Moldova (MD) — internal licensing notes

Source research, 2026-09-03 — tier-1 check done, GeoNames used with real cleanup

Checked Poșta Moldovei (the national postal operator) first: posta.md offers only a post-office locator/search tool, no bulk code list or open data download found. No official open bulk source found — GeoNames used.

The source — GeoNames MD

https://download.geonames.org/export/zip/MD.zip      # 1,753 rows

Licence CC-BY-4.0, verified against export/zip/readme.txt on 2026-09-03 — Moldova carries no per-country caveat.

This is the messiest raw source in this session's batch – 49 distinct admin_name1 values instead of Moldova's real 37 ISO units, because the source mixes real district names, typo variants of the same district, municipality labels, and individually-named Transnistrian towns with no official code.

Verified directly against the raw source file:

• Typos, confirmed against the codes already present for the correctly- spelled version: "Cantrmir" → Cantemir (code 65 already used by "Cantemir"), "Canul" → shares Cahul's own code (64) directly in the raw data, "Uhnheni"/"Ungheni" → same district, one spelling has the code (92), the other doesn't.
• Municipality duplicates: "Balti" (coded 60) and "Mun.Balti" (uncoded) are the same municipality; same for "Mun.Chisinau".
• Transnistria: GeoNames labels breakaway-region towns individually with a "Tr." suffix (Bender, Camenca, Grigoriopol, Ribnita, Slobozia, Tiraspol) and none carry a code. ISO 3166-2:MD folds all of Transnistria into a single unit, MD-SN — Wikipedia's own page states this plainly: "due to limited international recognition, Transnistria is considered part of Moldova and does not have its own ISO 3166-1 code" (verified 2026-09-03). All "Tr."-suffixed rows mapped to MD-SN.
• Judgment call, disclosed in ATTRIBUTION.txt: Bender has its own ISO code (MD-BD), distinct from Transnistria, but is under real de facto Transnistrian control and the source labels it "Bender Tr." — mapped to MD-SN to reflect administrative reality over the de jure code. This means MD-BD is not represented in the shipped data at all; disclosed, not hidden.
• Dubasari raion is genuinely split in the source between "Dubasari Cr." (government-controlled, → MD-DU) and "Dubasari Tr." (Transnistrian-controlled, → MD-SN) — this reflects the raion's real administrative partition, not an error to "fix" by merging.
• "Comrat" (Gagauzia's capital) → MD-GA.
• 1 row genuinely unresolved: place_name "Lazo" at postal code MD-5219 has admin_name1 literally "Lazo" with no code in the raw source. A different "Lazo" locality (postal code MD-4222) is already correctly coded under Ștefan Vodă (MD-SV) — Moldova has more than one village named Lazo, so guessing which raion the uncoded one belongs to was judged too risky without further research. Shipped with admin_code1 blank, disclosed, not guessed.
• Known coverage gap, disclosed not fabricated: Basarabeasca (MD-BS) has zero rows in this source at all.
• 0 exact duplicate rows, 0 (postal_code, place_name) collisions, 0 literal "nan" strings.

licence_status

{
  "channel": "postal-codes-r2",
  "source_url": "https://download.geonames.org/export/zip/MD.zip",
  "licence_id": "CC-BY-4.0",
  "collected_at": "2026-09-03",
  "decision": "compatible",
  "decided_by": "claude-session-2026-09-03",
  "decided_at": "2026-09-03"
}

Proposed tier

B ($49). Genuinely the most involved cleanup pass in this session's batch — real typo consolidation, a disputed-territory administrative judgment call, and an honest unresolved-row disclosure, not a mechanical crosswalk.

Build — 2026-09-03

Built via datasets/md/scripts/build_postal_codes_product.py.

Frictionless validation: valid: True.

Status: built and validated locally only. Nothing published to R2, nothing committed to git yet.