Malta (MT) — internal licensing notes

Source research, 2026-09-11 — tier-1 check done before considering GeoNames

Checked MaltaPost (the official national postal operator) first: a real official postcode finder exists at maltapost.com/PostCode/, but it is a one-address-at-a-time lookup tool — no bulk export, CSV, or API found.

Why GeoNames is NOT a usable fallback here (unlike most other countries)

Fetched GeoNames MT directly (download.geonames.org/export/zip/MT.zip) to check: only 73 rows, and GeoNames' own readme.txt states explicitly: "For Malta we have only the first letters of the full postal codes (for copyright reasons)" — confirmed directly (rows look like ATD, BZN, BRG, locality-level prefixes only, not the real AAA 9999 format).

Malta's real postal codes are alphanumeric, AAA 9999 (3 letters + 4 digits), introduced 1991, revised 2007. Independent estimate: ~27,426 real postal codes across ~74 localities (GeoPostcodes' public figure, not independently verified here, cited only for scale). GeoNames' 73-row truncated file represents a ~99.7% coverage loss versus that scale — the same class of unusable-fallback finding as Jersey's GeoNames export earlier in this project.

What this means

Unlike JP/CZ/etc. in this batch, GeoNames is not even a viable fallback for Malta — it's deliberately truncated at the source for copyright reasons, not just incomplete. The only real official source (MaltaPost's lookup tool) is not bulk-fetchable with what was checked in this pass.

Verdict

No build in this pass. A real bulk source was not found — official (MaltaPost, lookup-only) and fallback (GeoNames, deliberately truncated) both fail. Options for a future pass: contact MaltaPost directly for a bulk export, or investigate whether the site's lookup tool has an underlying API that could be queried systematically (would need a separate ToS/systematic-extraction check before attempting, same diligence as the OSM/Nominatim gate elsewhere in this project). data_available unchanged (false).

Re-check, 2026-09-15 — fresh search, conclusion unchanged

Re-ran source discovery from scratch (not just re-reading the above) per a request to conclusively confirm or overturn the 2026-09-11 verdict. Checked, this pass:

1. MaltaPost (maltapost.com/postcode/) — fetched directly again. The page is a JS-rendered SPA with an empty static body (no server-rendered content, no exposed API endpoint found in the static HTML). Still no bulk export, open-data page, or documented API. Unchanged from the prior pass.
2. NSO Malta (nso.gov.mt) — site blocks automated fetches (Cloudflare 403 to both WebFetch and curl). Web search around it surfaced nothing suggesting NSO publishes postal codes; NSO's own publications are census/economic/demographic, not postal geography.
3. data.gov.mt / portal.data.gov.mt — also Cloudflare-blocked to automated tools (confirmed via curl: literal Cloudflare "Attention Required" block page). site:portal.data.gov.mt web search for postal/postcode/localities returned nothing relevant.
4. Malta GeoHub (geohub.gov.mt), run by the Planning Authority — this one IS reachable and has a real public DCAT-US API (/api/feed/dcat-us/1.1.json). Pulled the full catalog: 63 real datasets (schools, bus stops, health centres, protected areas, local council boundaries, etc.). No postal-code or address-point dataset of any kind. Malta's INSPIRE "Addresses" (AD) theme, which EU member states are formally obligated to publish, was searched for specifically (MSDI portal, INSPIRE Malta ArcGIS hub inspire-catalogue-planningmt.hub.arcgis.com) — MSDI's public offering is environmental/INSPIRE themes only (terrain, protected areas, water bodies); no Addresses theme dataset was found exposed through either portal's public catalog/API.
5. GitHub — searched again for a complete AAA 9999-format Malta postcode repo. Found none; only Malaysia postcode repos surface for generic "postcode" GitHub searches.
6. Re-verified the aggregators are all the same GeoNames file — fetched postal.datahub.io/mt/mt.csv (the live resource behind datasets/mt/datapackage.yml in this repo, i.e. our own existing legacy listing) directly: 73 data rows, locality-prefix codes (ATD, BZN, BRG, …), identical in shape and content to the GeoNames MT.zip export checked in the prior pass. postalcodes.info independently advertises "73 geocoded records" for Malta — same number, confirming it's the same underlying GeoNames extract resold, not an independent or more-complete source. GeoPostcodes.com advertises 27,426 Malta codes but is a paid commercial product with no accessible sample large enough to verify, and buying/relicensing a competitor's commercial extract is out of scope for a source- discovery pass.

