Add Computed Accessibility Properties section by spectranaut · Pull Request #2800 · w3c/aria

spectranaut · 2026-05-28T23:02:02Z

🚀 Netlify Preview:
🔄 this PR updates the following sspecs:

core-aam preview — core-aam diff

For the accessible node testing initiative: WICG/aom#203

Link to added section computed accessibility properties for review

netlify · 2026-05-28T23:02:08Z

✅ Deploy Preview for wai-aria ready!

Name	Link
🔨 Latest commit	`dee5a49`
🔍 Latest deploy log	https://app.netlify.com/projects/wai-aria/deploys/6a270d9ff7c3eb0008ad7f53
😎 Deploy Preview	https://deploy-preview-2800--wai-aria.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
🤖 Make changes	Run an agent on this branch

To edit notification comments on pull requests, go to your Netlify project configuration.

github-actions · 2026-06-02T21:18:13Z

🚀 Deployed on https://deploy-preview-2800--wai-aria.netlify.app

cookiecrook · 2026-06-02T23:05:22Z

+              <tr>
+                <th>Computed Accessibility Property</th>
+                <td>
+                  Not mapped, or, referenced element should have `focus: true` ?


does focus here mean focused or focusable?

removed this one for now!

Not removed? "referenced element should have focus: true"

Nm... the comment was not obvious from prior diff I was looking at. Can the value be commented instead of the row?

Oh is that what we discussed? Do you think it's less odd to have it table cell empty rather than the whole row missing? I suppose I don't have a strong opinion.

I recalled ReSpec removed (or can be configured to remove) the empty rows. Am I mistaken?

I just tried this, when the second column is empty, it does not remove the row, and I didn't see any settings related to this in the respec documentation.

I don't think I like the empty column look, so I'd like to stick to commenting out the row. Alternatively, we can write something like "Not specified." or "Intentionally not specified."

spectranaut · 2026-06-03T21:19:44Z

Hi @jcsteh, @cookiecrook, @rahimabdi, @lucasradaelli -- this is ready for review based on the feedback from yesterday's meeting. Here is a link to the added section

I only included "Computed Accessibility Properties" for the following rows:

aria-description
aria-checked
aria-pressed
aria-label
aria-labelledby
aria-level
aria-multiline
aria-orientation
aria-required
aria-roledecription
aria-selected

Let me know what you think of this list, we can remove or add things!

jcsteh · 2026-06-05T05:53:40Z

+            </ol>
+          </aside>
+          <p>
+            When the state or property does not have a default value other than undefined, is not required on a given role, and is missing, undefined, or set to a invalid token, user agents MUST NOT include it the computed accessibility properties. When a state or property has a default value other than undefined and is supported but not required on a given role, user agents MAY include it in the computed accessibility properties.


An alternative (or addition) to the "MAY" case here is something like:

"When a state or property has a default value other than undefined, is supported but not required on a given role and is not mapped on all platforms, user agents MUST NOT include it in the computed accessibility properties."

This would mean that we have a clear rule for aria-required="false", for example: we do not include it in the property bag.

That said, I see you already have "Do not expose required" for aria-required="false". Does that already fit one of the rules here and I'm just missing it?

Hmm.. right. So below, in the tables, when something is not mapped, it seems to make sense to write "Do not expose". I guess, without thinking about it too clearly, it seemed like I didn't have to put the case you suggested into the prose because it would always be covered below.

But I suppose it's fine to be overly pedantic, this is a technical specification, I'm going to include your sentence (and the "may") for now.

