[derek73]

Part of #762. This is one of three parallel batches in #762's final round. Six files in docs/design/decisions/ get a ## Standing block and a ## Log heading, and their keys come off tests/v2/decisions_without_standing.json:

• differential-ledger
• parse-cost
• comma-suffix-arc
• deviates-registry
• review-agent-single-source
• 3-0-reevaluations

No parser code or rules/ file changes.

What each block says

differential-ledger

• Three lines on the Replace the differential ledgers with a committed snapshot of corpus readings #664 retirement: the snapshot, compare.py staying unclassified, and the CLDR agreement ratchet. They are told apart by the twin key words, because all three entries share the head 2026-10-13 #664.
• Three lines for the tooling arc's surviving Declined entries:
  • Policy annotations stay narrow;
  • check_source checks against <repo>/nameparser;
  • DEFAULT_BASELINE rotates in the release checklist.
• The superseded arcs were already marked as superseded, so there is no Not-standing line.
• The two ## arc headings are demoted to ### under ## Log.

parse-cost

• The ±2% call band per interpreter, and the 5s wall-clock backstop.
• Moving a row requires a decision entry.
• The current rows, matching _CALL_BASELINE: 307/344 on 3.11 and 288/325 on 3.12–3.15. call_count.py reads 307/344 on 3.11 today.
• The link pin: 2,478 and 2,459.
• The parked post_rules scan collapse.
• copy_with and its guard.
• Group's direct predicate calls.

comma-suffix-arc

• The multi-word entries were removed rather than split.
• A multi-word credential is matched through the run predicate, so SUFFIX_PHRASES and A6 are declined.
• Multi-word merges happen at the render tag.
• An unlisted abbreviation at the end of a name gets no structural fallback.
• Elsewhere lines point to C8 for the audit table as it shipped, H5 for the trailing slot, and S9 for the dotted shape.

deviates-registry, review-agent-single-source

• One line each. The second file has a second line for its Declined heading test.

3-0-reevaluations

• One line per (A)/(B) bullet that still stands.
• The FAMILY_COMMA line is scoped to Latin script, with an Elsewhere line to cjk-comma-demotion.
• The pack lines point to vocabulary-model, and Phase 3: language packs that prefer or restrict their language #753 is listed under Open.

How undated bullets got pointers

The pointer grammar only reads bullets that open with a date. So I added dated Log lines that name the bullets. These lines add no decisions, and existing entries are only touched by the edit kinds listed under Edits to existing entries below.

• comma-suffix-arc
  • A bundle line dated by the title (2026-07-26).
  • A line for the "Amended 2026-08-23" paragraph.
• 3-0-reevaluations
  • A bundle line dated 2026-08-15. That is the preamble's promotion date, and git log -S --reverse over docs/design/ confirms it for every undated bullet it covers.
  • One line for each bullet the bundle date doesn't fit. Bullets whose text gives a date: 2026-07-06, 08-17, 08-18, 08-29, 09-13 ×2, 09-20, 09-23 and 10-02.
  • One more for the legacy-pickle bullet, whose text gives no date. It was added by c2381f6 on 2026-08-16, a day after the promotion.
  • All these lines sit at the top of the Log, so the bullets keep their order and their "above" references.
• README.md: the ## Standing bullet now describes this form. That covers:
  • a bundle in a file with no dated title;
  • a bullet dated in its own text;
  • a bullet whose text gives no date, dated by the earliest git log -S --reverse hit over docs/design/. The file's own history starts at the 2026-10-17 split, and a phrase that hits first at the 2026-08-16 unwrap crossed a wrapped line.

Edits to existing entries

Edit kind 1 (superseded notes):

• parse-cost:
  • perf(pipeline): copy stage state without dataclasses.replace #546's four conditions, by test: guard coverage for copy_with and C1's settled path (#555) #561.
  • Each row-setting entry's rows, by the next row move. I checked each against _CALL_BASELINE's git history.
  • perf(pipeline): copy stage state without dataclasses.replace #546's link pin, by Rethink the maiden-marker rule (M2): a marker stands after the current name and takes words up to the trailing suffix run #601's re-pointed shape.
• comma-suffix-arc:
  • The audit table's ms/sa, by C8's fix(assign,vocab): a one-word family comma reads its credential run as suffixes (#296, #325) #428 review.
  • The trailing-abbreviation decline for the dotted spelling, by S9's 2026-09-14 entry.
  • Plan amendment 1's condition, by C8's fix(assign,vocab): a one-word family comma reads its credential run as suffixes (#296, #325) #428.
• 3-0-reevaluations:
  • The harness-baseline bullet, by differential-ledger's Replace the differential ledgers with a committed snapshot of corpus readings #664.
  • FAMILY_COMMA, by cjk-comma-demotion's Should docs/ document input shapes for the family-first name orders, and does a trailing particle belong in the notation? #469.
  • The 2026-08-18 bullet, by P8's Should Jose E Maria Santos join like Jose e Maria Santos does? (the single-letter connective's Latin-capital veto was never adjudicated) #383 + JOSE E MARIA SANTOS reads the conjunction E as an initial; an all-caps name carries no case signal to read it from #479 and P3's Josep Carod i Rovira reads family Rovira — Catalan and Polish link surnames with i, which is not a conjunction #397.

