECOMMENDED that only URIs are minted that are normalized according to Section 6.2 of [RFC3986]. This is easily achieved when the URIs are defined in CRI form (in which they also become part of the dictionary), as this avoids many common non-normalized forms of URIs by construction.¶

Non-normalized forms that are still to be avoided include:¶

• Uppercase characters in scheme names and domain names¶

• Explicitly stated HTTP default port (e.g., <http://example.com/> is preferable over <http://example.com:80/>)¶

• Punycode-encoding of Internationalized Domain Names in URIs¶

• URIs that are not in Unicode Normalization Form C¶

URIs that identify vocabulary do not need to be registered. The inclusion of domain names in URIs allows for the decentralized creation of new URIs without the risk of collisions.¶

However, URIs can be relatively verbose and impose a high overhead on a representation. This can be a problem in constrained environments [RFC7228]. Therefore, CoRAL alternatively allows the use of packed references that abbreviate CBOR data items from a dictionary, as specified in Section 3.2. These impose a much smaller overhead but instead need to be assigned by an authority to avoid collisions.¶

5.3. Expressing Registered Link Relation Types

Link relation types registered in the Link Relations Registry [LINK-RELATIONS], such as collection [RFC6573] or icon [W3C.REC-html52-20171214], can be used in CoRAL by appending the registered name to the URI <http://www.iana.org/assignments/relation/>:¶

• #using iana = <http://www.iana.org/assignments/relation/>
  
  iana:collection </items>
  iana:icon       </favicon.png>
  

  ¶

The convention of appending the relation type name to the prefix <http://www.iana.org/assignments/relation/> to form URIs is adopted from the Atom Syndication Format [RFC4287]; see also Appendix A.2 of [RFC8288].¶

Note that registered relation type names are required to be lowercase ASCII letters (see Section 3.3 of [RFC8288]).¶

5.4. Expressing Simple RDF Statements

In RDF [W3C.REC-rdf11-concepts-20140225], a statement says that some relationship, indicated by a predicate, holds between two resources. Existing RDF vocabularies can therefore be a good source for link relation types that describe resource metadata. For example, a CoRAL document could use the FOAF vocabulary [FOAF] to describe the person or software that made it:¶

• #using rdf = <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
  #using foaf = <http://xmlns.com/foaf/0.1/>
  
  foaf:maker null {
     rdf:type        <http://xmlns.com/foaf/0.1/Person>
     foaf:familyName "Hartke"
     foaf:givenName  "Klaus"
     foaf:mbox       <mailto:klaus.hartke@ericsson.com>
  }
  

  ¶

5.5. Expressing Natural Language Texts

Text strings can be associated with a Language Tag [RFC5646] and a base text direction (right-to-left or left-to-right) by using CBOR tag 38.¶

• #using base = <http://coreapps.org/base#>
  #using iana = <http://www.iana.org/assignments/relation/>
  
  iana:terms-of-service </tos> {
     base:title 38(["de", "Nutzungsbedingungen"])
     base:title 38(["en-US", "Terms of use"])
     base:title 38(["az", "ltr", "İstifadə şərtləri"])
  }
  

  ¶

[ Maturity note: Whether direction will actually be expressed in an updated tag 38, how precisely that is done, or whether a new tag will be allocated for text with direction is currently still under discussion. ]¶

5.6. Embedding Representations in CoRAL

When a document links to many Web resources and an agent needs a representation of each of them, it can be inefficient to retrieve each representations individually. To minimize round-trips, documents can embed representations of resources.¶

A representation can be embedded in a document by including a link of type <http://coreapps.org/base#representation>:¶

• #using base = <http://coreapps.org/base#>
  #using http = <http://coreapps.org/http#>
  #using iana = <http://www.iana.org/assignments/relation/>
  
  iana:icon </favicon.gif> {
     base:representation
        b64'R0lGODlhAQABAAAAACH5BAEKAAEALAAAAAABAAEAAAIAOw==' {
        http:type "image/gif"
     }
  }
  

  ¶

An embedded representation SHOULD have a nested link of type <http://coreapps.org/http#type> or <http://coreapps.org/coap#type> that indicates the content type of the representation.¶

The link relation types <http://coreapps.org/base#representation>, <http://coreapps.org/http#type>, and <http://coreapps.org/coap#type> are defined in Appendix A.¶

7.1. Media Type "application/coral+cbor"

This document registers the media type application/coral+cbor according to the procedures of [RFC6838].¶

Type name:

application¶

Subtype name:

coral+cbor¶

Required parameters:

