Drop export to "role" attribute

[marcoscaceres]

The ARIA spec now defines and exports the role attribute:
w3c/aria@5aa4442

So it might be prudent for HTML to no longer export it, and reference ARIA (HTML already references ARIA, but it should probably link directly to the ARIA spec for the definition).

Role is also defined by SVG2 (I've filed a bug there too):
https://svgwg.org/svg2-draft/struct.html#roleattribute

What would be amazing is if the HTML editors' could take a look at the role definition in Aria and see if it needs fixing up or additional things, so it's defined in the same manner as other attributes in HTML 🙏.

[annevk]

That commit against w3c/aria makes it seem that HTML needs to define a role attribute ("host language will provide") that meets the requirements given there. It doesn't seem to actually define the attribute itself or is that done elsewhere?

[marcoscaceres]

It doesn't seem to actually define the attribute itself or is that done elsewhere?

It defines it in as far as it <dfn>s it. But agree that it's not amazingly defined. I'm kind opening up for discussion who should own it, but I think it's confusing that we have SVG + HTML both defining it.

The HTML spec says:

The role attribute is defined in Accessible Rich Internet Applications (ARIA), as are the following roles: [ARIA]

Which leaves me feeling like it's underspecified.

[annevk]

This also came up in #6810 (comment).

Apart from who defines role (I guess we're ignoring the other ARIA attributes for now) there's a question as to where it works:

• All elements, any namespace. I.e., is it superglobal and ought to be defined in DOM?
• Global attribute for HTML, SVG, and MathML.
• Something else?

And then separately there's the question of who defines it. We don't have a good story at the moment for attributes (or elements) that are largely shared between HTML, SVG, and MathML.

[jyasskin]

ARIA dropped its <dfn> for the role attribute in 2022 (w3c/aria@a56ae18), so I've filed w3c/aria#2145 for them to figure out where the attribute should actually be defined. As I said there, I don't have a strong opinion, but other HTML folks might.