Repository navigation
Upcoming WHATNOT meeting on 2026-03-19 #12275
Description
Activity
So far up to date issues to be discussed on the meeting (issues/PRs with Agenda+ that weren't discussed last week):
Minutes:
- Add Shadow Root Slot Assignment: Keith Cirkel introduced a Pull Request (PR) to add the property
shadow root slot assignmenttoattach shadow, which is currently the only property that cannot be used declaratively. This feature enables a declarative shadow root with manual slot assignment, though assigned slots will not exist untilslot assignis called. The participants agreed that the feature is a small lift and does not require the stages process, and Keith Cirkel requested reviews for the specification and tests. - Discussion on HTML Canvas for Stage 2: Philip Jägenstedt gave an update on the progress of HTML canvas, noting they are working to move it to Stage 2 and presented several demos to illustrate its functionality. The demos showed that HTML elements within the canvas, such as text labels, can be selectable, searchable via find-in-page, and accessible via tabbing, which was a change from previous understandings. Philip Jägenstedt clarified that the implementation currently supports find-in-page highlights, and the spec and explainer will be updated to reflect this capability.
- Demo and Hit Testing in HTML Canvas: Philip Jägenstedt demonstrated a complex WebGPU slider example where an input element and a text div are composited on the canvas, showing visual effects that are impossible with standard CSS. Jake Archibald raised questions regarding hit testing for non-flat surfaces, such as curved or repeated text, noting that selection in the demo worked incorrectly without the
inertattribute. Philip Jägenstedt suggested that for complicated effects, applying theinertattribute may be the appropriate solution, though it was acknowledged that providing a low-level transform mapping from canvas space to element coordinates could potentially address complex selection issues. - Complexity of Hit Testing and Event Handling: Emilio Cobos Álvarez and Noam Rosenthal noted that complex hit testing, especially for elements drawn multiple times, is difficult to implement manually and that existing designs were cautious about adding complex callbacks for event mapping. There was a discussion about text selection being hard to implement manually and the challenges of managing trusted events, leading to an agreement to revisit how these complex cases are handled. Philip Jägenstedt confirmed they would take the feedback regarding non-linear text selection and the perceived convenience of using the
inertattribute back to the team. - Hit Testing and Z-Indexing: Jake Archibald pointed out a rough edge in the hit-testing API where, while a transform is provided to position elements correctly, the developer still needs to manually manage the correct Z-indexing for overlapping elements. Philip Jägenstedt confirmed that the order of elements in the DOM should match the painting order, or developers could attempt to use Z-index on child elements, as the API does not automatically manage the rendering order.
- Accessibility, Tooltips, and Paint Event Timing: Philip Jägenstedt provided updates on several concerns raised previously, stating that tooltips (from the title attribute) would paint on top of the canvas and not be accessible within the canvas. Initial screen reader testing for accessibility seemed positive, and the role hierarchy did not appear to be an issue, though adjustments could be made if needed. They also confirmed that the timing of the
paintevent has been added to the spec, occurring right afterrequestAnimationFrameand immediately after element snapshots are updated. - Paint Event Timing and Spec Placement: Emilio Cobos Álvarez raised concerns that placing the
paintevent after Intersection Observer steps could lead to strange behavior, as any changes from the arbitrary script run in thepaintevent would not be reflected in the Intersection Observer. They suggested integrating thepaintevent earlier, possibly within the Resize Observer loop, which is where similar synchronous script reactions and layout-changing events are handled. Noam Rosenthal noted this would require additional design to prevent multiple paint events from firing in a single rendering loop. - Concerns with Readback and Privacy: The discussion transitioned to the issue of readback and its privacy implications, as it makes it easy to introduce new observable behaviors into the platform. Philip Jägenstedt noted that due to WebGL and WebGPU, preventing readback entirely is difficult because shaders can introduce feedback loops, so the current design does not assume pixels are tainted. Jake Archibald argued that readback is necessary for legitimate use cases, such as compositing HTML elements to generate video files.
- Privacy Preserving Paint and System Colors: Luke inquired about the behavior of "privacy preserving paint" regarding default OS rendering, confirming that it uses light mode by default. The discussion centered on whether system information, such as high contrast mode or system colors, would be leaked through the canvas, with Luke suggesting using
force color adjust noneto prevent reading back Windows themes. Emilio Cobos Álvarez and Philip Jägenstedt noted that system colors might already be readable through other means, but the goal is to prevent the revelation of sensitive information. - Font Enumeration and Form Control Concerns: Keith Cirkel expressed concern over form control representation on Linux due to numerous desktop environments and themes, suggesting it lacks a canonical representation. They also raised a potential concern that allowing arbitrary text rendering could enable font enumeration attacks by reading back pixel data of rendered fonts or emojis. Philip Jägenstedt asked if this was different from existing canvas text rendering APIs, and the group concluded that this is a known area of concern that requires attention.
- Toolbar Element Proposal (Stage 1): Luke introduced a proposal for a new HTML
<toolbar>element to simplify creating a common pattern for web applications. The proposed element aims to provide built-in arrow key navigation and similar behaviors tofieldset, such as being able to disable the entire toolbar. The proposal received strong support from both Google and Mozilla participants for Stage 1 approval, signaling interest in solving this problem space. - Update on Out-of-Order Streaming: Noam Rosenthal provided an update on the out-of-order streaming feature, confirming that the
markerattribute has been removed based on previous discussions, as it did not solve any practical cross-site scripting (XSS) problems. The feature is at Stage 2 and is near ready for a PR, and the participants were invited to review the explainer and three associated PRs that cover parsing, DOM APIs, and template content.
- Add Shadow Root Slot Assignment: Keith Cirkel introduced a Pull Request (PR) to add the property
What is the issue with the HTML Standard?
Last week we held our weekly triage call (#12244) and the meeting notes will be posted there shortly. The next one is scheduled for March 19, 2026, at 1AM PDT. Note that this is 1 week later in the APAC-EMEA friendly time.
PLEASE ALSO NOTE THAT DUE TO THE TIME OF YEAR, THIS WILL BE ONE HOUR EARLIER FOR EUROPEAN PARTICIPANTS!
People interested in attending the next call who need the invite please respond here or reach out privately to @cwilso, or the editors. Please tag issues and PRs for the next call using agenda+ in all WHATWG repositories across issues and pull requests and we would like to invite anyone that can contribute to join us.