Add more global attributes from HTML

[fred-wang]

The global attributes are defined here:
https://mathml-refresh.github.io/mathml-core/#global-attributes

Some of them are taken from HTML:
https://mathml-refresh.github.io/mathml-core/#attributes-common-to-html-and-mathml-elements

After discussions in w3c/mathml-core#129, we agreed that we want to add some IDL for all MathML elements:
https://mathml-refresh.github.io/mathml-core/#dom-and-javascript

(1) GlobalEventHandlers, DocumentAndElementEventHandlers would mean adding all the onsomething event attributes:
https://html.spec.whatwg.org/multipage/webappapis.html#globaleventhandlers
https://html.spec.whatwg.org/multipage/webappapis.html#documentandelementeventhandlers
Rob mentioned that SVG has something to avoid to explicitly mention them:
https://svgwg.org/svg2-draft/interact.html#EventAttributes

(Side comment: do we really want to keep the non-normative https://mathml-refresh.github.io/mathml-core/#relaxng ? Now that we are relying a lot on HTML and CSS definitions (in particular for attribute values) instead of having a standalone document, it seems a bit tedious to keep things in sync. Maybe the RelaxNG schema should be moved in a different repository?)

(2) HTMLOrSVGElement would add this:
https://html.spec.whatwg.org/multipage/dom.html#htmlorsvgelement

'data-' attributes:
https://html.spec.whatwg.org/multipage/dom.html#attr-data-

nonce attribute:
https://html.spec.whatwg.org/multipage/urls-and-fetching.html#nonce-attributes

tabIndex attribute:
https://html.spec.whatwg.org/multipage/interaction.html#attr-tabindex

(3) ElementCSSInlineStyle has a style attribute
https://www.w3.org/TR/cssom-1/#elementcssinlinestyle
which is already here:
https://mathml-refresh.github.io/mathml-core/#style
so nothing to add.

In several places, the HTML spec mentions that IDL attributes reflect content attributes:
https://html.spec.whatwg.org/multipage/common-dom-interfaces.html#reflect
Maybe we need to mention that too.

[bkardell]

Hmm, somehow until now I didn't catch that math elements can have an href. would it have to link to/specify/mixin idl like https://html.spec.whatwg.org/#api-for-a-and-area-elements ?

Which annotation attributes from https://html.spec.whatwg.org/#linkTypes does it support (I assume the same as a/area?) and that then all of the things about building requests/CORs etc are exactly the same too?

This seems kind of considerably underspecified, do you know how those that support linking (Safari definitely does) handles those questions?

[bkardell]

ahh, I assume all of the 'location' things in CSS would treat this like links too? Basically all of these:

https://www.w3.org/TR/selectors-4/#location

[fred-wang]

ahh, I assume all of the 'location' things in CSS would treat this like links too? Basically all of these:

https://www.w3.org/TR/selectors-4/#location

Yes, I think so. The spec actually has such a selector in the UA stylesheet https://mathml-refresh.github.io/mathml-core/#dfn-href

[fred-wang]

Hmm, somehow until now I didn't catch that math elements can have an href. would it have to link to/specify/mixin idl like https://html.spec.whatwg.org/#api-for-a-and-area-elements ?

That seems sensible to me. IIRC, the MathML href implementations in Gecko/WebKit/Chromium basically rely on or copy what is done for the a element, so I expect they would behave the same.

(BTW, as a side comment MathML 2 used to rely on xlink for hyperlinks. I think SVG does it too)

[NSoiffer]

The UA Style sheet in the spec uses

:any-link {
  color: blue;
  text-decoration: none;
}

While I understand what is desired, isn't that too broad as it affects all links, not just those in MathML?

FYI: href was added in MathML 3 because XML and its linking doesn't really fly in HTML. It was added to all elements because we thought people might want to link out to a definition of some component. I think it is useful mainly on the math element and on the leaf elements, at least in terms of linking out to definitions. I'm sure that others can find cases where an mrow or msup or whatever is a better point for adding a definition link though. You can still add xml:xlink to any element if you are in an XML context.

[fred-wang]

The UA Style sheet in the spec uses

:any-link {
  color: blue;
  text-decoration: none;
}

While I understand what is desired, isn't that too broad as it affects all links, not just those in MathML?

The complete stylesheet is https://mathml-refresh.github.io/mathml-core/#user-agent-stylesheet

And this rule forces all rules to apply only to element in the MathML namespace:

@namespace url(http://www.w3.org/1998/Math/MathML);

[ExE-Boss]

Well, MathML doesn’t currently support event handler attributes at all, and I’d like it to stay that way, as event handler attributes are mothballed in favour of addEventListener, which is currently the only way to add event listeners to MathML elements.

---

TL;DR: I object to adding on<something> event handler attributes to MathML elements, since they’ve never been supported for them in the past.

[bkardell]

@ExE-Boss it seems very strange for mathml to be special in this regard. The default position should be, in my mind, that working like everything else should be the default goal and that being special should be only if we can give a really good justification for why it has to be, no? Is there some very strong argument as to why it should continue to be different really?

[ExE-Boss]

Because event handler attributes are deprecated, and a legacy mistake, IMO.

[bkardell]

They aren't deprecated. They have even expanded in the not distant past whatwg/html@940a2d4

[fred-wang]

I think @davidcarlisle raised the issue during one meeting. The conclusion was that as long as HTML supports them, it is not a big extra code to support this for MathML and as Brian said it makes things more consistent with SVG/HTML. Also, it's sometimes more convenient (e.g. for quick testing) to use on[event] attribute than to register events via addEventListener. I believe people still use onload a lot for example.

[fred-wang]

Remaining work was done by

w3c/mathml-core@c2d770c

and we have several tests for that. So I think we can close this.