[derek73]

Closes #1039.

What changes

junior (mul.toml) and júnior (pt.toml) gain given and family readings. They become suffix words that are also names, so the words-to-spare count gates them (rules/S11).

parse("Silva, Junior Carlos")   suffix 'Junior Carlos', no given  →  given 'Junior', middle 'Carlos'   (as 2.2.0–2.3.0)
parse("SILVA, JUNIOR")          suffix 'JUNIOR', no given         →  given 'JUNIOR'
parse("Kofi Junior")            suffix 'Junior', no family        →  family 'Junior'                    (as 1.4.0)
parse("John Smith Junior")      suffix 'Junior'                   →  suffix 'Junior' (now reports suffix-or-name)

A new reading property: title_case_is_name.

• Rules S6 and S11 read a gated suffix word's case: lower case means the suffix, Title case means the name. That lean is now a property of the reading rather than of the generation kind, as decided.
• It is set where the lean is true:
  • fils, which French writes in lower case as the generation;
  • the Czech, Slovak, Slovenian and Croatian st.;
  • kc, pc and dl (Should common British post-nominals (KC, FREng, DL, FRS, PC…) be in the vocabulary, so Jane Doe KC keeps family Doe? #965);
  • père, which Locale packs for Alexandre Dumas fils and Jiří Dienstbier ml.: generational words that collide outside their language #684 turned on by default while this PR was open, gated as fils is. Without the property it would have lost Locale packs for Alexandre Dumas fils and Jiří Dienstbier ml.: generational words that collide outside their language #684's lean (Jean Olivier Père → suffix 'Père'), so it carries it and reads as on master.
• Without the property, a gated word's writing says nothing and the count decides, as for an agnomen. Junior is written in Title case for the generation too, so it doesn't carry the property, and John Smith Junior keeps the suffix.
• Adding the property moved no parse: the suite and the committed snapshot were unchanged before junior's readings went in, and a 7,200-parse grid over every gated word under three orders and the cs/pt packs moved nothing.
• Callers: Lexicon.add(w, kind="generation", title_case_is_name=True). The build allows it on a generation reading or on a credential written as a word.
• 2.x: a 2.x Lexicon holding a non-agnomen word in suffix_words_ambiguous translates to a reading with the property, the lean 2.x gave it. The refusal message and the migrate.rst row say so.

What moves

Measured with tools/differential/measure.py --base against master 39839b9. The committed snapshot moves only the rules' and case table's Junior names. The CLDR, GND and Wikidata records don't move.

corpus	entries differ	of which a new suffix-or-name report alone	notable
tse-donors	14,839	14,763	38 two-word ADELSON JUNIOR → family 'JUNIOR'
orcid-researchers	505	441
jrc-names	391	209	177 Fabio Junior → family 'Junior'
patentsview-inventors	212	211
psc-individuals	97	86	Mr Oliver Junior → family 'Junior'; (Junior) → nickname, as (Neto)
ctgov-officials	87	77	… SILVA JUNIOR, DOCTORATE → family '… JUNIOR' (base already reversed)
fec-individuals	61	10	42 comma names gain the given name; 9 BRITO JUNIOR, RAIMUNDO → family 'BRITO JUNIOR'

The other six local corpora don't move.

Agreement with the filed boxes:

• orcid-researchers: 9 records progress, 4 regress.
• psc-individuals: 5 progress, 1 regresses.
• patentsview-inventors: no move.

Decided, and accepted without a decision yet

• Decided: give Junior a name reading; keep it a generation word, not an agnomen, so Reading.kind stays 'generation'; and make the lean a property that marks the words whose case leans, so the default is no lean.
• Not yet put to a decision, recorded as ACCEPTED in decisions/S11: in front of another suffix word that is also a name, Junior is a name word, as Neto already is on master (Pedro Neto Filho → family 'Neto').
  • So Pedro Junior Filho reads family 'Junior', suffix 'Filho', where 1.4.0 through 2.8.1 read middle 'Junior', family 'Filho'. Likewise Pedro Junior fils.
  • The rows that pinned the Gate generation readings by kind: fils, the Brazilian agnomens, default = false (#752, phase 2h) #890 rule are re-spelled with Jnr, which has no name reading, so they keep testing it.
  • Smith, Junior. now reads title 'Junior.' by rules/H2's period shape, as Smith, Neto. does. 2.2.0 through 2.8.1 read the suffix. No corpus name writes it.

Docs

• Rules amended:
  • S5, S6, S7 and S11 state the property.
  • S7 and S11 state that a suffix word that is also a name counts as a name word in front of another, unless its writing settles it (S6).
• Decisions: a dated S11 entry, with head notes on the two entries whose "the kind says what the writing means" it supersedes.
• Elsewhere: the vocab README property row, root AGENTS.md, the RFC table, customize.rst, migrate.rst, and a release-log bullet.

Review

• Code review: a grid of 12 words × 5 spellings × 20 shapes under 6 parsers found every move to be a junior or júnior name. It found the 2.x translation dropping the lean (fixed), S5 left stale (fixed), and Smith, Junior. (recorded).
• Design-docs review: S11's precedence and Then clauses contradicted their own examples; the Standing attribution, counts and history notes needed corrections. All fixed.
• Second-round review of the fix commit: the new name-word clause dropped S6's exception, migrate.rst still gave the lean-less 2.x spelling, and the FEC split was off by one. All fixed in the third commit.

🤖 Generated with Claude Code