Co-authored-by: James Teh <jamie@jantrid.net>

…ccessibilityPropertiesForAccessibilityNode commands https://bugs.webkit.org/show_bug.cgi?id=299508 rdar://161303091 Reviewed by NOBODY (OOPS!). This patch implements two new WebDriver automation commands used to support more robust WPT accessibility testing: - GetAccessibilityPropertiesForElement - GetAccessibilityPropertiesForAccessibilityNode For this initial implementation, test_driver.get_accessibility_properties...() will return an accessibility property bag containing the following computed properties: - accessibilityId - role - label - checked - pressed - children (array of serialized AXIDs) - parent (serialized AXID) As these new WebDriver commands are test-only interfaces, there's been in-depth discussion around how to treat computed AX properties when a property is missing & supported, or present & unsupported; see WICG/aom#203. Some additional resources: - WebDriver PR for the new commands: w3c/webdriver#1960 - ARIA PR for the initial set of computed accessibility properties, and treatment of undefined/missing/unsupported values: w3c/aria#2800. At a high level, the WebDriver client (i.e., Web Platform Tests' testdriver.js) initiates these new commands which is then dispatched to the UI Process (WebAutomationSession). The WebAutomationSession then parses the payload and extracts the protocol-specific element or accessibility node, after which it identifies the target (WebPageProxy) which invokes an asynchronous IPC message to the corresponding Web Process. Note that GetAccessibilityPropertiesForElement uses the same DOM node/frame-based approach as many other WebDriver commands however, GetAccessibilityPropertiesForAXNode does not require frame-based process routing (it retrieves computed accessibility properties using the page's axObjectCache, rather than a specific frame tied to a DOM node); thus, GetAccessibilityPropertiesForAccessibilityNode has a different message signature. Note that to fully support these new WebDriver commands, corresponding WebDriver HTTP endpoints must be implemented in Webdriver for Safari (safaridriver): - session/{session_id}/element/{elId}/accessibilityproperties: Returns accessibility properties for DOM element {elId} - session/{session_id}/accessibility/properties/{accessibilityId}: Returns accessibility properties for accessibility node {accessibilityId} Due to the existence of two AccessibilityProperties protocol object definitions, documentation for each has been updated: - Source/WebKit/UIProcess/Automation/Automation.json — specification conformance and interoperability testing - Source/JavaScriptCore/inspector/protocol/DOM.json — uWeb Inspector DOM element inspection As an unrelated small fix, the return type for GetComputedLabel (WebAutomationSessionProxy IPC message) incorrectly names the first string parameter as "role" instead of "label" so this patch also corrects the tuple to return (label, errorType). * LayoutTests/resources/testdriver-vendor.js: (window.test_driver_internal.get_accessibility_properties_for_element): (window.test_driver_internal.get_accessibility_properties_for_accessibility_node): * Source/JavaScriptCore/inspector/protocol/DOM.json * Source/WebKit/UIProcess/Automation/Automation.json: * Source/WebKit/UIProcess/Automation/WebAutomationSession.cpp: (Ref<JSON::ArrayOf<String>> buildArrayForAXChildren) (WebKit::WebAutomationSession::getAccessibilityPropertiesForElement): (WebKit::WebAutomationSession::getAccessibilityPropertiesForAccessibilityNode): * Source/WebKit/UIProcess/Automation/WebAutomationSession.h: * Source/WebKit/UIProcess/WebPageProxy.cpp: * Source/WebKit/WebProcess/Automation/WebAutomationSessionProxy.cpp: (WebKit::WebAutomationSessionProxy::getAccessibilityPropertiesForElement): (WebKit::WebAutomationSessionProxy::getAccessibilityPropertiesForAccessibilityNode): (WebKit::WebAutomationSessionProxy::getAccessibilityObjectForAXNode) (WebAutomationSessionProxy::ComputedAXProperties getComputedAccessibilityProperties) * Source/WebKit/WebProcess/Automation/WebAutomationSessionProxy.h: * Source/WebKit/WebProcess/Automation/WebAutomationSessionProxy.messages.in:

