[derek73]

Part of #762, batch 2. It goes first because #752 (vocabulary phase 2) will add entries to this file.

What changed

• docs/design/decisions/vocabulary-model.md gains a ## Standing block, 24 present-tense lines, each pointing at its entry, followed by a ## Log heading.
  • The 20 model decisions of 2026-10-15/16 are covered, along with the Describe each word by its readings: per-word TOML vocabulary compiled to today's sets (phase 1) #737, Vocabulary model and language packs (tracking) #747, Classify the particles in und.toml by language and kind #749, Classify the titles in und.toml by language and kind #750 and Assign languages to the credentials and the remaining und.toml words #751 entries.
  • Two pairs of entries share a date and issue: the two 2026-10-19 #750 entries and the two 2026-10-20 #751 entries. Each pointer to one of them carries the words its entry's text opens with.
  • Decisions not built yet are worded as plans ("is to…"): phase-3 pack narrowing, the e/y/i one-case fork, the modifier rule, the contested-head gate and the patronymic field.
• A Not standing: line covers the plans the classifications carried out. One is phase 1's interim conversion into und.toml; the other is Classify the particles in und.toml by language and kind #749's "stay where they were until classified". It also covers Vocabulary model and language packs (tracking) #747's reason for exempting und.toml from the order check, which no longer holds.
• One new log entry, 2026-10-21 Roll the decisions Standing block out past C1 #762: und.toml stays exempt from check 10 because it holds no language's words, the reason nameparser/vocab/README.md already gives. Approved by Derek in chat, 2026-10-21. It re-grounds an existing exemption rather than deciding anything new.
• tools/vocab/build.py: the UNORDERED comment now gives that reason. It used to cite vetting comments ("Same word as عبد below") that left with Classify the particles in und.toml by language and kind #749–Assign languages to the credentials and the remaining und.toml words #751.
• vocabulary-model is off tests/v2/decisions_without_standing.json.
• Open: Phase 2: derive ambiguity from clashing readings and extend the gate to every reading kind #752, Phase 3: language packs that prefer or restrict their language #753, Phase 4: order and shapes, the rules a pack can choose, and the output model #754, Should bon, dí, heer, herren and heren stay in the vocabulary? None is attested in a person's name #766. Should suffix split into generation and credentials? (Smith, Jr., PhD) #326 is left off: it is a sub-issue of Phase 4: order and shapes, the rules a pack can choose, and the output model #754, and decisions/suffix-field-composition already declined it for 2.x.

Checked

• Before writing each line I grepped docs/design/ for every issue cited. No other file reverses part of an entry here, so no entry head needed a "superseded by" note.
• Parsed on the tree:
  • Dean Smith → given Dean
  • Wm Smith → title Wm
  • Vice John Smith → title Vice, while John Smith Vice → family Vice
• Read from the vocabulary and the build: e, y and i are conjunctions; dean, justice and major are not titles; und.toml holds only bon and dí; no credential is in mul.toml; the phase-2–4 key names and addresses = "full"/"any" are refused by the build.
• Mutating the pointers on purpose made the check fail as it should: a wrong twin word or a dropped twin word fails as ambiguous, and a wrong date fails as no entry.
• uv run pytest -q passes on 3.11 and on 3.15: 22464 passed, 125 skipped, 3 xfailed each time. That run was before the review fixes, which touch only docs and a comment; the doc and vocab tests pass on the final tree.
• The design-docs reviewer ran over the branch (10 findings, all fixed), then over the fix commit (4 findings, all fixed in f2b00ab). The second round found a false claim the first fix had introduced: that modifier words are titles in every position.

Found stale, not fixed here

• The Describe each word by its readings: per-word TOML vocabulary compiled to today's sets (phase 1) #737 entry says Simplified and Traditional Han is "in practice" what a script subtag is for. Since Link zh Traditional and Simplified spellings with same_as (#747) #790 both spellings sit in one zh.toml, linked by same_as, and the Standing line states only the general rule.
• The file's title, "(2026-10-15 → 2026-10-18)", and its index line still give a date range that ends before the Vocabulary model and language packs (tracking) #747–Roll the decisions Standing block out past C1 #762 entries. Titles are not read as entries, so I left them.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.05%. Comparing base (49fb035) to head (78de760).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #800   +/-   ##
=======================================
  Coverage   99.05%   99.05%           
=======================================
  Files          47       47           
  Lines        4642     4642           
=======================================
  Hits         4598     4598           
  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.