Hungary (HU) — internal licensing notes

Source research, 2026-08-29

GeoNames is the only source with a verified open licence. The official Magyar Posta file exists and is directly downloadable, but its redistribution terms could not be confirmed — treat as unresolved, not assumed clean.

GeoNames HU postal export — verified clean, but admin2 is thin

https://download.geonames.org/export/zip/HU.zip   # 3,571 rows

Licence CC BY 4.0, verified against export/zip/readme.txt on 2026-08-29 — no per-country caveat for Hungary. admin_name1 (megye/county) filled on all 3,571 rows, 20 distinct — matches Hungary's real structure (19 counties

• Budapest = 20) exactly. admin_name2 (járás/district) is EMPTY on every single row (0/3,571) — a genuine gap in the postal export, not a formatting defect. No primary-key collisions, but 3,046 distinct codes across 3,571 rows (some codes legitimately shared by multiple place names).

Checked whether the same-source gazetteer-join trick (proven for South Africa this session) would fix the admin2 gap — it would not, meaningfully. GeoNames' general dump for Hungary (export/dump/HU.zip) has admin2 codes populated on only 83 of 16,020 populated-place entries (56 distinct codes, out of Hungary's real ~174 járás) — far too sparse to close this gap by joining. This is a real, disclosed coverage limitation of GeoNames for Hungary specifically, not something a build can engineer around within this same source family.

Official source — Magyar Posta irányítószám adatbázis

https://www.posta.hu/static/internet/download/Iranyitoszam-Internet_uj.xlsx

Listed on Magyar Posta's own site under "közérdekű adatok" (public-interest / freedom-of-information disclosure data), not under an explicit open-data licence grant. Hungarian FOI-disclosure obligations do not automatically carry a commercial-redistribution right — this is the same shape of gap already treated as unresolved elsewhere in this catalogue (Tanzania, Mauritius, Vietnam, Senegal): a file is published, but no licence text was found addressing reuse or resale. Not checked further this pass (no ToS/ licence page found in the search results); would need a direct read of Magyar Posta's site terms before this could be upgraded past unresolved.

KSH (Központi Statisztikai Hivatal) — mixed, does not clearly help either

KSH's own generated website content (tables, figures) is CC BY 4.0. But its own copyright page explicitly EXCLUDES data extracted from KSH's internal databases on individual request from that licence, stating users "do not have rights for commercial use of such datasets." The settlement registry (helységnévtár, ksh.hu/docs/helysegnevtar/hnt_letoltes_2025.xlsx) is exactly this kind of extracted-on-request database file, not the general website content — so the CC BY 4.0 grant likely does not cover it. Treat as unresolved, same reasoning as Magyar Posta above.

licence_status (proposed — needs human sign-off before shipping)

{
  "channel": "postal-codes-r2",
  "source_url": "https://download.geonames.org/export/zip/HU.zip",
  "licence_id": "CC-BY-4.0",
  "collected_at": "2026-08-29",
  "decision": "compatible",
  "decided_by": "claude-session-2026-08-29",
  "decided_at": "2026-08-29",
  "note": "Covers county (admin_name1) only. Magyar Posta and KSH sources for a real district (jaras) tier were both checked and left unresolved -- no verified open licence found for either, and neither should be scraped for a paid product until one is confirmed."
}

Proposed tier

B ($49). Single source (GeoNames) normalised in place, county-tier only.

Verdict

Build it as a county-only product on GeoNames, OR pause and get a real answer on Magyar Posta/KSH's terms first if district-level coverage is wanted. No licensing obstacle to shipping the GeoNames-only version; the admin2 gap is real and disclosed, not fixable within this source family.

Build — 2026-08-30 (local only, not published)

Built via datasets/hu/scripts/build_postal_codes_product.py. Confirmed directly: admin_name2 is empty on all 3,571 rows (no county-only build compromise, this is the actual source shape). admin_code1 turned out to need no crosswalk table lookup beyond a simple prefix: GeoNames' own 2-letter county codes (BA, BZ, BU, …) are IDENTICAL to the real ISO 3166-2:HU codes (verified against Wikipedia, all 20 match) — just add "HU-". Kept GeoNames' pre-2020 name "Csongrad" as the source's own text rather than silently renaming to the 2020-updated "Csongrad-Csanad"; the ISO code (HU-CS) is unaffected either way.

Frictionless validation: valid: True.

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