circ:cn-xinjiang-1 and circ:cn-xinjiang-2 are the seed's only ordinal-qualified IDs. Both carry the note "Two homonymous circumscriptions in the source index; qualifier pending committee review." This issue supplies the information needed to retire the ordinals, and asks which form replaces them.
The ordinals were fairly criticised in #1 as encoding arrival order in the source index rather than any fact about either circumscription. That criticism stands, and the fix is available.
What the source actually gives us
{"diocese_name": "Xinjiang", "diocese_id": "xinjia_cn"}
{"diocese_name": "Xinjiang", "diocese_id": "xinjin_cn"}Both rows are the bare string Xinjiang. No province field exists anywhere in the Chinese data, so rule 2's civil-region qualifier had nothing to draw on — which is why the generator fell back to ordinals. Xinjiang is the only duplicated name among China's 143 seeded entries. The differing diocese_ids are upstream collision-avoidance artefacts, not distinguishing information.
They are not, in fact, homonyms
The two jurisdictions have different Chinese names, differing in both characters and tone:
|
northwestern |
Shanxi |
| Chinese |
新疆 |
新绛 |
| Pinyin with tones |
Xīnjiāng |
Xīnjiàng |
| Civil location |
Xinjiang Uyghur Autonomous Region |
Xinjiang County, Yuncheng, Shanxi Province |
| Episcopal see |
Ürümqi (Cathedral of the Immaculate Conception) |
Xinjiang County (Cathedral of St. Anthony) |
| Canonical rank |
Apostolic Prefecture |
Apostolic Prefecture |
| Also known as |
Xinjiang–Urumqi |
Jiangzhou / Yuncheng |
| History |
Mission sui iuris split from Lanzhou 1930; Apostolic Prefecture 1938 |
Erected 1936 as Apostolic Prefecture of Jiangzhou/Xinjiang |
The collision is manufactured by tone-stripped romanization. These names are distinct in Chinese and distinct in fully-marked pinyin; they merge only after tones are dropped and the result is ASCII-folded. That matters beyond this pair — see the general question at the bottom.
Both are apostolic prefectures, not dioceses. The source index calls everything a diocese, so type is wrong-by-omission for both; apostolic_prefecture is already in the planned enumeration in docs/schema-proposal.md.
Options
A. Apply the existing rule properly — no qualifier needed. The schema defines <slug> as "the see name (not the full styled title)", with boston and roma as examples: the see city. The northwestern prefecture's see is Ürümqi; the Shanxi prefecture's see is Xinjiang County. Read that way there was never a collision:
circ:cn-urumqi — 新疆, see at Ürümqi
circ:cn-xinjiang — 新绛, see at Xinjiang County, Shanxi
Both slugs are currently free. This needs no new rule, no ordinal, and no qualifier. Cost: circ:cn-urumqi is not the jurisdiction's official title, which is Xinjiang, so the ID diverges from the name a reader would look up.
B. Official compound usage. Rule 2 says to follow "official usage where it exists", and both are commonly listed in compound form:
circ:cn-xinjiang-urumqi
circ:cn-xinjiang-jiangzhou (or -yuncheng)
Symmetric, keeps Xinjiang visible in both, and matches how reference works list them. Cost: encodes a compound that the Annuario may not use verbatim.
C. Rule 2 literally — civil region qualifier. Degenerate for the northwestern one, whose civil region is Xinjiang:
circ:cn-xinjiang-in-xinjiang ← unusable
circ:cn-xinjiang-in-shanxi
Listed for completeness; the asymmetry means rule 2 alone cannot resolve this pair.
D. Asymmetric mix. circ:cn-urumqi for the northwestern, circ:cn-xinjiang-in-shanxi for the Shanxi one — each disambiguated by whatever works for it. Cost: two different mechanisms for one pair.
My reading is that A is the strongest: it is the existing rule applied correctly rather than a new exception, and it makes the ordinals unnecessary rather than merely replacing them. B is the better answer if the committee decides the identifier should track the jurisdiction's official title rather than its see city — which is a question rule 2 does not currently settle and probably should.
Whichever is chosen
Both entries go into the MANUAL override table in scripts/generate_seed.py, replacing the current ordinal entries, with type: apostolic_prefecture and a note recording the Chinese name and civil location — the upstream index cannot distinguish these rows, so the resolution has to live here.
The historical details above should be confirmed against the Annuario Pontificio before merge; they are offered as the basis for the decision, not as sourced registry data.
