There's more to Performance than meets the Eye
summary
Shows measurable effects: screen readers rely on the browser’s accessibility tree, and Chrome mitigates latency by caching that tree into the application process, while Edge forces slow OS API roundtrips so out‑of‑process queries run an estimated 12–50× slower. Firefox’s multi‑process design added delays that land much closer to Chrome after Firefox 60. The virtual model still provides features (like two‑marker selection) that raw API queries do not, so dropping it requires API and UX redesign.
AI-written summary · about
chapters (28)
- 0:00Intro
- 0:49Web performance
- 1:07Browsers
- 1:29Screen readers
- 2:23Screen reader demo: text
- 2:53Screen reader demo: language
- 3:25In the late 90s Windows
- 4:13Screen reader demo: headings
- 4:46Screen reader demo buttons
- 8:05The accessibility tree details element
- 9:48Single process model
- 10:23Communication architectures
- 12:55Multi-process Chrome & firefox
- 13:18Chrome initial approach
- 13:54Chrome current approach
- 14:16Chrome accessibility performance
- 14:34Chrome screen reader TTI
- 16:03Firefox approach
- 16:53Firefox accessibility performance
- 17:53Multi-process Edge
- 18:36Edge accessibility performance
- 19:03Edge: Jaws & NVDA
- 19:44Edge: Narrator
- 20:28Performance drop the virtual model?
- 20:58Screen reader demo: selecting content
- 21:42Screen reader demo: select using placeholders
- 22:55Screen reader demo: searching with
- 23:45Performance code wel
