Personal ordinariates: two of three are supranational under rule 5, and all three are styled differently

[JohnRDOrazio]

The three personal ordinariates erected under Anglicanorum coetibus are all present in the seed, but all three are styled inconsistently with each other, and two of the three have a nation that rule 5 says they should not have.

Split out of #4, which settles slug language for military ordinariates. This cohort is a different problem and should not inherit that decision by default.

Current state

entry	nation	province
circ:us-personal-ordinariate-of-the-chair-of-st-peter	US	Texas
circ:gb-our-lady-of-walsingham	GB	—
circ:au-our-lady-of-the-southern-cross	AU	—

Problem 1 — two of the three are supranational

Rule 5 reserves the int segment for structures spanning more than one nation, and circ:int-opus-dei already applies it. Verified extents:

• Chair of Saint Peter — the United States and Canada, from its erection in 2012. Seeded as circ:us- with nation: US.
• Our Lady of the Southern Cross — Australia, Guam, Japan, New Zealand and the Philippines. Seeded as circ:au- with nation: AU.
• Our Lady of Walsingham — England, Wales and Scotland; all within the United Kingdom. circ:gb- with nation: GB is correct and needs no change.

So two of the three are in exactly the position Opus Dei was in before 5067fb8: a supranational structure carrying the country code of the nation where its chancery happens to sit. Under rule 5 these become circ:int-chair-of-st-peter and circ:int-our-lady-of-the-southern-cross, with nation: null.

The Southern Cross case is the starker of the two — its jurisdiction reaches across five territories in two hemispheres, and au asserts something the entity is not.

Problem 2 — province: Texas is not a province

circ:us-personal-ordinariate-of-the-chair-of-st-peter carries province: Texas. That is where the chancery sits (Houston), not a civil subdivision of its jurisdiction — the ordinariate's territory is two entire countries. province elsewhere in the seed narrows a circumscription's extent, so this reads as a claim that the ordinariate is confined to Texas. It should be dropped when the entry moves to int.

Problem 3 — the three are styled three different ways

circ:us-personal-ordinariate-of-the-chair-of-st-peter    type word + article + dedication
circ:gb-our-lady-of-walsingham                           dedication only
circ:au-our-lady-of-the-southern-cross                   dedication only

Two carry only the dedication; one carries the full styled title. Under rule 1 the type word should not be in the slug — and unlike military ordinariates (#4), these genuinely do have a name beneath the type word, so there is no rule 5 justification for keeping it. circ:int-chair-of-st-peter is the consistent form.

Worth noting this cohort is also why the sweep in #4 undercounted: an ordinariate whose slug is pure dedication matches no type-word search. Any future audit of "which entries carry a type word" needs to run off type once it is populated, not off slug pattern matching.

Problem 4 — type is null for all three

type is null throughout the seed pending an authoritative pass, but these three have an unambiguous value already available: personal_ordinariate, which is in the planned enumeration in docs/schema-proposal.md. They could be typed in the same pass that resolves the IDs.

Proposed resolution

1. circ:us-personal-ordinariate-of-the-chair-of-st-peter → circ:int-chair-of-st-peter, nation: null, drop province, type: personal_ordinariate
2. circ:au-our-lady-of-the-southern-cross → circ:int-our-lady-of-the-southern-cross, nation: null, type: personal_ordinariate
3. circ:gb-our-lady-of-walsingham — ID unchanged, type: personal_ordinariate

All three would go in the MANUAL override table in scripts/generate_seed.py alongside Opus Dei, with a note recording the extent and the rule 5 basis, as that entry already does.

Open question this raises

Rule 5's int token flattens two distinct situations: a structure with no territorial nation (Opus Dei, worldwide) and a structure spanning a specific, enumerable set of nations (Chair of St Peter → US + CA; Southern Cross → AU + GU + JP + NZ + PH). Both become int, and the difference is then only recoverable from an attribute.

Is that acceptable, or should the multi-national case carry the list of nations in a nations array while nation stays null? The latter costs nothing at the identifier level and makes "which circumscriptions operate in Japan?" answerable, which it currently is not.

Related: #1 (identifier stability), #3 (territorial prelatures and abbacies), #4 (military ordinariate slug language), rule 5 and open question 2 in docs/schema-proposal.md.

Sources for the territorial extents: Chair of Saint Peter · CCCB · Our Lady of the Southern Cross · Our Lady of Walsingham · CBCEW