No new source clears the bar (order-of-thousands rows, official or clearly-licensed, distinguishable from the deliberately-truncated GeoNames extract). Conclusion from 2026-09-11 stands unchanged: do not build. data_available remains false. datasets/mt/datapackage.yml was not touched.

Note for a future pass, sharper than the prior one: the one lead that hasn't been run to ground is whether MaltaPost's postcode-finder page calls an internal JSON API per-lookup that could — separately from this pass, and only after an explicit ToS/systematic-extraction review like the Nominatim gate elsewhere in this project — be queried address-by- address rather than scraped in bulk. Not attempted here: no ToS permission was found for systematic querying, and the page's JS bundle was not decompiled to find the endpoint.

Build — EXECUTED 2026-09-15, per explicit instruction to build from GeoNames anyway

The 2026-09-11/15 verdict above was that GeoNames is NOT a usable fallback because it's deliberately truncated to locality-prefix level. On explicit instruction, this was reversed: build it anyway, with the truncation disclosed prominently rather than treated as disqualifying.

Result: 73 rows, all 68/68 ISO 3166-2:MT local councils covered. postal_code = the 3-letter GeoNames locality prefix (e.g. ATD) — explicitly NOT a complete deliverable Malta postcode (real codes are AAA 9999; the 4-digit suffix does not exist in this source at all). This is disclosed in the dataset title, description, README, and ATTRIBUTION, not silently shipped as if complete.

admin_code1 required a genuinely careful crosswalk, not a numeric prepend: GeoNames' own admin_code1 ("01".."68") does NOT match ISO 3166-2:MT order past code 10 (e.g. GeoNames codes Mdina "20", ISO assigns Mdina MT-29). Two real island-disambiguation traps were resolved by name + latitude, not guessed: GeoNames' "Rabat" (Malta mainland) = ISO's "Rabat Malta" (MT-46); GeoNames' "Victoria" (Gozo) = ISO's "Rabat Għawdex" (MT-45, Victoria being the modern common name for the same place). GeoNames' two "Żebbuġ" entries (Malta mainland vs Gozo) were resolved the same way. Two English-name variants required care, not swapping: "Saint Julians" = ISO's "San Ġiljan" (MT-48); "Saint Paul's Bay" = ISO's "San Pawl il-Baħar" (MT-51). Also found and fixed a build-script bug mid-pass: GeoNames' own admin_name1 field turned out to be a Maltese-language restatement of place_name (not a real council- grouping field) — the actual stable grouping key is GeoNames' own numeric admin_code1, verified unambiguous by grouping every raw row by it before building the crosswalk.

Frictionless: valid: True. licence_status.decision: compatible (clean CC BY 4.0, no licence ambiguity here — the issue was always completeness, never rights).

Tier decision: NOT premium. Per docs/PRICING_TIERS.md's own rule ("is_premium is a separate flag… only when the full version is genuinely larger than the published 100-row sample… PR #24 stripped this from 41 datasets of 1–99 rows"), Malta's 73-row full dataset is smaller than a 100-row sample would be — the sample IS the whole dataset. is_premium left unset, no premium_tier.

Published: git (datasets/mt/{datapackage.yml,scripts/}), main R2 bucket (mt/{mt.csv,datapackage.json,README.md,ATTRIBUTION.txt,mt.zip}, byte-verified), samples bucket (full 73-row file, mt/mt.csv), and _meta/catalog.json (backed up, mutated, re-verified at 248 entries).