#69 Capacity advice should measure what users wait for, then recalibrate comfortable sizes
–
- github
- github/openchronology/timeline-app
- github/openchronology/timeline-app#69
- issue
- state:closed
The UI benchmarks (#51) check that wherever an interaction misses its budget, the "consider another platform" advice is already showing. In the browser it usually isn't. A run at 1,000, 4,000 and 10,000 moments, at full speed and with the CPU throttled 4×, found **37 advice gaps**. For example: - at 10,000 moments, full speed: wheel zoom takes 279 ms and drag frames 217 ms, with no advice; - at 4,000 moments, 4× throttled: wheel zoom takes 766 ms, and opening takes 1.70 s against its 1.5 s limit, with no advice. **Why:** `src/capacity.ts` compares the median of the editor's own *drawing* time with `THRESHOLDS.redraw` (150 ms). Users wait from the input event to the next paint, which adds input handling and painting. The benchmark measures that wait directly. At 10,000 moments the drawing is about 140 ms, under the threshold, while the wait is about 280 ms. **Proposal:** 1. **Measure the wait in the app:** for zooming, panning and keyboard navigation, take the time from the input event's `timeStamp` to the frame after the view is drawn (the same rAF-then-message timing the benchmark probe uses), and feed those samples to `advise()` in place of the drawing time. 2. **Recalibrate the comfortable sizes** in `src/capacity.ts` (`COMFORTABLE`) and `docs/capacity.md` from `bench/ui.mjs` rather than from drawing time alone. Today's numbers suggest the browser is comfortable for navigation up to about 1,000–4,000 moments at full speed, and below 1,000 with a 4× slower CPU. 3. **Make CI enforce it:** extend the CI gate to the comfortable sizes once they're honest (or once the gestures issue raises them), so advice gaps can't return unnoticed. 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://github.com/openchronology/timeline-app/issues/69
Related
- #102 Advise by how long users wait for views; recalibrate comfortable sizes
- #51 thorough ui benchmarks verifying user interaction within thresholds defined by systems benchmarks
- #68 Make panning, zooming and pinching meet the 150 ms budget at the documented sizes
- #70 UI interaction benchmarks against the capacity budgets (browser and platform)
- #74 UI benchmarks on the desktop app, plus two desktop fixes
- Benchmark UI interactions against the capacity budgets
- Advise by how long users wait for views, as the benchmarks measure it