derek73 · Nov 13, 2026 · Nov 13, 2026 · Nov 13, 2026 · Nov 13, 2026 · Nov 13, 2026
diff --git a/AGENTS.md b/AGENTS.md
@@ -300,7 +300,7 @@ The 2.0 rewrite landed as underscore-private modules beside the v1 code, which 2
 
 **Adding a curated sub-set of an existing vocabulary set** (a marker over a base set, as `AFTER_GIVEN_TITLES` is over `TITLES`) — it is a behavioral PROPERTY of a reading, not a second word list: a row in `nameparser/_reading_index.py`'s `BEHAVIORAL` table, a branch in its `compile_readings` adding the words whose reading carries it and a key in its `Compiled` TypedDict, a constant of the same name in `tools/vocab/build.py`'s `CONSTANTS` roster (an import-time assert there holds the two to one list), and, for `Lexicon` to read it, `_lexicon._COMPILED_FIELDS` and `_LexiconSets` (no module re-exports a compiled set since 3.0, #1007). A property the compiler reads is kept in the reading index by `indexed()`, which reads `BEHAVIORAL`, so a new one reaches a choice of languages too. The subset relation then holds by construction, since the property cannot exist without its reading, so it needs no check. A marker that records a clash with a NAME is the other shape: derived from the word's name reading beside the clashing one (`SUFFIX_ACRONYMS_AMBIGUOUS`, `SUFFIX_WORDS_AMBIGUOUS`, and the complement `NON_GIVEN_NAME_PARTICLES` works from), not written as a property, so it too holds by construction (decisions/vocabulary-model, 2026-10-25). A relation construction does not give (`∩ == ∅` with a set the sub-set cannot logically overlap; a sub-set member that is also a title is silently inert, title handling consuming it first) goes in `check_pool`, with a mutation in `tests/v2/test_vocab.py`'s table. Do **not** forbid `titles ∩ particles` — that overlap is intentional (`freiherr`, `st`). A caller states a property on a reading (`add(w, kind="salutation", after_given=True)`), so a caller's word reaches the base and the sub-set together, as a file's does (3.0's 4i-b). Two traps a gating set met (#752 phase 2i). Where a caller listing a word in the BASE set has always meant "read it as this everywhere" (the 1.4 recipe `c.titles.add("dean")`, spelled `remove("dean", kind="given")` since 3.0's 4i-b), the gated words compile OUT of the base rather than into both, or the gate overrides the caller's listing: `TITLES_AMBIGUOUS` is not a subset of `TITLES` for that reason, and a word in both reads as a title. And `_VOCAB_FIELDS` is read by three rosters: two in `_pipeline/_vocab.py`, the capitals shape's "unlisted" (`in_any_wordlist`, over `_LISTED_FIELDS`, the roster less four fields) and the comma contrast's "claimed as non-name" (`_NON_NAME_FIELDS`, those less `surnames` and `bound_given_names`), and `_assemble._listed_whole`, which reads it whole to tell a word the parse's own lexicon lists for `Reading.languages`. A new field joins all three by default; a field whose words may be names leaves the two in `_vocab`, which the grid found as `John Smith, DEAN` reversed where C6 had read a credential; #789's `given_name_abbreviations` met it again (`John Smith, JAS` and `Jas SMITH, XYZ` reversed), found by review because its grid held only the dotted spellings the set was for -- a new field's grid holds the spellings it must NOT move too.
 
