Problem
Components 4 has exhaustive Chromium evidence across four renderers, 261 stories, light/dark appearances, interactions, and axe. That is strong bounded automation, but it is not Firefox/WebKit coverage or manual assistive-technology certification.
Direction
Add a layered evidence model rather than multiplying every expensive case immediately:
- Keep the complete Chromium matrix as the exhaustive baseline.
- Run a representative, explicitly documented cross-browser interaction profile in Firefox and WebKit covering forms, overlays, focus restoration, keyboard collections, native reset, hydration, and the three adapters.
- Define a manual AT protocol for at least VoiceOver/Safari and NVDA/Firefox or the project-supported equivalents.
- Store evidence by package, renderer, browser/AT version, operating system, profile, and commit.
Any sampling must be visible and justified; never describe it as universal certification.
Acceptance criteria
- Browser support policy names exact engines/versions and selection rationale.
- Cross-browser profile passes for built-in, MUI, PrimeReact 11, and PrimeReact 10.
- Manual protocol covers accessible names, focus order/restore, live regions, validation, selection, switch/checkbox/radio, dialogs, dropdowns, and date entry.
- Results and known gaps are published as bounded evidence in adapter
CONFORMANCE.md files.
- CI cost and flake policy are documented.
Problem
Components 4 has exhaustive Chromium evidence across four renderers, 261 stories, light/dark appearances, interactions, and axe. That is strong bounded automation, but it is not Firefox/WebKit coverage or manual assistive-technology certification.
Direction
Add a layered evidence model rather than multiplying every expensive case immediately:
Any sampling must be visible and justified; never describe it as universal certification.
Acceptance criteria
CONFORMANCE.mdfiles.