Should a French choice read M. as a title, now that French data shows almost no initials? (M. Jean Dupont gives given 'M.') #1161

M. (Monsieur) is excluded in fr.toml (#658, decided 2026-11-03), under every choice including French. The reason: it's written exactly like an initial, and 782 ORCID records and 168 PatentsView inventors open with M. before two or more name words, every one an initial. Those counts are worldwide. French data looks different (2026-11-07):

• France's register of elected officials: 1 of 38,863 given-name fields opens with a lone letter.
• ORCID's agreement sample: 0 of 1,505 records from France open with M.; 2 of 2,622 from Belgium, Switzerland, Luxembourg and Monaco do (M. Ángeles, M. Malek).

So under a French choice the competing reading, M. as an initial, is rare. The exclusion's cost, M. Jean Dupont read as given 'M.', middle 'Jean', is borne by every French text that writes the salutation.

How it would work: a default = false salutation reading on fr.toml's m, replacing the exclusion, which a French choice turns on. No reading can apply only under strict=True: strictness is the caller's, never data (decisions/vocabulary-model, 2026-10-31). Plain and strict French choices would read it alike, because the competing reading is a shape (an initial), not another language's file. There's no locales.FR today, but a caller can make the choice: parser_for(Locale(code="fr", lexicon=Lexicon(), languages=frozenset({"fr"}))).

Questions:

• Is a French choice enough licence? The caller is stating the data is French, which is the evidence the worldwide counts lacked.
• What's the cost under that choice? M. Smith in French data, where M is an initial (Martin, Marie), would read title 'M.', given ''. How often do French sources write a bare initial plus a surname?
• Should this wait for a French prose source that writes the salutation (the Journal officiel's nominations, through DILA's open data)? That would measure the gain, not just the cost.