First pass at a media types registry.

_config.yml

@@ -18,6 +18,11 @@ collections:
     name: Format Registry
     output: true
     permalink: /registry/:collection/:title
+  media-type:
+    slug: media-type
+    name: Media Type Registry
+    output: true
+    permalink: /registry/:collection/:title
   extension:
     slug: extension
     name: Specification Extension Registry

_includes/media-type-entry.md

@@ -0,0 +1,22 @@
+# <a href=".">{{ page.collection | replace 'media-type', 'Media Type' | replace '-', ' ' }}</a>
+
+## {{ page.description }}
+
+**[Media Type](https://spec.openapis.org/oas/latest.html#media-types):** `{{ page.media_type }}` {% if value.unregistered %}_unregistered_ {% endif %} ([{{ page.specification.name }}]({{ page.specification.url }}))
+
+**OAS Reference:** [{{ page.reference.section }}](https://spec.openapis.org/oas/latest.html#{{ page.reference.anchor }})
+
+{{ include.summary }}
+
+{% if page.issue %}
+### GitHub Issue
+
+* [#{{ page.issue }}](https://github.com/OAI/OpenAPI-Specification/issues/{{ page.issue }})
+{% endif %}
+
+{% if page.remarks %}
+### Remarks
+
+{{ page.remarks }}
+{% endif %}
+

registries/_media-type/application_json_seq.md

@@ -0,0 +1,20 @@
+---
+owner: handrews
+issue:
+description: JSON Text Sequences
+specification:
+    name: RFC7464
+    url: https://www.rfc-editor.org/rfc/rfc7464.html
+media_type: application/json_seq
+reference:
+    section: Sequential JSON
+    anchor: sequential-json
+versions: "3.2+"
+layout: default
+---
+
+{% capture summary %}
+JSON Text Sequences uses the same approach as all sequential JSON media types.
+{% endcapture %}
+
+{% include media-type-entry.md summary=summary remarks=remarks %}  

registries/_media-type/application_jsonl.md

@@ -0,0 +1,21 @@
+---
+owner: handrews
+issue:
+description: JSON Lines
+specification:
+    name: JSONL
+    url: https://jsonlines.org/
+media_type: application/jsonl
+media_type_unregistered: true
+reference:
+    section: Sequential JSON
+    anchor: sequential-json
+versions: "3.2+"
+layout: default
+---
+
+{% capture summary %}
+JSON Lines uses the same approach as all sequential JSON media types.
+{% endcapture %}
+
+{% include media-type-entry.md summary=summary remarks=remarks %}  

registries/_media-type/application_octet-stream.md

@@ -0,0 +1,19 @@
+---
+owner: handrews
+issue:
+description: Binary or Unknown
+media_type: application/octet-stream
+specification:
+    name: RFC2046 §4.5.1
+    url: https://www.rfc-editor.org/rfc/rfc2046.html#section-4.5.1
+reference:
+    section: Working with Binary Data
+    anchor: working-with-binary-data
+layout: default
+---
+
+{% capture summary %}
+Binary data (also including `image/*`, `video/*`, `audio/*` and other binary media types) is modeled using an empty Schema Object, in accordance with JSON Schema's guidance regarding [non-JSON instances](https://www.ietf.org/archive/id/draft-bhutton-json-schema-01.html#name-non-json-instances).  Note that as specified in the linked reference section ("Working with Binary Data"), modeling binary data that has been encoded into a string is handled differently from raw binary data, with two variations: One when an Encoding Object is involved, and one when no Encoding Object is involved.
+{% endcapture %}
+
+{% include media-type-entry.md summary=summary remarks=remarks %}  

