YAML Media Type

1.1. 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. These words may also appear in this document in lower case as plain English words, absent their normative meanings.¶

The terms "content", "content negotiation", "resource", and "user agent" in this document are to be interpreted as in [HTTP].¶

The terms "fragment" and "fragment identifier" in this document are to be interpreted as in [URI].¶

The terms "presentation", "stream", "YAML document", "representation graph", "tag", "serialization detail", "node", "alias node", "anchor" and "anchor name" in this document are to be interpreted as in [YAML].¶

Figures containing YAML code always start with the "%YAML 1.2" directive to improve readability.¶

1.2. Fragment identification

A fragment identifies a node in a stream.¶

A fragment identifier starting with "*" is to be interpreted as a YAML alias node (see Section 1.2.1).¶

For single-document YAML streams, a fragment identifier that is empty or that starts with "/" is to be interpreted as a JSON Pointer [JSON-POINTER] and is evaluated on the YAML representation graph, walking through alias nodes; in particular, the empty fragment identifier references the root node. This syntax can only reference the YAML nodes that are on a path that is made up of nodes interoperable with the JSON data model (see Section 3.4).¶

A fragment identifier is not guaranteed to reference an existing node. Therefore, applications SHOULD define how an unresolved alias node ought to be handled.¶

 %YAML 1.2
 ---
 one: &foo scalar
 two: &bar
   - some
   - sequence
   - items
 ...
 %YAML 1.2
 ---
 &document_2
 one: &foo [a, sequence]

2.1. Media Type application/yaml

The media type for YAML text is application/yaml; the following information serves as the registration form for this media type.¶

Type name:

application¶

Subtype name:

yaml¶

Required parameters:

N/A¶

Optional parameters:

N/A; unrecognized parameters should be ignored¶

Encoding considerations:

binary¶

Security considerations:

see Section 4 of this document¶

Interoperability considerations:

see Section 3 of this document¶

Published specification:

[YAML]¶

Applications that use this media type:

Applications that need a human-friendly, cross language, Unicode based data serialization language designed around the common native data types of dynamic programming languages.¶

Fragment identifier considerations:

See Section 1.2¶

Additional information:¶

• Deprecated alias names for this type: application/x-yaml, text/yaml, text/x-yaml. These names are used, but not registered.¶
• Magic number(s) N/A¶
• File extension(s): "yaml" (preferred), "yml". See Section 3.3.¶
• Macintosh file type code(s): N/A¶
• Windows Clipboard Name: YAML¶

Person and email address to contact for further information:

See Authors' Addresses section.¶

Intended usage:

COMMON¶

Restrictions on usage:

None.¶

Author:

See Authors' Addresses section.¶

Change controller:

IESG¶

2.2. The +yaml Structured Syntax Suffix

The suffix +yaml MAY be used with any media type whose representation follows that established for application/yaml. The media type structured syntax suffix registration form follows. See [MEDIATYPE] for definitions of each of the registration form headings.¶

Name:

YAML Ain't Markup Language (YAML)¶

+suffix:

+yaml¶

References:

[YAML]¶

Encoding considerations:

see Section 2.1¶

Fragment identifier considerations:

Differently from application/yaml, there is no fragment identification syntax defined for +yaml.¶

A specific xxx/yyy+yaml media type needs to define the syntax and semantics for fragment identifiers because the ones in Section 2.1 do not apply unless explicitly expressed.¶

Interoperability considerations:

See Section 2.1¶

Security considerations:

See Section 2.1¶

Contact:

httpapi@ietf.org or art@ietf.org¶

Author:

See Authors' Addresses section¶

Change controller:

IESG¶

3.1. YAML is an Evolving Language

YAML is an evolving language and, over time, some features have been added and others removed.¶

This [YAML] media type registration is independent of YAML version. This allows content negotiation of version-independent YAML resources.¶

Implementers concerned about features related to a specific YAML version can specify it in YAML documents using the %YAML directive (see Section 6.8.1 of [YAML]).¶

3.2. YAML streams

A YAML stream can contain zero or more YAML documents.¶

When receiving a multi-document stream, an application that only expects one-document streams, ought to signal an error instead of ignoring the extra documents.¶

Current implementations consider different documents in a stream independent, similarly to JSON Text Sequences (see [RFC7464]); elements such as anchors are not guaranteed to be referenceable across different documents.¶

3.3. Filename extension

The "yaml" filename extension is the preferred one; it is the most popular and widely used on the web. The "yml" filename extension is still used. The simultaneous usage of two filename extensions in the same context might cause interoperability issues (e.g., when both a "config.yaml" and a "config.yml" are present).¶

3.4. YAML and JSON

When using flow collection styles (see Section 7.4 of [YAML]) a YAML document could look like JSON [JSON], thus similar interoperability considerations apply.¶

When using YAML as a more efficient format to serialize information intended to be consumed as JSON, information not reflected in the representation graph and classified as presentation or serialization detail (see Section 3.2 of [YAML]) can be discarded. This includes comments (see Section 3.2.3.3 of [YAML]), directives, and alias nodes (see Section 7.1 of [YAML]) that do not have a JSON counterpart.¶

 %YAML 1.2
 ---
 # This comment will be lost
 # when serializing in JSON.
 Title:
   type: string
   maxLength: &text_limit 64

 Name:
   type: string{{Section 8.1 of JSON}}
   maxLength: *text_limit  # Replaced by the value 64.

