You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Should locales.PT turn on two_family_names? Maria Santos Silva reads family Silva under the pack, where Portugal's own boxes say Santos Silva #1453
decisions/O6 (two family names) records that locales.PT leaves the switch off "until a Portuguese given-name list exists" (2026-10-30, #753 phase 3g). This measures what such a list buys, and finds the list is not what is missing.
Measured 2026-11-18 on master c878443, ORCID holders' own given and family boxes (orcid-researchers), the first 3,000 records a country with both boxes filled, every box agreeing (local_agreement.read). The Portuguese list is the given-box words of the other 32,478 Portuguese and Brazilian records, at least 5 uses and at least 80% of uses in the given box, 677 words.
locales.PT with
Portugal
Brazil
Angola
Mozambique
nothing (today)
71.9%
50.2%
76.0%
85.5%
the switch
83.0%
74.3%
57.1%
50.6%
the switch and locales.ES_AR's list
89.4%
78.7%
66.7%
71.7%
the switch and the ORCID list
89.6%
79.3%
65.1%
66.2%
The ORCID list holds 168 words ES_AR's lacks (matheus, goncalo, vitoria, rute, vasco). Together the two score 89.5% in Portugal and 79.0% in Brazil, no better than either.
What is left, under both lists: a family box of three words or more, of which the switch takes two (Brazil 9.3% of records, Portugal 3.7%); and a second given name still taken (Brazil about 5.6%, Portugal about 3%).
Angola and Mozambique lose under every row with the switch on.
The ORCID list is one cut of four tried (share 80%, 90% or 95%; 5 or 10 uses), the best shown; its words were not read by hand, and no census source for Portugal or Brazil was looked for.
Questions.
Does locales.PT turn the switch on with the Spanish list as it stands, which gains 17 points in Portugal and 28 in Brazil, or does the Angolan and Mozambican loss keep it off for a pack chosen by language?
If on: is it locales.PT, or locales.PT_BR and a Portugal pack apart, as locales.ES_AR stands apart from locales.ES?
Does a pack carry another language's list? locales.ES_AR's is measured on Spain's census. The alternative is a Portuguese list from a source the project may measure (ORCID's boxes are CC0), which this run says would change little.
Is a family of three or more surnames a reading two_family_names should reach under a Portuguese choice, and by what signal? It is the larger remaining gap in Brazil.
Not asked here: whether Parser.suggest should propose the switch (#1452).
Activity