Repository navigation
id: validation gap — P4.11 DataScheme update-cost measurement at 100/1,000/5,000 members was never recorded #1396
Description
Gap found by the #1075 validation audit (Phase 4, P4.11) — 2026-10-27
The plan of record (doc/2026-10-11T11-08-42Z_design_remaining_work/plan-phases.md, P4.11) requires: "Measure DataScheme update cost at 100, 1,000 and 5,000 members with our adapter, for the BeeKEM comparison — recorded in the phase document; no pass or fail."
A fresh search on main 452d69e6 (2026-10-27) finds no such measurement recorded: no update-cost figures at 100/1,000/5,000 members in the phase documents under doc/2026-10-11T11-08-42Z_design_remaining_work/, in pkgs/id/src/encryption/tests.rs, or elsewhere in the repo (the BeeKEM assessment records why BeeKEM was not adopted, but not our adapter's measured update cost at those scales). The related size evidence that does exist is welcome-size only (test_each_epoch_adds_about_forty_bytes_to_the_welcome, ~40.0 B/epoch, P4.4).
Acceptance criteria
- DataScheme update cost (time and/or message bytes) measured with the real adapter at 100, 1,000 and 5,000 members and recorded in the Phase 4 phase document, in a form the BeeKEM comparison can cite.
- The bench/test that produces the numbers is committed and re-runnable (name it in the document).
Found by: queue task 1256 (validation audit of #1075), slice 3 (Phases 2–4 walk). Parent audit: #1075. Path-to-Phase-12 queue: #1235.
Activity
developing-today commentedon Oct 27, 2026
CODE-488
dezren39 commentedon Nov 2, 2026
🤖 Agent note · state: picked up
dezren39 commentedon Nov 2, 2026
🤖 Agent note · state: done
Triage verdict — #1396: ALREADY DONE (audit miss); close, no follow-up packet
The P4.11 measurement was recorded on main two weeks before this issue was filed — verified fresh this run against origin/main (
0e26fa4b3), and back-checked at the issue's own search base (452d69e6, 2026-10-27):pkgs/id/src/encryption/tests.rs—measure_update_cost()plustest_the_update_cost_measurement_grows_with_the_membersplus the ignored release benchmarkbench_update_cost()iterating[100, 1_000, 5_000]members (run:cargo test --release --lib encryption::tests::bench_update_cost -- --ignored --nocapture). Present verbatim at452d69e6(lines ~2388–2397), so it predates this issue. Added by commit2a3de9a9"test(rust): DataScheme update cost measurement [P4.11]" (2026-10-12).doc/2026-10-12T14-33-36Z_phase_4_groups_encryption/2026-10-12T14-33-36Z_phase_4_groups_encryption.md, section "P4.11: DataScheme update cost (measured)" — full table at 100 / 1,000 / 5,000 members (update bytes 30.7 KB / 310 KB / 1.54 MB; ~309 B and ~100 µs per member), side-by-side BeeKEM comparison table, fan-out break-even, and reading. That doc's own status table marks "P4.11 DataScheme cost measurement … done … numbers below". Recorded by commitf5dc8041"docs(id): record P4.8 to P4.11 and the DataScheme update cost [phase 4]" (2026-10-12).adc1cad0, merged 2026-10-12) — 15 days before this issue's 2026-10-27 filing.Why the audit missed it: the audit searched
doc/2026-10-11T11-08-42Z_design_remaining_work/(the plan directory, which holds the P4.11 requirement, not results) and did not surface the results doc (doc/2026-10-12T14-33-36Z_phase_4_groups_encryption/) or the tests.rs benchmark, both of which already existed at its search base.Why no follow-up packet: there is no remaining work — the instrument exists and the numbers are recorded with the BeeKEM comparison the plan asked for. (One optional nit for the owner, not queue work: the phase-4 doc notes the BeeKEM assessment itself was deliberately not edited with the measured break-even; owner call.)