-**Shipping data from a new source: name its licence in the package metadata.** Data the project derives by measuring (counts, its own cutoffs and filters) is its own; a curated list copied whole keeps its licence, and goes in an extra package rather than the core if that licence is a burden (decisions/shipped-data-licences, FACTS WE MEASURE). A list derived from a third-party dataset (a census table, a name dictionary, a Unicode database) that lands in `nameparser/` adds that dataset's SPDX identifier to `pyproject.toml`'s `license` expression and a line to README.rst's License section, and, where its licence asks its text or notice to travel with copies, a `LICENSE-<source>` file at the root, listed in `license-files` and in `tests/v2/test_licensing.py`'s `_THIRD_PARTY`; a module docstring crediting the source is not enough on its own. The INE (CC BY 4.0, 3.0.0), Taiwan (OGDL, 2.5.0) and Unihan (Unicode License v3, 2.5.0, in `locales/zh.py` through `tools/zh_surnames.py`, which zh.py's docstring did not name) data shipped behind an expression naming the LGPL alone, and nam_dict's GFDL data reached the tree with no copy of the GFDL, until decisions/shipped-data-licences (2026-11-12, #1370).
+**Shipping data from a new source: name its licence in the package metadata.** Data the project derives by measuring (counts, its own cutoffs and filters) is its own; a curated list copied whole keeps its licence, and goes in an extra package rather than the core if that licence is a burden (decisions/shipped-data-licences, FACTS WE MEASURE). A list derived from a third-party dataset (a census table, a name dictionary, a Unicode database) that lands in `nameparser/` adds that dataset's SPDX identifier to `pyproject.toml`'s `license` expression and a line to README.rst's License section, and, where its licence asks its text or notice to travel with copies, a `LICENSE-<source>` file at the root, listed in `license-files` and in `tests/v2/test_licensing.py`'s `_THIRD_PARTY`; a module docstring crediting the source is not enough on its own. The INE (CC BY 4.0, 3.0.0), Taiwan (OGDL, 2.5.0) and Unihan (Unicode License v3, 2.5.0, in `locales/zh.py` through `tools/zh_surnames.py`, which zh.py's docstring did not name) data shipped behind an expression naming the LGPL alone, and nam_dict's GFDL data reached the tree with no copy of the GFDL, until decisions/shipped-data-licences (2026-11-12, #1370); the Unihan data has since been replaced by Wikidata's, CC0, and its licence left the expression (that file's 2026-11-12 UNIHAN REPLACED).
 
 **Comparing configurations: use `tools/differential/config_diff.py`, not a throwaway script.** Parsing the corpora under two or more configurations in one tree (the default against a choice, plain against strict, a lexicon with a word removed via `expr:`) is `config_diff.py diff`, parallel, with corpus, country, regex and sample filters, the moves grouped by field and the words whose role changed ranked; `config_diff.py gaps` is the per-language strict-gap report on top of it, and `--readings` writes every name's readings so two trees' runs compare line by line. #1000, #1005, #1017 and #1019 each wrote a single-process script for this (tools/differential/README.md, "Measuring a change").
 

diff --git a/LICENSE-unicode b/LICENSE-unicode
diff --git a/README.rst b/README.rst
@@ -305,9 +305,6 @@ under other licences, which apply to those parts:
 - the group suggester's common given names from Jörg Michael's
   nam_dict: GNU Free Documentation License 1.2 or later, in
   ``LICENSE-nam_dict``;
-- ``locales.ZH``'s simplified surname forms and the group suggester's
-  pinyin surnames from the Unicode Consortium's Unihan database: Unicode
-  License v3, in ``LICENSE-unicode``;
 - ``locales.ES`` and ``locales.ES_AR``'s given names from Spain's
   Instituto Nacional de Estadística: Creative Commons Attribution 4.0,
   credited in ``nameparser/locales/_es_given_names.py`` and
@@ -317,7 +314,7 @@ under other licences, which apply to those parts:
   ``nameparser/locales/zh.py``.
 
 The package's licence expression is
-``LGPL-2.1-or-later AND Apache-2.0 AND GFDL-1.2-no-invariants-or-later AND CC-BY-4.0 AND OGDL-Taiwan-1.0 AND Unicode-3.0``.
+``LGPL-2.1-or-later AND Apache-2.0 AND GFDL-1.2-no-invariants-or-later AND CC-BY-4.0 AND OGDL-Taiwan-1.0``.
 
 .. |Build Status| image:: https://github.com/derek73/python-nameparser/actions/workflows/python-package.yml/badge.svg
    :target: https://github.com/derek73/python-nameparser/actions/workflows/python-package.yml