registries/_media-type/application_x-ndjson.md

@@ -0,0 +1,21 @@
+---
+owner: handrews
+issue:
+description: Newline Delimited JSON
+specification:
+    name: NDJSON
+    url: https://github.com/ndjson/ndjson-spec/blob/master/README.md
+media_type: application/x-ndjson
+media_type_unregistered: true
+reference:
+    section: Sequential JSON
+    anchor: sequential-json
+versions: "3.2+"
+layout: default
+---
+
+{% capture summary %}
+Newline Delimited JSON uses the same approach as all sequential JSON media types.
+{% endcapture %}
+
+{% include media-type-entry.md summary=summary remarks=remarks %}  

registries/_media-type/application_x-www-form-urlencoded.md

@@ -0,0 +1,19 @@
+---
+owner: handrews
+issue:
+description: URL-Encoded Forms
+media_type: application/x-www-form-urlencoded
+specification:
+    name: WHATWG URL
+    url: https://url.spec.whatwg.org/#application/x-www-form-urlencoded
+reference:
+    section: Support for x-www-form-urlencoded Request Bodies
+    anchor: support-for-x-www-form-urlencoded-request-bodies
+layout: default
+---
+
+{% capture summary %}
+URL-Encoded forms use the Encoding Object to control how the JSON-like structure defined by the Schema Object maps to the URL query string-like format.  Note that this is separate from how URL query parameters are managed, which is done with the Parameter Object.
+{% endcapture %}
+
+{% include media-type-entry.md summary=summary remarks=remarks %}  

registries/_media-type/application_xml.md

@@ -0,0 +1,19 @@
+---
+owner: handrews
+issue:
+description: XML
+media_type: application/xml
+specification:
+    name: RFC7303
+    url: https://www.rfc-editor.org/rfc/rfc7303.html
+reference:
+    section: XML Object
+    anchor: xml-object
+layout: default
+---
+
+{% capture summary %}
+XML is modeled using the OAS's `xml` extension keyword for JSON Schema, which has an XML Object as its value.  This allows fine-grained control over how each part of the JSON Schema description maps to XML elements or attributes.
+{% endcapture %}
+
+{% include media-type-entry.md summary=summary remarks=remarks %}  

registries/_media-type/multipart_form-data.md

@@ -0,0 +1,19 @@
+---
+owner: handrews
+issue:
+description: Multipart Form Data
+media_type: multipart/form-data
+specification:
+    name: RFC7578
+    url: https://www.rfc-editor.org/rfc/rfc7578.html
+reference:
+    section: Encoding multipart Media Types
+    anchor: encoding-multipart-media-types
+layout: default
+---
+
+{% capture summary %}
+Multipart forms use the Encoding Object to control how the JSON-like structure defined by the Schema Object maps to each part.  Multipart media types that do not use named parts cannot be handled with this technique, although it may be possible to use `Content-Disposition: form` with a name parameter with such media types, but as no specification recommends this, support is unlikely to be dependable.
+{% endcapture %}
+
+{% include media-type-entry.md summary=summary remarks=remarks %}  

registries/_media-type/text_event-stream.md

@@ -0,0 +1,21 @@
+---
+owner: handrews
+issue:
+description: SSE Events
+specification:
+    name: WHATWG HTML
+    url: https://html.spec.whatwg.org/multipage/iana.html#text/event-stream
+media_type: text/event-stream
+media_type_unregistered: true
+reference:
+    section: Server-Sent Event Streams
+    anchor: server-sent-event-streams
+versions: "3.2+"
+layout: default
+---
+
+{% capture summary %}
+Event streams build on the sequential media type support used by sequential JSON media types by further defining a mapping of the individual event format into the Schema Object's data model.
+{% endcapture %}
+
+{% include media-type-entry.md summary=summary remarks=remarks %}  

registries/_media-type/text_plain.md

@@ -0,0 +1,19 @@
+---
+owner: handrews
+issue:
+description: Plain Text
+media_type: text/plain
+specification:
+    name: RFC2046
+    url: https://www.rfc-editor.org/rfc/rfc2046.html
+reference:
+    section: Encoding Object
+    anchor: encoding-object
+layout: default
+---
+
+{% capture summary %}
+Plain text is modeled as a single string.  Note that unlike JSON strings, the contents of the string representing the plain text are not quoted when serializing to a document.  While a Schema Object of `{type: string, const: foo}` for JSON validates the JSON value `"foo"`, for plain text it validates `foo`, without quotes.
+{% endcapture %}
+
+{% include media-type-entry.md summary=summary remarks=remarks %}  

