implement headingoffset & headingreset attributes

[keithamus]

This specifies the headingoffset and headingstart attributes as proposed in #5033.

Closes #5033

• At least two implementers are interested (and none opposed):
  • Chromium: Actively prototyping.
  • Mozilla: (positive) Headingoffset & Headingreset attributes mozilla/standards-positions#1263 (comment)
  • WebKit: (no signal) Headingoffset & Headingreset attributes WebKit/standards-positions#523
• Tests are written and can be reviewed and commented upon at:
  • [html/semantics/sectioning] Test headingoffset & headingreset. web-platform-tests/wpt#54294
• Implementation bugs are filed:
  • Chromium: https://issues.chromium.org/issues/333628468
  • Gecko: https://bugzilla.mozilla.org/show_bug.cgi?id=1974383
  • WebKit: https://bugs.webkit.org/show_bug.cgi?id=295092
• Corresponding HTML AAM & ARIA in HTML issues & PRs: [html-aam] Set heading level based on HTML computed heading level w3c/aria#2598
• MDN issue is filed: Headingoffset mdn/mdn#711
• The top of this comment includes a clear commit message to use.

(See WHATWG Working Mode: Changes for more details.)

---

/dom.html ( diff )
/index.html ( diff )
/infrastructure.html ( diff )
/rendering.html ( diff )
/sections.html ( diff )

[scottaohara]

thank you for working on this.

aside from the comments i left in my quick review, does it make sense to explicitly call out as a note that aria-level is not considered/taken into account with these attributes?

from re-reading the issue this stems from, i'm assuming people are likely thinking that if someone declares an aria-level on a heading, the value of that attribute is what the heading level will be exposed as, yes?

Or to be very clear, do people agree with the following:

<div headingoffset=1>
  <h1 aria-level=4> level 4 </h1>
  <h1> level 2 </h1>
  <h1 headingoffset=20 headingreset aria-level=3><!-- 3 --></h1>
</div>

if HTML doesn't have a note for this, then we can get it into html aam, aria in html, or both - and i'd assume MDN author guidance.

[domenic]

I haven't looked at the PR text itself yet, but I think such a note/example would be valuable.

[keithamus]

I've added a note mentioning aria-level as well as an example demonstrating their combinations.

[annevk]

Since this is not marked as draft and is still getting updates I'm going to make it extra clear that I consider #5033 (comment) to be blocking concerns.

[jakearchibald]

Heads-up: The previews are stuck on some earlier version.

[keithamus]

How can I kick the build previews in OP? They don't seem to be generating.

[jakearchibald]

They're usually automatic, so something has gone wrong.

[jakearchibald]

Btw, I think this is great, and a much-needed addition to the platform.

[annevk]

Not both Reflect and ReflectRange. The latter implies the former.

[keithamus]

I cribbed from other examples such as TableColElement which has both Reflect & ReflectRange. Shall I file an issue for cleaning those up?

[annevk]

Oh wait. I guess you need to leave this as-is and if this lands first, then #11518 needs to remove it. (And vice versa. :-()

[annevk]

Very close now I think.

[annevk]

@keithamus I just noticed you removed all the style sheet changes. While I don't think we have to change that example, I do think we have to change the user agent style sheet as part of this change. Do we have test coverage for the user agent style sheet being changed as part of this change?

[annevk]

I'm happy with this. I'm not happy with the variant discussed on WHATWG Chat whereby dialog in its model state can end up changing heading levels. Heading levels should stay consistent regardless of the state of controls.