…ccessibilityPropertiesForAccessibilityNode commands https://bugs.webkit.org/show_bug.cgi?id=299508 rdar://161303091 Reviewed by NOBODY (OOPS!). This patch implements two new WebDriver automation commands used to support more robust WPT accessibility testing: - GetAccessibilityPropertiesForElement - GetAccessibilityPropertiesForAccessibilityNode For this initial implementation, test_driver.get_accessibility_properties...() will return an accessibility property bag containing the following computed properties: - accessibilityId - role - label - checked - pressed - children (array of serialized AXIDs) - parent (serialized AXID) As these new WebDriver commands are test-only interfaces, there's been in-depth discussion around how to treat computed AX properties when a property is missing and supported, or present & unsupported; see WICG/aom#203. Some additional resources: - WebDriver PR for the new commands: w3c/webdriver#1960 - ARIA PR for the initial set of computed accessibility properties, and treatment of undefined/missing/unsupported values: w3c/aria#2800. At a high level, the WebDriver client (i.e., Web Platform Tests' testdriver.js) initiates these new commands which is then dispatched to the UI Process (WebAutomationSession). The WebAutomationSession then parses the payload and extracts the protocol-specific element or accessibility node, after which it identifies the target (WebPageProxy) which invokes an asynchronous IPC message to the corresponding Web Process. Note that GetAccessibilityPropertiesForElement uses the same DOM node/frame-based approach as many other WebDriver commands however, GetAccessibilityPropertiesForAXNode does not require frame-based process routing (it retrieves computed accessibility properties using the page's axObjectCache, rather than a specific frame tied to a DOM node); thus, GetAccessibilityPropertiesForAccessibilityNode has a different message signature. Note that to fully support these new WebDriver commands, corresponding WebDriver HTTP endpoints must be implemented in Webdriver for Safari (safaridriver): - session/{session_id}/element/{elId}/accessibilityproperties: Returns accessibility properties for DOM element {elId} - session/{session_id}/accessibility/properties/{accessibilityId}: Returns accessibility properties for accessibility node {accessibilityId} Due to the existence of two AccessibilityProperties protocol object definitions, documentation for each has been updated: - Source/WebKit/UIProcess/Automation/Automation.json — specification conformance and interoperability testing - Source/JavaScriptCore/inspector/protocol/DOM.json — Web Inspector DOM element inspection As an unrelated small fix, the return type for GetComputedLabel (WebAutomationSessionProxy IPC message) incorrectly names the first string parameter as "role" instead of "label" so this patch also corrects the tuple to return (label, errorType). * LayoutTests/resources/testdriver-vendor.js: (window.test_driver_internal.get_accessibility_properties_for_element): (window.test_driver_internal.get_accessibility_properties_for_accessibility_node): * Source/JavaScriptCore/inspector/protocol/DOM.json * Source/WebKit/UIProcess/Automation/Automation.json: * Source/WebKit/UIProcess/Automation/WebAutomationSession.cpp: (Ref<JSON::ArrayOf<String>> buildArrayForAXChildren) (WebKit::WebAutomationSession::getAccessibilityPropertiesForElement): (WebKit::WebAutomationSession::getAccessibilityPropertiesForAccessibilityNode): * Source/WebKit/UIProcess/Automation/WebAutomationSession.h: * Source/WebKit/UIProcess/WebPageProxy.cpp: * Source/WebKit/WebProcess/Automation/WebAutomationSessionProxy.cpp: (WebKit::WebAutomationSessionProxy::getAccessibilityPropertiesForElement): (WebKit::WebAutomationSessionProxy::getAccessibilityPropertiesForAccessibilityNode): (WebKit::WebAutomationSessionProxy::getAccessibilityObjectForAXNode) (WebAutomationSessionProxy::ComputedAXProperties getComputedAccessibilityProperties) * Source/WebKit/WebProcess/Automation/WebAutomationSessionProxy.h: * Source/WebKit/WebProcess/Automation/WebAutomationSessionProxy.messages.in:

…ccessibilityPropertiesForAccessibilityNode commands https://bugs.webkit.org/show_bug.cgi?id=299508 rdar://161303091 Reviewed by NOBODY (OOPS!). This patch implements two new WebDriver automation commands used to support more robust WPT accessibility testing: - GetAccessibilityPropertiesForElement - GetAccessibilityPropertiesForAccessibilityNode For this initial implementation, test_driver.get_accessibility_properties...() will return an accessibility property bag containing the following computed properties: - accessibilityNodeId - role - label - checked - pressed - children (array of serialized AXIDs) - parent (serialized AXID) As these new WebDriver commands are test-only interfaces, there's been in-depth discussion around how to treat computed AX properties when a property is missing and supported, or present & unsupported; see WICG/aom#203. Some additional resources: - WebDriver PR for the new commands: w3c/webdriver#1960 - ARIA PR for the initial set of computed accessibility properties, and treatment of undefined/missing/unsupported values: w3c/aria#2800. At a high level, the WebDriver client (i.e., Web Platform Tests' testdriver.js) initiates these new commands which is then dispatched to the UI Process (WebAutomationSession). The WebAutomationSession then parses the payload and extracts the protocol-specific element or accessibility node, after which it identifies the target (WebPageProxy) which invokes an asynchronous IPC message to the corresponding Web Process. Note that GetAccessibilityPropertiesForElement uses the same DOM node/frame-based approach as many other WebDriver commands however, GetAccessibilityPropertiesForAXNode does not require frame-based process routing (it retrieves computed accessibility properties using the page's axObjectCache, rather than a specific frame tied to a DOM node); thus, GetAccessibilityPropertiesForAccessibilityNode has a different message signature. Note that to fully support these new WebDriver commands, corresponding WebDriver HTTP endpoints must be implemented in Webdriver for Safari (safaridriver): - session/{session_id}/element/{elId}/accessibilityproperties: Returns accessibility properties for DOM element {elId} - session/{session_id}/accessibility/properties/{accessibilityId}: Returns accessibility properties for accessibility node {accessibilityId} Due to the existence of two AccessibilityProperties protocol object definitions, documentation for each has been updated: - Source/WebKit/UIProcess/Automation/Automation.json — specification conformance and interoperability testing - Source/JavaScriptCore/inspector/protocol/DOM.json — Web Inspector DOM element inspection As an unrelated small fix, the return type for GetComputedLabel (WebAutomationSessionProxy IPC message) incorrectly names the first string parameter as "role" instead of "label" so this patch also corrects the tuple to return (label, errorType). * LayoutTests/resources/testdriver-vendor.js: (window.test_driver_internal.get_accessibility_properties_for_element): (window.test_driver_internal.get_accessibility_properties_for_accessibility_node): * Source/JavaScriptCore/inspector/protocol/DOM.json * Source/WebKit/UIProcess/Automation/Automation.json: * Source/WebKit/UIProcess/Automation/WebAutomationSession.cpp: (Ref<JSON::ArrayOf<String>> buildArrayForAXChildren) (WebKit::WebAutomationSession::getAccessibilityPropertiesForElement): (WebKit::WebAutomationSession::getAccessibilityPropertiesForAccessibilityNode): * Source/WebKit/UIProcess/Automation/WebAutomationSession.h: * Source/WebKit/UIProcess/WebPageProxy.cpp: * Source/WebKit/WebProcess/Automation/WebAutomationSessionProxy.cpp: (WebKit::WebAutomationSessionProxy::getAccessibilityPropertiesForElement): (WebKit::WebAutomationSessionProxy::getAccessibilityPropertiesForAccessibilityNode): (WebKit::WebAutomationSessionProxy::getAccessibilityObjectForAXNode) (WebAutomationSessionProxy::ComputedAXProperties getComputedAccessibilityProperties) * Source/WebKit/WebProcess/Automation/WebAutomationSessionProxy.h: * Source/WebKit/WebProcess/Automation/WebAutomationSessionProxy.messages.in:

