#39 Advise users when a timeline outgrows its platform
–
- github
- github/openchronology/timeline-app
- github/openchronology/timeline-app#39
- issue
- state:closed
## Background Measurements (see the capacity benchmarks, and the discussion that led to this issue): - **In-memory (browser guest editing, offline HTML)** strains on processor time, not RAM. On a fast desktop: - opening takes 0.4 s at 10,000 moments, 3.2 s at 50,000 and 23 s at 200,000; - the first whole-timeline frame takes 92 ms, 0.7 s and 3.4 s; - the draft save after each edit pauses the page for about 35 ms, 180 ms and 850 ms. Memory is about 1.4 KB per moment (272 MB at 200,000). Phones are estimated at 3–5× slower. - **SQLite (desktop)** is not strained up to the 200,000-moment cap: it opens in 1.1 s, draws the whole timeline in 48 ms, saves in 2.7 ms, and the file is 68 MB. - **PostgreSQL (platform)** is in the same range and has the same 200,000-moment cap per timeline. So moving from the browser to the desktop or the platform pays off from about 20,000 moments on a desktop browser, and sooner on slow devices. Moving from the desktop to the platform does not make a single user faster; it is worth suggesting for collaboration and on slow hardware. ## Proposal 1. **Browser (in-memory) advisory.** A non-blocking banner: - **Trigger:** measurements on the user's own device: opening takes more than ~1.5 s, the median whole-timeline frame is above ~150 ms, or the pause after an edit is above ~100 ms. Fall back to a moment count of ~20,000 where measurements are unavailable. - **Actions:** export and open in the desktop app; when signed in, save to your account. The platform's editor loads only what is visible, so it scales like the database. - **Dismissal:** remembered per timeline. 2. **Desktop advisory.** Suggest the platform when the device measures slow (opening or whole-timeline views consistently above ~1 s), and mention it for collaboration. Not based on timeline size, since the platform would not be faster. 3. **Approaching the cap, every platform.** From ~150,000 moments (or durations or relationships), explain that a timeline holds at most 200,000 and suggest splitting it. No platform lifts the cap. 4. Thresholds come from the capacity benchmarks (real Chromium at normal speed and with 4× processor throttling); the docs explain them. ## Acceptance criteria - Each advisory appears at its measured or fallback threshold and not before. Browser tests cover them with large generated timelines and with throttled measurements. - Advisories never block editing, and their dismissal persists. - The docs explain the limits per platform, with the benchmark data. https://github.com/openchronology/timeline-app/issues/39
Related
Open #39 Advise users when a timeline outgrows its platform in the timeline