N/A¶

Optional parameters:

dictionary - See Section 3.2 of [I-D.ietf-core-coral].¶

Encoding considerations:

binary - See Section 3 of [I-D.ietf-core-coral].¶

Security considerations:

See Section 6 of [I-D.ietf-core-coral].¶

Interoperability considerations:

N/A¶

Published specification:

[I-D.ietf-core-coral]¶

Applications that use this media type:

See Section 1 of [I-D.ietf-core-coral].¶

Fragment identifier considerations:

As specified for application/cbor.¶

Additional information:

Deprecated alias names for this type:

N/A¶

Magic number(s):

N/A¶

File extension(s):

.coral.cbor¶

Macintosh file type code(s):

N/A¶

Person & email address to contact for further information:

See the Author's Address section of [I-D.ietf-core-coral].¶

Intended usage:

COMMON¶

Restrictions on usage:

N/A¶

Author:

See the Author's Address section of [I-D.ietf-core-coral].¶

Change controller:

IESG¶

Provisional registration?

No¶

7.2. CoAP Content Formats

This document registers CoAP content formats for the content types application/coral+cbor and text/coral according to the procedures of [RFC7252].¶

• Content Type:

  application/coral+cbor¶

  Content Coding:

  identity¶

  ID:

  TBD3¶

  Reference:

  [I-D.ietf-core-coral]¶

[[NOTE TO RFC EDITOR: Please replace all occurrences of TBD3 in this document with the code points assigned by IANA.]]¶

[[NOTE TO IMPLEMENTERS: Experimental implementations may use content format ID 65087 for application/coral+cbor until IANA has assigned code points.]]¶

A.1. Base

Link Relation Types:¶

<http://www.w3.org/1999/02/22-rdf-syntax-ns#type>

Indicates that the link's context is an instance of the class specified as the link's target, as defined by RDF Schema [W3C.REC-rdf-schema-20140225].¶

<http://coreapps.org/base#title>

Indicates that the link target is a human-readable label (e.g., a menu entry).¶

The link target MUST be a literal. The text string SHOULD be wrapped in a tag indicating language and, if necessary, direction if applicable.¶

<http://coreapps.org/base#representation>

Indicates that the link target is a representation of the link context.¶

The link target MUST be a byte string.¶

The representation may be a full, partial, or inconsistent version of the representation served from the URI of the resource.¶

A link with this link relation type can occur as a top-level element in a document or as a nested element within a link. When it occurs as a top-level element, it provides an alternate representation of the document's retrieval context. When it occurs nested within a link, it provides a representation of link target of the enclosing link.¶

Operation Types:¶

<http://coreapps.org/base#update>

Indicates that the state of the form's context can be replaced with the state described by a representation submitted to the server.¶

This operation type defaults to the PUT method [RFC7231] [RFC7252] for both HTTP and CoAP. Typical overrides by a form field include the PATCH method [RFC5789] [RFC8132] for HTTP and CoAP and the iPATCH method [RFC8132] for CoAP.¶

<http://coreapps.org/base#search>

Indicates that the form's context can be searched by submitting a search query.¶

This operation type defaults to the POST method [RFC7231] for HTTP and the FETCH method [RFC8132] for CoAP. Typical overrides by a form field include the POST method [RFC7252] for CoAP.¶

A.2. Collections

Link Relation Types:¶

<http://www.iana.org/assignments/relation/item>

Indicates that the link's context is a collection and that the link's target is a member of that collection, as defined in Section 2.1 of [RFC6573].¶

<http://www.iana.org/assignments/relation/collection>

Indicates that the link's target is a collection and that the link's context is a member of that collection, as defined in Section 2.2 of [RFC6573].¶

Operation Types:¶

<http://coreapps.org/collections#create>

Indicates that the form's context is a collection and that a new item can be created in that collection with the state defined by a representation submitted to the server.¶

This operation type defaults to the POST method [RFC7231] [RFC7252] for both HTTP and CoAP.¶

<http://coreapps.org/collections#delete>

Indicates that the form's context is a member of a collection and that the form's context can be removed from that collection.¶

This operation type defaults to the DELETE method [RFC7231] [RFC7252] for both HTTP and CoAP.¶

A.3. HTTP

Form Field Types:¶

<http://coreapps.org/http#method>

Specifies the HTTP method for the request.¶

The form field value MUST be a text string in the format defined in Section 4.1 of [RFC7231]. The possible set of values is maintained in the HTTP Methods Registry [HTTP-METHODS].¶

A form field of this type MUST NOT occur more than once in a form. If absent, it defaults to the request method implied by the form's operation type.¶