cookiecrook · 2026-06-08T16:33:45Z

Hi @jcsteh, @cookiecrook, @rahimabdi, @lucasradaelli -- this is ready for review based on the feedback from yesterday's meeting. Here is a link to the added section

I only included "Computed Accessibility Properties" for the following rows:

aria-description

aria-checked

aria-pressed

aria-label

aria-labelledby

aria-level

aria-multiline

aria-orientation

aria-required

aria-roledecription

aria-selected

Let me know what you think of this list, we can remove or add things!

Thoughts individually.

level: Should we add a clarifying note on some of the ones that have known correlate reconciliation with the host languages? For example, <h3> gets a level of 3 event though the ariaLevel IDL reflection should be… null for now unless it changes to Integer, and then would be 0 I think?
required: note to cross-ref html:required?
multiline: This is another we should reconsider. ARIA's choice to use textbox and textbox multiline instead of the existing input type=text and textarea was weird, especially in the context of contenteditable. IIRC, this was one I raised as a problem early as an ARIA editor, but [corp rep] didn't want to pay [AT dev] to change the implementation, so we left it in the spec. Had I been a more experienced editor at the time, I may have pushed back more firmly.
selected: We should discuss the name of this one in the context of ambiguity. "selected" can be confused with text range selection, so maybe we want to disambiguate the name further? Also, some implementations don't make as much of a distinction between checked, pressed, and selected in the API... I'm not opposed to this one as-is. Jut trying to anticipate future problems.

The rest look good to me. Thanks for trimming the big list down and proceeding more deliberately.

Co-authored-by: Lola <lolaodelola@gmail.com>

spectranaut · 2026-06-08T18:15:41Z

level: Should we add a clarifying note on some of the ones that have known correlate reconciliation with the host languages? For example, <h3> gets a level of 3 event though the ariaLevel IDL reflection should be… null for now unless it changes to Integer, and then would be 0 I think?

I don't think we need to make this note in the table... however, we could say something about this in the definition of computed accessibility properties, perhaps adding this after the first sentence (first sentence included):

The computed accessibility properties of an element are a set of platform independent key/value pairs of strings that represent the states and attributes of any accessible object in the accessibility tree. Many different things affect the value of the key in the accessibility tree, including but not limited to: defaults values for the state or attribute, default values when used on roles, host language states and attributes, and whether the host language overrides the ARIA values. These values do not always align with IDL reflection of the aria state or attribute.

No objections to keeping this in the list, right?

required: note to cross-ref html:required?

I don't think we should cross ref to html required, because html required already references Core-AAM: https://w3c.github.io/html-aam/#att-required

No objections to keeping this in the list, right?

multiline: This is another we should reconsider. ARIA's choice to use textbox and textbox multiline instead of the existing input type=text and textarea was weird, especially in the context of contenteditable. IIRC, this was one I raised as a problem early as an ARIA editor, but [corp rep] didn't want to pay [AT dev] to change the implementation, so we left it in the spec. Had I been a more experienced editor at the time, I may have pushed back more firmly.