registry/media-type.md

@@ -0,0 +1,46 @@
+---
+title: Media Type Registry
+layout: default
+permalink: /registry/media-type/index.html
+parent: Registry
+---
+
+# Media Type Registry
+
+This registry defines how to use the Schema Object, Media Type Object, and in some cases other Objects to model media types other than `application/json` or media types using a `+json` suffix.
+
+## Data Modeling vs Mapping
+
+JSON Schema operates on an in-memory [data model](https://www.ietf.org/archive/id/draft-bhutton-json-schema-01.html#name-instance-data-model) based on the [JSON RFC](https://www.rfc-editor.org/rfc/rfc8259.html#section-3), which is different from the set of types used by JSON Schema's `type` keyword.
+
+JSON Schema's data model description includes guidance on how to _map_ JSON documents into the data model, such as noting that whitespace and different lexical representations of numbers (such as `1` vs `1.0`) are **not** significant within the data model.
+
+All in-memory data described by the OpenAPI Specification (OAS) uses the same in-memory data model, as described under the "Data Types" section.
+However, the OAS defines _mappings_ for several additional media types, where JSON Schema is used on a JSON-like in-memory representation which may have a significantly different structure from the media type's representation in [HTTP content](https://www.rfc-editor.org/rfc/rfc9110.html#name-content).
+This registry documents those mappings, and in the future may document additional mappings not explicitly mentioned in the OAS.
+
+### Setting the Media Type
+
+JSON Schema draft 2020-12 offers [keywords for modeling embedded media types](https://www.ietf.org/archive/id/draft-bhutton-json-schema-validation-01.html#name-a-vocabulary-for-the-conten): `contentMediaType`, `contentEncoding`, and `contentSchema`, which can be used to set a media type, encoding, or schema for [certain types of data](https://spec.openapis.org/oas/latest.html#working-with-binary-data).  These keywords, most notably `contentMediaType`, can contradict media types set in the parent key of a Media Type Object, or by an Encoding Object (including by the default Encoding Object when an Encoding Object is relevant but not present).  In such cases, the Media Type Object key or the Encoding Object ***always*** take precedence over the JSON Schema keywords.
+
+## Specification Versions
+
+This registry is being created for the OpenAPI Specification (OAS) version 3.2, and requirements regarding its support will be included in that specification's text.
+
+Implementations MAY support these data modeling techniques in other OAS versions or other specifications such as Arazzo, as long as the necessary Objects and fields are supported in those specification versions.
+
+## Contributing
+
+While most OpenAPI Initiative registries invite community contributions, this registry is somewhat experimental.
+Please open a [Discussion](https://github.com/OAI/OpenAPI-Specification/discussions) explaining your use cases for any media type(s) you would like to see added, rather than proposing a solution.
+Solution proposals will be invited _after_ use cases are accepted.
+
+## Media Types
+
+**Note:** For any media type with structured suffix usage (e.g. `application/openapi+json` uses the structured suffix associated with `application/json`), the registered techniques for the media type also apply to media types using the related structured suffix.
+
+|Media Type|Name|Specification|OAS Reference|OAS Versions|
+|---|---|---|
+{% for value in site.media-type %}| <tt><a href="{{ value.slug }}">{{ value.media_type }}</a></tt> {% if value.unregistered %}_(unregistered)_ {% endif %} | {{ value.description }} | <a href="{{ value.specification.url }}">{{ value.specification.name }}</a> | <a href="https://spec.openapis.org/oas/latest.html#{{ value.reference.anchor }}">{{ value.reference.section }}</a> | {{ value.versions }} |
+{% endfor %}
+