posinset/setsize for radiogroups

[slender789]

Given:

<input name=test type=radio>radio1</input>
<input name=test type=radio>radio2</input>
<input name=test type=radio>radio3</input>
<input name=test type=radio>radio4</input>
<input name=test type=radio>radio5</input>
<input name=test type=radio>radio6</input>

html AAM says
"With aria-setsize value reflecting number of type=radio input elements within the radio button group and aria-posinset value reflecting the elements position within the radio button group."

The radio button group is defined by the html standard:

• The input element b's type attribute is in the Radio Button state.
• Either a and b have the same form owner, or they both have no form owner.
• Both a and b are in the same tree.
• They both have a name attribute, their name attributes are not empty, and the value of a's name attribute equals the value of b's name attribute.

Now, given:

<div role="radiogroup">
  <input name=test type=radio>radio1</input>
  <input name=test type=radio>radio2</input>
  <input name=test type=radio>radio3</input>
</div>
<div role="radiogroup">
  <input name=test type=radio>radio4</input>
  <input name=test type=radio>radio5</input>
  <input name=test type=radio>radio6</input>
</div>

neither html AAM nor core AAM give any guidance.

Should the role=radiogroup alter the computation of posinset/setsize by treating them as two isolated radiogroups?

[JAWS-test]

ARIA roles override HTML roles, so in your second example, role=radiogroup should take precedence when determining the radio button group.

[scottaohara]

What’s the use case for this?

Treating them as separate seems like a bad idea, since they’re all associated with each other.

But also, going off the example markup alone, without accname the radiogroups are likely to be ignored anyway. So….

[slender789]

@JAWS-test about aria overriding html, my concern is radio button groups are not roles in html but instead a concept in html defined by the listed criteria above.

@scottaohara (context)[https://g-issues.chromium.org/issues/40919211]
I also agree that if we were to treat the example as two separate, the developer would have to override native radio button keyboard behaviors.

What I think we're looking for is the documentation stating the precise means by which (radiogroup)[https://www.w3.org/TR/core-aamz-1.2/#role-map-radiogroup] overrides the computation of radio button group, e.g. role radiogroup overrides the posinset/setsize computation for html radio button groups.

[slender789]

.