Maiden marker counts only behind a surname; a credential starts a suffix run (#601, #602)

[derek73]

Closes #601. Closes #602. Grew out of #548 (Dr. nee Smith PhD Prof.), which was closed into these two.

Summary

• Should a name word after a credential join the suffix? John Smith PhD Jones reads middle Smith PhD #602 (rules.md#S2): an unambiguous credential after the name core (two name words, or the given word after a family comma) starts a suffix run to the end of its part. A title word in the run is a title, and an absorbed name word is reported as suffix-or-name. John Smith PhD Jones gives suffix PhD Jones, and Eric H. Holder Jr. Attorney General gives title Attorney General, as the comma spelling already did. Ambiguous credentials (MA), particles, bound given names, single letters and bare title words do not start a run.
• Rethink the maiden-marker rule (M2): a marker stands after the current name and takes words up to the trailing suffix run #601 (rules.md#M2): a maiden marker counts only in the name before any comma or in the surname part before a family comma, behind a name word, and not straight behind a title-only head, a suffix word or a connective. Anywhere else it is an ordinary word, as in 1.4.0: Doe, Jane nee Smith gives middle nee Smith. A particle counts as the surname, so Mai Le née Nguyen and Anh Do née Tran are unchanged. The clause ends where the trailing run of the clause-free name begins, and that run takes its suffix/title roles at the take, so no join reaches it. This replaces the Should Doe, Jane nee Smith MA read MA as a credential? The maiden name keeps it, where it gives up PhD and Jr. #533/Should Jane Doe nee Smith Prof. read Prof. as a title? The maiden name keeps it, where Jane Doe Prof. gives the title #535 release machinery (_a_name_word_ahead, _join_takes_the_member, _release_reads_off, the clause link exception, and TailReader, now ClauseSite).
• Docs: rules.md M2 is restated with its examples regenerated, and S2, H5, C1, C2 and P5/P6 are touched. decisions.md has a Rethink the maiden-marker rule (M2): a marker stands after the current name and takes words up to the trailing suffix run #601 entry (head rule, prototype variants, triage, accepted costs) and a Should a name word after a credential join the suffix? John Smith PhD Jones reads middle Smith PhD #602 entry. AGENTS.md is swept. The release log has Rethink the maiden-marker rule (M2): a marker stands after the current name and takes words up to the trailing suffix run #601 and Should a name word after a credential join the suffix? John Smith PhD Jones reads middle Smith PhD #602 bullets, and the Josep Carod i Rovira reads family Rovira — Catalan and Polish link surnames with i, which is not a conjunction #397 and Should a declared suffix delimiter inside a joined connective run be dropped? Smith, John, Puig - i Soler keeps the - in the suffix #549 bullets are corrected.
• Ledgers: every baseline classifies the moved names under new fix(#601)/fix(#602) rules, and all five gates exit 0.

Degenerate input

Jane Doe, PhD née Smith and Berg, Jane van der nee Smith DO read differently than in 2.3.0. Both are garbage in, garbage out, with no reading designed for them. decisions.md#M2 records the measured move so it isn't mistaken for a regression, and the release log leaves them out.

Review rounds

• Design-docs review: six wording fixes (scope of M2's reporting sentence and H5, the suffix-comma part, a stale "open on Rethink the maiden-marker rule (M2): a marker stands after the current name and takes words up to the trailing suffix run #601", stale perf prose).
• Code review: the given-part run absorbed a particle that P6 then attached, which gave two false reports. The run now leaves P6's particle tail alone. A fuzz over 15,657 comma shapes finds no token reported twice. Also fixed: stage-ownership prose.
• Fix-commit review: narrowed that skip to exactly the tail P6 attaches, and stated the do exception in S2.

Simplify pass and design docs

• /simplify (f7c1aa88) made no behavior change. Every field, ambiguity and initials string is identical over 95,813 texts under three policies. Ordinary names keep their frame counts, and the run and maiden shapes got cheaper (John Smith PhD MBA FRCS 395 → 358, Jane Doe nee Smith 238 → 221). Two cross-stage drift risks it surfaced are filed as Two stages hand-copy each other's trailing-word logic after #601/#602; share one predicate each #610.
• mechanisms.md gains READ-WITHOUT-THEN-BIND and LICENSED-BY-NEIGHBOR, the two patterns Rethink the maiden-marker rule (M2): a marker stands after the current name and takes words up to the trailing suffix run #601 used. docs/design/AGENTS.md gains the "rethinking a rule that grew by patching" signal. Whether C1, S2 or P6 should get the same treatment is Have C1, S2 or P6 grown by patching the way M2 had, and should they be rethought? #611.
• Master was merged twice (Liam Ó Murchú reads Ó as a middle initial: add the Irish particles Ó, Ní, Ua (and Malay binti) #604, then Salutations and titles in other languages read as given names: Herra Väinö Johansson gives given Herra #606/Add salutations, offices and ranks in other languages to TITLES #608). Each time only the 1.4.0 comma-family claim counts conflicted, and they were re-recorded over the merged corpus.

Notes

• The 1.4.0 ledger still has Doe, J. nee MA ba and Doe, Jane nee Smith PhD MEng claimed by the broad comma-family rule.
• Unclassified radar diffs at each baseline predate this branch.

Test plan

• uv run pytest (11,379 passed on the final merge), also run without namedivider
• uv run mypy, uv run ruff check
• sphinx-build -b doctest docs, python -m doctest README.rst
• differential gate at 1.4.0, 2.0.0, 2.1.0, 2.2.0 and 2.3.0, all exit 0, no moved shapes
• release-log readings checked against the 1.4.0, 2.0.0–2.3.0 wheels

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.97%. Comparing base (59b95bc) to head (af785fa).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #609      +/-   ##
==========================================
- Coverage   98.98%   98.97%   -0.01%     
==========================================
  Files          45       45              
  Lines        4241     4214      -27     
==========================================
- Hits         4198     4171      -27     
  Misses         43       43              

☔ 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.