A leading particle run should stop at the first name word when name_order is family-first #395

Behavior

                     GIVEN_FIRST (unchanged)      FAMILY_FIRST / FF_GIVEN_LAST
de Mesnil Juan       family='de Mesnil Juan'      family='de Mesnil'  given='Juan'
de la Vega Juan      family='de la Vega Juan'     family='de la Vega' given='Juan'
de Mesnil Garcia     family='de Mesnil Garcia'    family='de Mesnil'  given='Garcia'

The default order does not change. de Mesnil Juan reads exactly
like pennie von bergen wessels — the same shape, and #393's
discussion settled that the greedy reading is right there.

Why order decides it

Declaring a family-first order is precisely an assertion that what
follows the family is not more surname. A particle run's stopping
point is exactly the question of where surname material ends, so the
declaration is evidence about it.

Recorded at decisions.md#P1 (2026-08-17), which supersedes the
2026-08-16 sentence "no name_order moves that stopping point". The
grouping/assignment split stands everywhere else.

Accepted cost: de la Vega Juan under the default order stays wholly a
surname. A caller who means family de la Vega plus given Juan
writes the comma.

Carried forward from two failed attempts

#390 tried a position-only rule twice — PR #391 (claim in assign) and
PR #394 (claim in grouping). Both closed unmerged. What their reviews
established, all measured:

Traps that bit and will bite again

• A conjunction merge makes the leading piece span several tokens:
  de and Mesnil is one piece headed by and, which carries no
  particle tag. A guard testing only for the ABSENCE of
  vocab:particle-ambiguous passes vacuously.
• st, do and freiherr are in TITLES and the ambiguous half; a
  guard keyed on the leading piece hands them the chain.
• abdul is a P5 bound-given word; merging it into a particle group
  kills the P5 join and splits the pair across two fields.
• A conjunction-joined run counts as ONE word.
  de la Vega y Santos Juan → family='de la Vega y Santos' is
  correct, not a defect. Stated as a clause on rules.md#P3.

Verification is structurally weak here — build the pins first

• The differential sees ONE corpus name (de Mesnil Garcia), and under
  this direction the default order does not move at all, so it may see
  ZERO. Exit 0 will not be evidence.
• The two-leftover shape (de Mesnil Juan Carlos), where the two
  family-first orders diverge from each other, exists in no test, case
  row or rule example.
• tests/v2/pipeline/test_post_rules.py uses a reduced _LEX
  (particles={"de","la","van"}, titles={"mr","sir"}). Three tests
  have been silently inert because of it. Use Lexicon.default() for
  anything about shipped vocabulary.

Docs that need amending with the change

• rules.md#P2's "The final group reads as the family name" is false
  for every name this moves, and P2 has no interacts: line.
• rules.md#P1 and #P4 both say "one name word" where they mean one
  UNIT — see the conjunction point above.
• _post_rules.py carries a stale DEVIATION #364 comment.

Related

#365 is the same rule reached under FAMILY_FIRST and is still open.
#364 asked this question and was closed on an answer that has since
been superseded twice.