URI and URI-reference are overlapping types

[alanconway]

This came up while implementing a Go mapping for the cloudevents type system. Here's the definition of URI and URI-reference:

URI - Absolute uniform resource identifier.
String encoding: Absolute URI as defined in RFC 3986 Section 4.3.
URI-reference - Uniform resource identifier reference.
String encoding: URI-reference as defined in RFC 3986 Section 4.1.

The issue is that all legal values of the URI type are also legal values of the URI-reference type. It is unusual to have a separate type to distinguish absolute URIs from relative-or-absolute URIs. In Go, python and ruby there is a single type (named URI or URL), and it has a test such as u.IsAbs() to distinguish absolute from relative values.

The URI type is only used in one place and there the language is ambiguous:

dataschema
Type: URI
Description: Identifies the schema that data adheres to. Incompatible changes to the schema SHOULD be reflected by a different URI. See Versioning of Attributes in the Primer for more information.
Constraints:
OPTIONAL
If present, MUST be a non-empty URI-reference

Note the conflicting claims that this attribute is of type URI but MUST be a URI-reference (which may be an invalid URI!) Can we drop the URI type and use URI-reference everywhere instead?

[duglin]

The second ref should obviously be URI, it’s a typo. I believe we picked absolute URI for this one case because relative made no sense for something that is supposed to uniquely identify something in a global sense. Thanks, -Doug Sent from my iPhone

…

On Sep 20, 2019, at 12:21 AM, Alan Conway ***@***.***> wrote: This came up while implementing a Go mapping for the cloudevents type system. Here's the definition of URI and URI-reference: URI - Absolute uniform resource identifier. String encoding: Absolute URI as defined in RFC 3986 Section 4.3. URI-reference - Uniform resource identifier reference. String encoding: URI-reference as defined in RFC 3986 Section 4.1. The issue is that all legal values of the URI type are also legal values of the URI-reference type. It is unusual to have a separate type to distinguish absolute URIs from relative-or-absolute URIs. In Go, python and ruby there is a single type (named URI or URL), and it has a test such as u.IsAbs() to distinguish absolute from relative values. The URI type is only used in one place and there the language is ambiguous: dataschema Type: URI Description: Identifies the schema that data adheres to. Incompatible changes to the schema SHOULD be reflected by a different URI. See Versioning of Attributes in the Primer for more information. Constraints: OPTIONAL If present, MUST be a non-empty URI-reference Note the conflicting claims that this attribute is of type URI but MUST be a URI-reference (which may be an invalid URI!) Can we drop the URI type and use URI-reference everywhere instead? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[duglin]

I think we used to use relative URI everywhere and just missed that one when we switched to URI Thanks, -Doug Sent from my iPhone

…

On Sep 20, 2019, at 12:21 AM, Alan Conway ***@***.***> wrote: This came up while implementing a Go mapping for the cloudevents type system. Here's the definition of URI and URI-reference: URI - Absolute uniform resource identifier. String encoding: Absolute URI as defined in RFC 3986 Section 4.3. URI-reference - Uniform resource identifier reference. String encoding: URI-reference as defined in RFC 3986 Section 4.1. The issue is that all legal values of the URI type are also legal values of the URI-reference type. It is unusual to have a separate type to distinguish absolute URIs from relative-or-absolute URIs. In Go, python and ruby there is a single type (named URI or URL), and it has a test such as u.IsAbs() to distinguish absolute from relative values. The URI type is only used in one place and there the language is ambiguous: dataschema Type: URI Description: Identifies the schema that data adheres to. Incompatible changes to the schema SHOULD be reflected by a different URI. See Versioning of Attributes in the Primer for more information. Constraints: OPTIONAL If present, MUST be a non-empty URI-reference Note the conflicting claims that this attribute is of type URI but MUST be a URI-reference (which may be an invalid URI!) Can we drop the URI type and use URI-reference everywhere instead? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[duglin]

For reference, this is the PR that changed it to URI: #475 (comment)

I think it was during a weekly call that someone raised a concern about it being a URI-reference.

[duglin]

@alanconway I fixed the URI-ref -> URI typo as part of #512

[alanconway]

I guess I'll drop this since we allow type-morphing to canonical string and URI < URI-reference < String. I think most programming language mappings and bindings will end up mapping URI and URI-reference to the same native type but that's OK.