Should ben be a family-name particle in Maghrebi and Israeli surnames? (Karim Ben Salah gives middle 'Ben', family 'Salah') #1248

parse("Karim Ben Salah") gives given 'Karim', middle 'Ben', family 'Salah'. In Tunisian, Algerian and Moroccan names written in French spelling, and in Israeli names in Latin script, Ben X is very often the family name (Ben Salah, Ben Ammar, Ben David). ben has no reading today. It was left out on purpose: decisions/v1-xfail-triage (2026-09-01) keeps it out because of the given name Ben (Ahmad ben Husain reads middle 'ben'), and ar.toml records the exclusion. This issue reopens that call with measurements.

Measured (2026-11-08, config_diff.py score), with ben as an ambiguous particle, Lexicon.default().add("ben", kind="given").add("ben", kind="particle"), the way al reads:

• The given name Ben keeps its place where it opens a name: Ben Johnson still gives given 'Ben', now with a particle-or-given report. It moves when it stands in the middle: John Ben Smith gives family 'Ben Smith'. That is the exclusion's cost, and it is small: 1 broken row in ORCID GB, 0 in ORCID US, 2 in PatentsView US, each an English middle name Ben (John Ben Smith).
• Most of PatentsView's broken rows are filers who put Ben in the given box (given Dana Ben, family David). That's the filers' split, not the name's.
• A patronymic reading is worse. With add("ben", kind="patronymic") and no given reading, ben is never-given, so P1's fold (a never-given particle opening the name) reads Ben Davies as family 'Ben Davies': ORCID GB breaks 51 (96.39 to 96.14), ORCID US 20, PatentsView GB 33. It also fills R9's patronymic view with surnames.

The cost: Tunisian legal names with a ben chain. Official Tunisian lists write the name the way the civil register does: the given name, then ben + father, ben + grandfather, then the family name. Today ben stays unlisted, so the chain stays in the middle name and the family name is read right:

OpenSanctions' copy of Tunisia's counter-terrorism list (tn_cnlct, 158 people, given and family boxes separate, 148 given boxes holding a chain) reads 140 of 158 right today and 24 of 158 with ben listed. Self-written names (ORCID) almost never use the chain: ORCID TN's 3 broken rows are the only ones. This is the Latin counterpart of #1155 (a family name after a بن chain): the last ben of a chain joins forward and takes the family name with it.

Questions:

• List ben as an ambiguous particle? If so, in which file: ar.toml beside the exclusion, he.toml, or both?
• Does the chain need its own answer first? One option: a ben with another ben after it in the same chain is a patronymic link, and the word after the last link's name word is the family name. That's Should a family name after a بن chain stay out of the patronymic? (خالد بن سعد القحطاني gives patronymic 'بن سعد القحطاني') #1155's question in Latin.
• Should the lowercase writing (ben inside a mixed-case name) say anything? In tn_cnlct both cases are used in the chain, so case doesn't tell a chain link from a surname.