Repository navigation
id: server-side redraw batching (few-ms frame coalescing) for all presentations #561
Description
What and why
Accumulate redraws for a few milliseconds and send one frame instead of a full view per committed event. Server-side only, it helps every tier and transport, and per the transport comparison it matters more than SSH vs QUIC. Per the transport comparison, render batching matters more for latency than SSH vs QUIC, and it helps every presentation at once.
Source
- RPT §3 'Output path and batching'
- RPT §5 'Buffered output while staying interactive'
- RPT §13 (latency is mostly render batching)
- RPT §14 Tier 2
RPT = doc/2026-10-18T18-40-00Z_report_platform_world_state/2026-10-18T18-40-00Z_report_platform_world_state.md, REF = doc/2026-10-18T00-00-00Z_reference_platform_state/2026-10-18T00-00-00Z_reference_platform_state.md. Both are uncommitted reports in the platform-world-apps-migration worktree.
Current status
Not started. Host events already coalesce through a tokio watch (RPT cites world.rs watch); committed events still trigger an immediate full view per session (RPT cites world_session.rs:2110-2115). grep -n 'batch\|coalesce\|debounce' pkgs/id/src/world_session.rs finds nothing.
Acceptance criteria
- A short plan fixes the coalescing window and where it lives (session driver)
- Committed events inside the window produce one frame per session, not one full view per event
- SSH, web and CLI presentations all get it
- Frame count and bytes are compared against the F11.9 baseline
Size
M
Prerequisites
- id(world): F11.9 bytes-per-frame baseline table #542 — D1-009 F11.9 Bytes-per-frame baseline table
Activity
dezren39 commentedon Oct 24, 2026
id train triage (2026-10-24): too-big-needs-design — server-side redraw batching (D1-043), terminal-side counterpart of F12.11 and a prerequisite of F13.4.
dezren39 commentedon Oct 26, 2026
Intake review
Fleet-monitor intake pass (2026-10-26). Body already carries a full
## Acceptance criteriasection, so this pass is label normalization plus a liveness check — no body edit.Verified live on main
e3467560(2026-10-26):git grep -c 'coalesce\|debounce' origin/main -- pkgs/id/src/world_session.rsfinds nothing — committed events still trigger an immediate full view per session. Still live, exactly as filed.Related: prerequisite #542 (F11.9 bytes-per-frame baseline) is still open and unenriched, and the acceptance criteria's frame/byte comparison needs it. Terminal-side counterpart: #562. The plan-first acceptance bullet (fix the window and where it lives) is a design decision worth a short comment thread before code.
Risks / notes: too long a window makes the world feel laggy; too short gains nothing. Interactive input echo should not be coalesced the same way as committed-event redraws — worth stating explicitly in the plan.
Suggested approach: land the short plan first (window size, session-driver placement, echo exemption), then implement in the session driver so SSH, web, and CLI presentations all inherit it, and measure against the #542 baseline.
22 remaining items
dezren39 commentedon Nov 1, 2026
agent-note:v1
{"v":1,"kind":"parked","task":4832,"lane":"f","note":"PR #3905 OPEN for #561 (head 149880a, MERGEABLE/CLEAN after merging current main; sole conflict was Cargo.toml dev-deps, union-resolved). Codex round 1 on the implementation head: no findings; re-review requested for the merge head. Targeted paused-time test (coalesces) compiling/running on the laptop, log /tmp/4832-test.log. Parked 30m: resume = test pass + Codex clean on the new head, then merge PR #3905 and complete task 4832."}
dezren39 commentedon Nov 1, 2026
agent-note:v1
{"v":1,"kind":"picked_up","task":4832,"lane":"openchamber-scheduled-lane","note":"Resumed queue task 4832. Fresh state: #561 OPEN; sole matching PR #3905 OPEN at head 149880a, latest Codex review clean, no hard-stop labels or competing implementation PR. Existing targeted test log currently shows compile failure in id-core tests (unresolved vt100); checking whether this reproduces on current main, which has advanced to bc76a7e since the PR merge head. Next: establish test status against clean main/current head and then merge if verified, otherwise park with exact resume condition."}
Merge pull request #3905 from developing-today/queue/4832-561-server-…
dezren39 commentedon Nov 1, 2026
agent-note:v1
{"v":1,"kind":"done","task":4832,"lane":"openchamber-scheduled-lane","note":"Merged PR #3905 at 26bed62; GitHub closed #561. The targeted paused-time regression passed on the merged-main head (1 passed, 0 failed); rustfmt and diff checks passed. No required CI checks were configured; CodeRabbit skipped. Queue task completed; sandbox backlog/sweep bookkeeping deferred to fleet reconciliation."}