Ukraine (UA) — internal licensing notes

Source research, 2026-08-31

Clean GO on GeoNames; no official bulk source exists; real admin-tier recovery is available via the same-source dump-join technique (RS/ZA pattern).

What's currently live

R2's ua/ua.csv (29,571 rows, matches _meta/catalog.json's stale 2026-07-30 entry) is a near-raw GeoNames postal export, corrupted the same way as PL/AL/TH/RS/BG: admin_name3 is the literal string "nan" on every row (verified: 29,571/29,571), and accuracy is float-formatted ("4.0" instead of "4") — both classic pandas round-trip artifacts. admin_code1 is GeoNames' own internal numbering ("01"–"27"), not ISO 3166-2:UA — using it as-is would ship the wrong codes. datapackage.yml's declared resource (7,465 bytes, dated 2024-11-07) doesn't match the live object (2,373,590 bytes) either — stale metadata on top of the data defect, same pattern as every other legacy placeholder audited this batch.

Source 1 — GeoNames UA postal export (the base)

https://download.geonames.org/export/zip/UA.zip     # 29,571 rows

Licence CC BY 4.0, verified against export/zip/readme.txt on 2026-08-31 — no per-country caveat for Ukraine.

Real limitation, verified directly against the raw file (not the sold copy): the postal export itself has no raion (district) tier at all — admin_name2 simply duplicates admin_name1 (the oblast name) on every row, and admin_code2/admin_name3/admin_code3 are blank. This is a genuine source gap, not something introduced downstream.

Source 2 — GeoNames UA gazetteer dump (admin enrichment, same publisher)

https://download.geonames.org/export/dump/UA.zip    # 45,380 entries

Same licence, same publisher, not a new source. Carries GeoNames' own admin1 (27 codes: 24 oblasts + Crimea + Kyiv City + Sevastopol City) and admin2 (raion-level, 146 codes) fields on 33,505 of its own populated-place entries.

Real finding, cross-checked against a primary account of Ukraine's own 2020 reform (Verkhovna Rada resolution, 2020-07-17, reported by The Ukrainian Weekly and KyivPost): Ukraine consolidated 490 old raions into 136 new raions (including 10 in Crimea, de facto outside Ukrainian control since 2014). GeoNames' dump carries 146 UA admin2 codes — exactly 136 (the new national raion count) + 10 (Kyiv City's own internal city districts, a separate tier that sits under Kyiv City rather than being one of the 136 national raions). This is strong, specific evidence the dump reflects the post-reform structure, not a stale pre-2020 one — the opposite risk to the one caught for Norway's fylke reform, checked here before trusting the dump at scale.

Join feasibility, checked before committing to a build

The postal export's place names are Cyrillic (Черкаси); the dump's name/asciiname fields are Latin transliterations (Cherkasy) — a naive name-to-name join against name/asciiname alone matches zero rows. The dump's alternatenames column does carry the Cyrillic form as one of several comma-separated aliases. Matching against name ∪ asciiname ∪ alternatenames recovers 22,224/29,571 (75%) exact matches before any coordinate disambiguation or nearest-neighbour fallback (the RS/ZA pattern) — a workable base for a future build, though this needs the fallback pass run and measured (not yet done — this is a research pass, not a build) before a real coverage number can be quoted.

❌ Ukrposhta — no bulk source, not a licensing block

No official Ukrposhta open-data bulk download or API was found (checked 2026-08-31) — same shape as Serbia's Posta Srbije and Poland's situation before its PDF-only book edition: real-world lookup exists (services.ukrposhta.com/postindex_new/) but nothing bulk/open. This is a data-availability gap, not a resale prohibition.

ISO 3166-2:UA — GeoNames' internal numbering is NOT the ISO scheme

Verified against en.wikipedia.org/wiki/ISO_3166-2:UA on 2026-08-31. GeoNames' admin1 codes (01–27) are an internal, unrelated numbering. A verified name-based crosswalk (all 27 GeoNames admin1 names match one ISO entry each, zero misses):

Disclosure needed for any future build — factual, not political

Crimea (UA-43) and Sevastopol City (UA-40) are Ukrainian territory under ISO 3166 and under Ukraine's own law, and are included on that basis — same treatment this project already gives other administratively-defined regions. But they have been outside actual Ukrainian civil administration since Russia's 2014 annexation, and parts of Donetsk (UA-14), Luhansk (UA-09), Zaporizhzhia (UA-23) and Kherson (UA-65) oblasts have been under Russian occupation since 2022. This is a real postal-service currency question, not an administrative-boundary question: any rows in occupied territory should be labelled as sourced from GeoNames' (pre-war) postal data rather than implied to reflect current, operating Ukrainian postal service in those areas. Disclose in README/ATTRIBUTION if built — do not omit the regions (that would misrepresent Ukraine's actual administrative structure) and do not silently imply full current postal operability either.

licence_status

{
  "channel": "postal-codes-r2",
  "source_url": "https://download.geonames.org/export/zip/UA.zip",
  "licence_id": "CC-BY-4.0",
  "collected_at": "2026-08-31",
  "decision": "compatible",
  "decided_by": "claude-session-2026-08-31",
  "decided_at": "2026-08-31",
  "note": "No official bulk source exists to consolidate onto GeoNames; admin enrichment is same-publisher/same-licence GeoNames-to-GeoNames (postal export + gazetteer dump), not a second source with its own terms."
}

Proposed tier

B ($49). Single publisher (GeoNames, both exports) normalised in place with real derivation work — same-source raion-tier recovery via the dump join, verified ISO 3166-2:UA crosswalk (GeoNames' numbering is unrelated to ISO's), and fixing the "nan"/float-accuracy corruption — but no independent second-source consolidation, so this doesn't clear tier A's bar per docs/PRICING_TIERS.md. Same shape as RS/ZA/BG.

Moat

Low-Medium. GeoNames is openly downloadable, and the dump-join technique is now standard in this catalogue (RS/ZA). What's not entirely trivial to redo: catching the Cyrillic/Latin naming mismatch that zeroes out a naive join, and independently verifying the dump reflects the post-2020 raion reform (146 = 136 + Kyiv's 10 city districts) rather than assuming it and shipping a stale structure.

Build — 2026-08-31 (local only, not published)

Built via datasets/ua/scripts/build_postal_codes_product.py. Results, measured directly:

Frictionless validation: valid: True.

Status: built and validated locally only. Nothing published to R2, nothing committed to git. datasets/ua/README.md and datapackage.yml updated in place (tier B, premium_source_key: ua/ua.zip, real sources/licenses, replacing the "Multiple sources" placeholder).