General question this raises
The seed romanizes 143 Chinese circumscriptions from a tone-free source, and slugify() then ASCII-folds. Two distinct names collided here and were caught because they collided exactly. The same mechanism can silently merge or confuse other pairs, and applies to any tonal language in scope.
Should the registry carry the native-script name (新疆 / 新绛) as a field, so that identity is verifiable in the source language rather than resting on a lossy romanization? That would also give the multilingual label layer discussed in #1 something authoritative to hold.
Related: #1 (identifier stability — this pair is one of its three cases), #3, #4, #5. Rule 2 and open question 4 in docs/schema-proposal.md.
circ:cn-xinjiang-1andcirc:cn-xinjiang-2are the seed's only ordinal-qualified IDs. Both carry the note "Two homonymous circumscriptions in the source index; qualifier pending committee review." This issue supplies the information needed to retire the ordinals, and asks which form replaces them.The ordinals were fairly criticised in #1 as encoding arrival order in the source index rather than any fact about either circumscription. That criticism stands, and the fix is available.
What the source actually gives us
{"diocese_name": "Xinjiang", "diocese_id": "xinjia_cn"} {"diocese_name": "Xinjiang", "diocese_id": "xinjin_cn"}Both rows are the bare string
Xinjiang. Noprovincefield exists anywhere in the Chinese data, so rule 2's civil-region qualifier had nothing to draw on — which is why the generator fell back to ordinals.Xinjiangis the only duplicated name among China's 143 seeded entries. The differingdiocese_ids are upstream collision-avoidance artefacts, not distinguishing information.They are not, in fact, homonyms
The two jurisdictions have different Chinese names, differing in both characters and tone:
The collision is manufactured by tone-stripped romanization. These names are distinct in Chinese and distinct in fully-marked pinyin; they merge only after tones are dropped and the result is ASCII-folded. That matters beyond this pair — see the general question at the bottom.
Both are apostolic prefectures, not dioceses. The source index calls everything a diocese, so
typeis wrong-by-omission for both;apostolic_prefectureis already in the planned enumeration indocs/schema-proposal.md.Options
A. Apply the existing rule properly — no qualifier needed. The schema defines
<slug>as "the see name (not the full styled title)", withbostonandromaas examples: the see city. The northwestern prefecture's see is Ürümqi; the Shanxi prefecture's see is Xinjiang County. Read that way there was never a collision:Both slugs are currently free. This needs no new rule, no ordinal, and no qualifier. Cost:
circ:cn-urumqiis not the jurisdiction's official title, which is Xinjiang, so the ID diverges from the name a reader would look up.B. Official compound usage. Rule 2 says to follow "official usage where it exists", and both are commonly listed in compound form:
Symmetric, keeps Xinjiang visible in both, and matches how reference works list them. Cost: encodes a compound that the Annuario may not use verbatim.
C. Rule 2 literally — civil region qualifier. Degenerate for the northwestern one, whose civil region is Xinjiang:
Listed for completeness; the asymmetry means rule 2 alone cannot resolve this pair.
D. Asymmetric mix.
circ:cn-urumqifor the northwestern,circ:cn-xinjiang-in-shanxifor the Shanxi one — each disambiguated by whatever works for it. Cost: two different mechanisms for one pair.My reading is that A is the strongest: it is the existing rule applied correctly rather than a new exception, and it makes the ordinals unnecessary rather than merely replacing them. B is the better answer if the committee decides the identifier should track the jurisdiction's official title rather than its see city — which is a question rule 2 does not currently settle and probably should.
Whichever is chosen
Both entries go into the
MANUALoverride table inscripts/generate_seed.py, replacing the current ordinal entries, withtype: apostolic_prefectureand a note recording the Chinese name and civil location — the upstream index cannot distinguish these rows, so the resolution has to live here.The historical details above should be confirmed against the Annuario Pontificio before merge; they are offered as the basis for the decision, not as sourced registry data.
General question this raises
The seed romanizes 143 Chinese circumscriptions from a tone-free source, and
slugify()then ASCII-folds. Two distinct names collided here and were caught because they collided exactly. The same mechanism can silently merge or confuse other pairs, and applies to any tonal language in scope.Should the registry carry the native-script name (
新疆/新绛) as a field, so that identity is verifiable in the source language rather than resting on a lossy romanization? That would also give the multilingual label layer discussed in #1 something authoritative to hold.Related: #1 (identifier stability — this pair is one of its three cases), #3, #4, #5. Rule 2 and open question 4 in
docs/schema-proposal.md.