<http://coreapps.org/http#accept>

Specifies an acceptable HTTP content type for the request payload. There may be multiple form fields of this type. If a form does not include a form field of this type, the server accepts any or no request payload, depending on the operation type.¶

The form field value MUST be a text string in the format defined in Section 3.1.1.1 of [RFC7231]. The possible set of media types and their parameters is maintained in the Media Types Registry [MEDIA-TYPES].¶

Link Relation Types:¶

<http://coreapps.org/http#type>

Specifies the HTTP content type of the link context.¶

The link target MUST be a text string in the format defined in Section 3.1.1.1 of [RFC7231]. The possible set of media types and their parameters is maintained in the Media Types Registry [MEDIA-TYPES].¶

A.4. CoAP

Form Field Types:¶

<http://coreapps.org/coap#method>

Specifies the CoAP method for the request.¶

The form field value MUST be an integer identifying a CoAP method (e.g., the integer 2 for the POST method). The possible set of values is maintained in the CoAP Method Codes Registry [CORE-PARAMETERS].¶

A form field of this type MUST NOT occur more than once in a form. If absent, it defaults to the request method implied by the form's operation type.¶

<http://coreapps.org/coap#accept>

Specifies an acceptable CoAP content format for the request payload. There may be multiple form fields of this type. If a form does not include a form field of this type, the server accepts any or no request payload, depending on the operation type.¶

The form field value MUST be an integer identifying a CoAP content format. The possible set of values is maintained in the CoAP Content Formats Registry [CORE-PARAMETERS].¶

Link Relation Types:¶

<http://coreapps.org/coap#type>

Specifies the CoAP content format of the link context.¶

The link target MUST be an integer identifying a CoAP content format (e.g., the integer 42 for the content type application/octet-stream without a content coding). The possible set of values is maintained in the CoAP Content Formats Registry [CORE-PARAMETERS].¶

C.1. RDF

[ TBD: Expand / introduce the common CURIEs used here. ]¶

RDF and the CoRAL Basic Information Model can be interconverted losslessly, as long as some basic restrictions are met:¶

• All involved IRIs (on the RDF side) and CRIs (on the CoRAL side) can be converted; that means that round-tripping IRIs through CoRAL converts them to the equivalent URIs.¶

  The precise limitations of what CRIs can not express are described in [I-D.ietf-core-href] and out of scope of this document.¶

  A possible extension to CoRAL that allows tagged URIs in place of CRIs could remove this limitation. (CRIs that can not be expressed as URIs are not valid anyway).¶

• A blank node of CoRAL can only have one incoming edge in serialization. RDF documents with multiply connected blank nodes need to undergo skolemization before they can be expressed in CoRAL.¶

• CoRAL supports arbitrary literal objects, including CBOR tags. For each object that is used in a literal, a mapping to a datatype (typically XSD) needs to be defined.¶

  When literals are normalized in RDF according to XSD rules, or the literal mappings to RDF datatypes are ambiguous on the CoRAL side, round-tripping CoRAL through RDF can be lossy to the extent of the normalization or ambiguity.¶

• As always with expressing arbitrary graphs of the Basic Information Model in serialization, if there is no directed tree spanning the directed graph, statements need to be introduced to reach some topics.¶

Each statement in RDF is mapped to a statement in CoRAL. Any IRI it contains in RDF is mapped to an equivalent CRI in CoRAL and vice versa. Any blank node of RDF is converted to a blank node (serialized as a null) in CoRAL. (Beware that depending on the context established in Section 4, the retrieval context may be a URI or a blank node). Literals are converted as follows:¶

• CBOR text strings are coverted to RDF string literals without a language tag.¶

• CBOR literals from the following list are converted to their corresponding text representations of the datatype from the following table:¶

[ TBD: Check compatibilities, give type for at least the basic tags. Directional text might wind up in tag 38, ]¶

• RDF literals are mapped to any CoRAL literal that yields an equivalent RDF literal in the opposite direction.¶

<foaf:Person rdf:about="#danbri" xmlns:foaf="http://xmlns.com/foaf/0.1/">
  <foaf:name>Dan Brickley</foaf:name>
  <foaf:homepage rdf:resource="http://danbri.org/" />
  <foaf:openid rdf:resource="http://danbri.org/" />
  <foaf:img rdf:resource="/images/me.jpg" />
</foaf:Person>