Interesting history :) I'm fine with leaving this one out. But I think that means I need to update the example that uses it....

selected: We should discuss the name of this one in the context of ambiguity. "selected" can be confused with text range selection, so maybe we want to disambiguate the name further? Also, some implementations don't make as much of a distinction between checked, pressed, and selected in the API... I'm not opposed to this one as-is. Jut trying to anticipate future problems.

I would like to leave this one it, but I'm totally fine if you think we should drop it for now. Chrome does differentiate between select and checked/pressed, it's just checked/pressed that are combined, and we are already surfacing that information. Since we are already sorting out the checked/pressed problem, and we might as well deal with all three?

As for the name, if we wanted to do something different then selected... I'm not sure, the whole web does have this name collision, not just accessibility... I wonder how it's dealt with in other contexts? I feel like if we want to surface selected text we should say "selectedText".

Add Computed Accessibility Properties section

61dd7d1

spectranaut added the spec:core-aam label May 28, 2026

Clean up

f810b86

github-actions Bot temporarily deployed to pull request June 2, 2026 21:18 Inactive

fix example

6a93c14

github-actions Bot temporarily deployed to pull request June 2, 2026 21:48 Inactive

cookiecrook reviewed Jun 2, 2026

View reviewed changes

Comment thread core-aam/index.html

cookiecrook reviewed Jun 2, 2026

View reviewed changes

Comment thread core-aam/index.html

cookiecrook reviewed Jun 2, 2026

View reviewed changes

Comment thread core-aam/index.html

Update prose

1104e96

github-actions Bot temporarily deployed to pull request June 3, 2026 18:19 Inactive

spectranaut added 2 commits June 3, 2026 11:37

comment out most rows

d4b7f9d

Fix examples

4e11d58

spectranaut marked this pull request as ready for review June 3, 2026 21:12

breakup paragraph

6ae1151

spectranaut requested review from ChrisC, jcsteh, lolaodelola, lucasradaelli and rahimabdi June 3, 2026 21:13

github-actions Bot temporarily deployed to pull request June 3, 2026 21:15 Inactive

Minor fix

4d955f4

github-actions Bot temporarily deployed to pull request June 3, 2026 21:55 Inactive

spectranaut requested a review from cookiecrook June 4, 2026 15:36

cyns self-requested a review June 4, 2026 17:10

jcsteh reviewed Jun 5, 2026

View reviewed changes

Apply suggestions from code review

85c2a43

Co-authored-by: James Teh <jamie@jantrid.net>

github-actions Bot temporarily deployed to pull request June 5, 2026 18:47 Inactive

spectranaut added 3 commits June 5, 2026 14:15

Fix line endings

0f60f03

Code review comments

7bbc288

Add not mapped description

dc9d5da

github-actions Bot temporarily deployed to pull request June 5, 2026 21:18 Inactive

rahimabdi mentioned this pull request Jun 6, 2026

WebDriver: Add support for GetAccessibilityPropertiesForElement, GetAccessibilityPropertiesForAccessibilityNode commands WebKit/WebKit#65058

Open

lolaodelola reviewed Jun 8, 2026

View reviewed changes

Comment thread core-aam/index.html Outdated

Apply suggestion from @lolaodelola

05ce4f9

Co-authored-by: Lola <lolaodelola@gmail.com>

github-actions Bot temporarily deployed to pull request June 8, 2026 16:39 Inactive

Update examples to only used allowed list

dee5a49

github-actions Bot deployed to pull request June 8, 2026 18:46 View deployment

spectranaut requested review from jcsteh and lolaodelola June 11, 2026 23:27

spectranaut mentioned this pull request Jun 11, 2026

Meeting: July 1, 2026, @ 9 AM PST web-platform-tests/interop-accessibility#236

Open

Conversation

spectranaut commented May 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

netlify Bot commented May 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Deploy Preview for wai-aria ready!

Uh oh!

github-actions Bot commented Jun 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

spectranaut commented Jun 3, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

cookiecrook commented Jun 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

spectranaut commented Jun 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

spectranaut commented May 28, 2026 •

edited

Loading

netlify Bot commented May 28, 2026 •

edited

Loading

github-actions Bot commented Jun 2, 2026 •

edited

Loading

cookiecrook commented Jun 8, 2026 •

edited

Loading

spectranaut commented Jun 8, 2026 •

edited

Loading