Edit kind 2 (in-place corrections, each noted):

• parse-cost's closure-inlining entry was dated 2026-10-07 and named only Share group's connective-join loop with the comma decision's title chain #617. It now names Inline group's last three predicate closures #619, the PR that landed commit 343e268.
• parse-cost's preamble said "every number below is py3.11". It now names the interpreters it actually quotes.
• Two pointers in 3-0-reevaluations no longer resolved:
  • "the order-independence decision above" now points to decisions/P1's 2026-08-16 and 2026-08-17 entries;
  • "the 2026-09-13 (A) bullet above" matched two bullets, and now names the one it means.

Review

I ran the design-docs-reviewer agent (opus) over the branch, then again over each fix commit until a round was clean: nine rounds.

Round 1 (branch):

• 3-0 said an add-only pack can't carry the single-letter policy. A pack can add a member; only removal is out of reach.
• parse-cost credited the closure inlining to Share group's connective-join loop with the comma decision's title chain #617; it was Inline group's last three predicate closures #619.
• 3-0's A Constants set entry with edge whitespace raises ValueError at the first parse: c.suffix_not_acronyms.add('ma ') then HumanName('John Smith', c) #541 line overstated what the filter drops, compared with decisions/R4.
• 3-0 said "Since Measure the CLDR corpus with each locale pack too, so a pack change shows its effect #672". It was Measure the CLDR corpus with each locale pack too #673 and Add a native-script Wikidata corpus: real names in their own scripts #693, and no decisions entry covers that change; it is now listed under Not standing.
• parse-cost's row moves were listed under Not standing instead of being marked superseded.
• comma-suffix-arc's plan amendment 1 was unmarked.
• The new dated lines broke the bullets' "above" references.
• The dated-line form wasn't described in the README.
• The legacy-pickle bullet was dated a day early.
• parse-cost's preamble was false.

Round 2:

• The preamble fix was still too broad.
• The "bridge discipline above" pointer was unclear; it now names the preamble.
• The two silently dropped TITLES typos were missing from the A Constants set entry with edge whitespace raises ValueError at the first parse: c.suffix_not_acronyms.add('ma ') then HumanName('John Smith', c) #541 line.
• The README sentence didn't cover the 08-16 bullet.
• perf(pipeline): copy stage state without dataclasses.replace #546's link pin was unmarked.

Rounds 3–8: wording of the README sentence. The two clauses were folded into one, and the dated line now takes the decision's own date. A date the text gives for something else (when a code change landed) doesn't count. The git recipe now reads across the 2026-10-17 split and the 2026-08-16 unwrap, and takes the earliest hit.

Round 9: clean.

Found but left

• parse-cost: Rethink the maiden-marker rule (M2): a marker stands after the current name and takes words up to the trailing suffix run #601's 2026-10-04 re-pointing of _LINK_BASELINE (2,301 → 2,678) has no parse-cost entry. The perf(pipeline): copy stage state without dataclasses.replace #546 note names it, but no entry records it as a row decision.
• parse-cost: the Rethink the credential-run rule (S2): read a name's trailing run once, over name units, and bind it before any join #614 entry mentions _counted_as_one, which Should S2's trailing read count the particle chain's units instead of merging a copy of them? #620 removed. This is dated history, left as written.
• tests/v2/test_benchmark.py: the comment block above _CALL_BASELINE stops at Should S2's trailing read count the particle chain's units instead of merging a copy of them? #620 and doesn't mention the Where does Freiherr St John Smith MA start: should a title that is also a particle open the name? #624 and Merge the trailing run's particle runs in group, so S2's read needs no BOUND stop or report gathering #625 moves.
• 3-0-reevaluations: the bullets' tag-then-date form (- (A) 2026-08-17 — …) is left as written. The dated Log lines index them instead.
• comma-suffix-arc: the audit-table paragraph entered the folder on 2026-08-16 (233a7db), a day after the rest of the bundle, and dates itself 2026-07-30/08-01, inside the title's range. It is covered by the 2026-07-26 bundle line. No Standing line rests on it; the audit is indexed through the Elsewhere line to C8.
• cjk-full-stops (not in this batch): one bullet records a decision approved 2026-09-09 under a 2026-09-10 bundle line, with no line of its own.
• tools/differential/README.md: the per-pack agreement views (Measure the CLDR corpus with each locale pack too #673, Add a native-script Wikidata corpus: real names in their own scripts #693) are recorded only there, in no decisions entry.

Tests

• uv run pytest -q tests/v2/test_design_docs.py tests/v2/test_doc_citations.py: 153 passed.
• Full suite on 3.11: 22772 passed, 125 skipped, 3 xfailed.
• Full suite on 3.15: 22772 passed, 125 skipped, 3 xfailed.

---

Generated by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.05%. Comparing base (a565544) to head (febfb53).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #832   +/-   ##
=======================================
  Coverage   99.05%   99.05%           
=======================================
  Files          47       47           
  Lines        4659     4659           
=======================================
  Hits         4615     4615           
  Misses         44       44           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.