Implementers need to ensure that relevant information will not be lost during the processing. For example, they might consider acceptable that alias nodes are replaced by static values.¶

In some cases an implementer may want to define a list of allowed YAML features, taking into account that the following ones might have interoperability issues with [JSON]:¶

• multi-document YAML streams;¶
• non UTF-8 encoding. Before encoding YAML streams in UTF-16 or UTF-32, it is important to note that Section 8.1 of [JSON] mandates the use of UTF-8 when exchanging JSON texts between systems that are not part of a closed ecosystem, and that Section 5.2 of [YAML] recommends the use of UTF-8;¶
• mapping keys that are not strings;¶
• circular references represented using anchor (see Section 4.2 and Figure 4);¶
• .inf and .nan float values, since JSON does not support them;¶
• non-JSON types, including the ones associated with tags like !!timestamp that were included in the default schema of older YAML versions;¶
• tags in general, and specifically the ones that do not map to JSON types like custom and local tags such as !!python/object and !mytag (see Section 2.4 of [YAML]);¶

 %YAML 1.2
 ---
 non-json-keys:
   0: a number
   [0, 1]: a sequence
   ? {k: v}
   : a map
 ---
 non-json-keys:
   !date 2020-01-01: a timestamp
 non-json-value: !date 2020-01-01
 ...

3.5. Fragment identifiers

To allow fragment identifiers to traverse alias nodes, the YAML representation graph needs to be generated before the fragment identifier evaluation. It is important that this evaluation will not cause the issues mentioned in Section 3.4 and in Security considerations (Section 4) such as infinite loops and unexpected code execution.¶

Implementers need to consider that the YAML version and supported features (e.g., merge keys) can affect the generation of the representation graph (see Figure 9).¶

In Section 2.1, this document extends the use of specifications based on the JSON data model with support for YAML fragment identifiers. This is to improve the interoperability of already consolidated practices, such as the one of writing OpenAPI documents [OAS] in YAML.¶

Appendix A provides a non-exhaustive list of examples that could help understand interoperability issues related to fragment identifiers.¶

4.1. Arbitrary Code Execution

Care should be used when using YAML tags, because their resolution might trigger unexpected code execution.¶

Code execution in deserializers should be disabled by default, and only be enabled explicitly. In those cases, the implementation should ensure - for example, via specific functions - that the code execution results in strictly bounded time/memory limits.¶

Many implementations provide safe deserializers addressing these issues.¶

4.2. Resource Exhaustion

YAML documents are rooted, connected, directed graphs and can contain reference cycles, so they can't be treated as simple trees (see Section 3.2.1 of [YAML]). An implementation that treats them as simple trees risks going into an infinite loop while traversing the YAML representation graph. This can happen:¶

• when trying to serialize it as JSON;¶
• or when searching/identifying nodes using specifications based on the JSON data model (e.g., [JSON-POINTER]).¶

 %YAML 1.2
 ---
 x: &x
   y: *x

Even if a representation graph is not cyclic, treating it as a simple tree could lead to improper behaviors (such as the "billion laughs" or "Exponential Entity Expansion" problem).¶

 %YAML 1.2
 ---
 x1: &a1 ["a", "a"]
 x2: &a2 [*a1, *a1]
 x3: &a3 [*a2, *a2]

This can be addressed using processors limiting the anchor recursion depth and validating the input before processing it; even in these cases it is important to carefully test the implementation you are going to use. The same considerations apply when serializing a YAML representation graph in a format that does not support reference cycles (see Section 3.4).¶

4.3. YAML streams

Incremental parsing and processing of a YAML stream can produce partial results and later indicate failure to parse the remainder of the stream; to prevent partial processing, implementers might prefer validating all the documents in a stream beforehand.¶

Repeated parsing and re-encoding of a YAML stream can result in the addition or removal of document delimiters (e.g., --- or ...) as well as the modification of anchor names and other serialization details, which can break signature validation.¶

A.1. Unreferenceable nodes

In this example, a couple of YAML nodes that cannot be referenced based on the JSON data model since their mapping keys are not strings.¶

 %YAML 1.2
 ---
 a-map-cannot:
   ? {be: expressed}
   : with a JSON Pointer

 0: no numeric mapping keys in JSON

A.2. Referencing a missing node

In this example the fragment #/0 does not reference an existing node¶

 %YAML 1.2
 ---
 0: "JSON Pointer `#/0` references a string mapping key."

A.3. Representation graph with anchors and cyclic references

In this YAML document, the #/foo/bar/baz fragment identifier traverses the representation graph and references the string you. Moreover, the presence of a cyclic reference implies that there are infinite fragment identifiers #/foo/bat/../bat/bar referencing the &anchor node.¶

 %YAML 1.2
 ---
 anchor: &anchor
   baz: you
 foo: &foo
   bar: *anchor
   bat: *foo

Many YAML implementations will resolve the merge key "<<:" defined in YAML 1.1 in the representation graph. This means that the fragment #/book/author/given_name references the string Federico and that the fragment #/book/<< will not reference any existing node.¶

 %YAML 1.1
 ---
 # Many implementations use merge keys.
 the-viceroys: &the-viceroys
   title: The Viceroys
   author:
     given_name: Federico
     family_name: De Roberto
 book:
   <<: *the-viceroys
   title: The Illusion

Since draft-ietf-httpapi-yaml-mediatypes-02

• clarification on fragment identifiers #50.¶

Since draft-ietf-httpapi-yaml-mediatypes-01

• application/yaml fragment identifiers compatible with JSON Pointer #41 (#47).¶