51([[cri'http://danbri.org/'], [<<-3, "xmlns.com", ["foaf", "0.1"], null>>], [], [
  [2, cri'http://www.iana.org/assignments/relation/carries-information-about', cri'/me.xml#danbri',
    [2, cri'http://www.w3.org/1999/02/22-rdf-syntax-ns#type', 6(<<'Person'>>)],
    [2, 6(<<'name'>>), "Dan Brickley"],
    [2, 6(<<'homepage'>>), 6(0)],
    [2, 6(<<'openid'>>), 6(0)],
    [2, 6(<<'img'>>), cri'/images/me.jpg']
  ]
]])

C.2. CoRE Link Format

Generic information in Web Links as described in [RFC8288] can not be converted to CoRAL in any practical way: Attributes are not managed, and it is not clear from the syntax whether an attribute is making a statement about the link or its target. (See Section 2.3.3.1 for an example).¶

Applications that use links with the attribute semantics common in the CoRE ecosystem (typically used with [RFC6690] Link Format) can use this conversion. It defines terms for common properties used for discovering resources, and describes a way to compatibly extend the mapping.¶

The same mechanism (but probably with a different mapping between names and attributes, and different rules about the necessity of packing entries) can be defined for any data model that builds on [RFC8288] semantics, e.g., the links sent in headers or payloads about [RFC7089] mementos, or applications building on [I-D.ietf-httpapi-linkset].¶

In several points the mapping describes URIs to necessarily have an entry in the packing table; this refers to the profiling described further down. Parts of a Link Format document that would need an entry but do not have one can not be converted; these are ignored in the conversion unless the converter is configured to be strict and fail the complete conversion in that case.¶

This mapping from Link Format to CoRAL is performed as follows: * For each relation in a link, a statement is created mapping the link context to the subject, the link target to the object and the relation to the predicate.¶

If the relation is of ext-rel-type, it is used as a URI as is. Otherwise it is a registered value, prefixed with http://www.iana.org/assignments/relation/ and necessarily packed using table TBD. (This is equivalent to the RPP mechanism for attribute values).¶

• Each target attribute is converted to one or more statements by the mechanism indicated for the attribute name in the following table. Statements produced from a link have the target as its subject, the attribute name without any trailing asterisk (prefixed with https://TBD/ [ to be picked together with IANA as it'll be a registry ]) as its predicate, and the object(s) depending on the mechanism.¶

  Attributes are necessarily listed in this table.¶

Available mechanisms are:¶

• SPSP (space split): Link format values are split at space characters (SP in the RFC6690 ABNF), and all values treated using another mechanism.¶

• string: The attribute value is stored as a text string literal. If the Link Format attribute is language tagged (i.e. when the attribute name ends with an asterisk and the value is of ext-value shape), the literal is encapsulated in a CBOR language tag (38).¶

• int: The target attribute is processed as an ASCII encoded number and expressed as an integer literal. A failing conversion is treated like an unknown registered value: It is ignored unless configured otherwise.¶

• RPP (registered-prefix / packed): The input value (often the result of the SPSP mechanism) is parsed according to the relation-type ABNF production. If it is of ext-rel-type, it is expressed as that URI. If it is prefixed with the string indicated with the mechanism, and necessarily compressed through table TBD.¶

All currently registered link attributes are used in the CoRE ecosystem as indicating a property of the target that is independent of the link being followed. If this conversion is to be extended to cover attributes that pertain to the full link being followed (typically along with one or more link relations), the relevant relations are not expressed as a single statement, but as a form, i.e. as two statements linking the context to a blank node and the blank node to the target; the attributes are attached to the blank node. The precise mechanism out of scope for this document, and left to those who first register such an attribute.¶

Some structure can be carried over from Link Format to the structured model: The sequences of links gets reused, and the set and sequence of attributes in a particular occurrence of a link get applied to the statement produced from the link (or all the statements, if the link has multiple link relations). Statements whose subject is not the document itself are attached to the retrieval context using the necessarily packed http://www.iana.org/assignments/relation/carries-information-about property. Statements about URLs mentioned elsewhere in the document can be expressed there instead.¶

Link relations of the reg-name form, link attributes, and attribute values from the RPP mechanism MUST be serialized using packed CBOR as initialized in table TBD. No other packing is used. A consumer MAY ignore any items compressed through the dictionary for which it does not know the expanded version: These necessarily represent statements that involve terms the consumer does not understand.¶

[ As an alternative, packing attributes together with their URIs is considered: Rather than [2, 6(/ attr:rt /), 6(/ rt:core.rd /)] we could have 6(rt-core) right away; unregistered values would stay [2, 6(/ attr:rt /), value] or maybe 254([value]) using prefix packing. ]¶