Albania (AL) — internal licensing notes

Source research, 2026-08-27 — sellability + reproducibility audit

GeoNames is the only source that is both legally sellable and reproducible. Every alternative was tested on 2026-08-27 and each fails on resale, on retrievability, or both. Findings below so this is not re-litigated.

The only usable source — GeoNames AL

https://download.geonames.org/export/zip/AL.zip       # 8.9 KB, 494 rows

Licence CC-BY-4.0, verified against export/zip/readme.txt on 2026-08-27 — Albania carries no per-country caveat (the readme flags only GB, CL, CN, IE, MT, AR, BR).

Ruled out — do not re-attempt without new evidence

If ASIG ever publishes a licensed bulk download of njësitë administrative, that is the upgrade path — it is the only remaining candidate that could be both official and sellable.

Defects in the currently-published file

al/al.csv on R2 is a GeoNames re-export (row count matches AL.zip exactly: 494) round-tripped through pandas and corrupted. All of these are fixable from GeoNames alone — no new source required:

1. admin_name3 is the literal string nan on every row. GeoNames leaves it empty; a pandas round-trip wrote the float NaN out as text.
2. admin_code2 is a GeoNames internal geonameId, not an official code. e.g. Bashkia Berat → 10944368. Confirmed against download.geonames.org/export/dump/admin2Codes.txt: all 61 bashkia are keyed by geonameId (AL.40.10944368), so this is GeoNames' own identifier and means nothing to a buyer. Either derive a real code or leave the column empty — do not ship an internal id as an admin code.
3. admin_code1 is the FIPS-style scheme (40, 41, 43, …), not ISO 3166-2:AL (AL-01…AL-12). Derive the ISO codes.
4. Integer accuracy written as float (4 → 4.0); 2 rows blank.
5. 494 rows but only 489 distinct postal codes — 5 duplicate codes; check against the primary key rule before publishing.

Reference structure for validating a rebuild: Albania has 12 counties (qark), 61 municipalities (bashki), 373–374 administrative units (njësi administrative). GeoNames covers tiers 1 and 2; tier 3 is unavailable from any sellable source (see the table above), so admin_name3 should ship empty, not as nan and not backfilled from a non-commercial source.

Attribution is absent from the sold product. R2 carries only al/al.csv — no README.md, datapackage.json, ATTRIBUTION.txt or al.zip. The repo datapackage.yml lists sources: [{title: "Multiple sources"}] with no licenses[]. CC BY 4.0 attribution is a licence condition — live breach on a paid product.

datapackage.yml also sets is_premium: true with no premium_tier and no premium_source_key, and no zip exists in R2 for a key to point at. At 494 rows against a 100-row public sample the premium value is thin — worth deciding deliberately whether AL should be premium at all.

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

{
  "channel": "postal-codes-r2",
  "components": [
    {
      "field_group": "all fields",
      "source_url": "https://download.geonames.org/export/zip/AL.zip",
      "licence_id": "CC-BY-4.0",
      "collected_at": "2026-08-27",
      "decision": "compatible",
      "decided_by": "claude-opus-5 (source research)",
      "decided_at": "2026-08-27"
    }
  ]
}

Proposed tier

B ($49). Single-source normalisation with nothing to reconcile. There is no second authoritative source to consolidate, so this cannot reach A on the criteria in docs/PRICING_TIERS.md — and it should not be priced as if it did.

Build — 2026-08-28 (local only, not published)

Built via datasets/al/scripts/build_postal_codes_product.py. 494 rows, validated clean with Frictionless.

Correction to this document's own earlier coverage claim: the "Reference structure" section above states Albania has 61 bashki and implies GeoNames reaches them; it does not say GeoNames covers all 61, but the build's first draft assumed exactly that and shipped it as "61/61 covered" in the README/ATTRIBUTION text before the raw source was re-checked. GeoNames' AL.zip only carries 31 of Albania's 61 bashki — verified directly by listing distinct admin_name2 values in the raw file. This is a real coverage gap in the source, disclosed in data/README.md and data/ATTRIBUTION.txt, not something to gloss over as "all bashki."

Other build decisions, beyond what this file already documented:

• admin_code1 derived via a verified ISO 3166-2:AL crosswalk (Wikipedia, checked 2026-08-28), keyed to the exact 12 admin_name1 strings GeoNames uses — not assumed from FIPS-code ordering.
• GeoNames' one inconsistent qark label ("Tirana", where the other 11 all read "Qarku i X") normalised to "Qarku i Tiranës" for consistency — disclosed as a correction, not silently changed.
• alternative_city_name re-derived fresh from GeoNames' gazetteer dump (283/494 rows populated) — same join method used for TH/RO, so the live file's 215-row figure is not lost.
• No duplicate (postal_code, place_name) pairs — confirmed directly; the "489 distinct postal codes" figure above is 5 codes shared across different place names, not a primary-key collision.

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