Jersey (JE) — internal licensing notes

Status: NOT live, correctly gated. No clean path exists without contacting a rightsholder — and the human has instructed not to. Reviewed 2026-08-24 under datahub-next#941 §2.

Product identity

• R2 je/je.csv is 508 bytes — not a 0-row placeholder, but not a real product either (a small sample, not the full build). No is_premium. Nothing meaningful is on sale.
• The real build (3,235 rows) lives in fedex-agents-prototype/countries/JEY/. No Jersey_Delivery/ folder exists — never sold to a customer.

The issue's specific claim was wrong, but the real problem is worse

datahub-next#941 §2 said: "No parish/vingtaine boundary dataset found on opendata.gov.je." False — verified directly: opendata.gov.je hosts a 2011-census vingtaine dataset (55 vingtaines with parish mapping) under OGL-Jersey v1.0, which states plainly: "The Chief Minister offers the information hosted on data.je for re-use… There is no need to apply for licence." That layer is genuinely clean and contact-free.

But the actual build doesn't use that. Traced 2026-08-24: the real parish/ vingtaine polygons come from a Jersey Government ArcGIS FeatureServer (services-eu1.arcgis.com/mUgkK1QjC37PAGmP/.../Parish_Boundaries and .../Vingtaine_Boundaries). Queried its ?f=json metadata directly: licenseInfo and copyrightText are both empty strings. No licence declared at the source. Probably the same Jersey Government infrastructure as the clean OGL-J portal — but that is an inference, not a confirmed fact, and confirming it requires asking Jersey's GIS team, which is off the table per instruction.

The postcode layer (3,235 rows) is the harder blocker, and re-sourcing without contact is not possible:

• Current sourcing: freemaptools.com's UK postcode dump (filtered to JE%)
  • britishredcrosssociety/channel-islands-geocoding on GitHub, which has no LICENSE file — GitHub's default is all-rights-reserved, not a redistribution grant.
• Code-Point Open does not cover this. Verified directly against Ordnance Survey's own product page: "An open dataset of all the current postcode units in Great Britain" — explicitly GB only (England/Scotland/Wales). Jersey is a Crown Dependency, not part of Great Britain, and is excluded. This rules out the fix the original review recommended for the UK build.
• GeoNames JE exists but is not a usable replacement. Fetched and checked directly: 5 rows, 4 distinct postcodes — just the outward codes JE1–JE4, no admin3, no street/full-postcode granularity. Our build has 3,235 full postcodes. Adopting GeoNames JE would be a ~99.9% coverage loss, not a rebuild.
• The rightsholder is Jersey Post / its appointed data partners (Digimap (Jersey) Ltd, Europa Technologies) — licensing from them is the only path found, and that requires contact.

Coordinates currently come from Nominatim/OpenStreetMap — same ODbL + Nominatim-policy problem documented for Singapore and Kuwait elsewhere in this review (see the OSM/ODbL gate in CLAUDE.md). A third, independent reason this can't ship as-is even before the postcode-source question is settled.

Recommendation

No action possible under the current constraint (no contact). Keep data_available: false. If the no-contact rule is ever relaxed for a single case, Jersey is the weaker candidate relative to Senegal (see datasets/sn/.licensing-notes.md) — it stacks three separate problems (unlicensed aggregator source, unconfirmed ArcGIS boundary licence, ODbL coordinates) rather than one clean gap.

Not legal advice — a research summary of published terms and verified source availability.

Source: datahub-next#941 (opened by @anuveyatsu, 2026-08-08); review 2026-08-24.

GENERATED AND PUBLISHED — 2026-08-25

Human decision, overriding the "no action possible without contact" hold above: generate and publish now, resolve licensing after. je.csv (3,235 rows) built from the FedEx-schema QA-passed dataset via datasets/je/scripts/build_postal_codes_product.py. Two of the three stacked problems documented above are avoided by construction, not resolved:

• No coordinates are included. The FedEx source has none, so this release ships with latitude/longitude/accuracy blank — the Nominatim/OSM/ODbL problem this file describes is real for a future coordinate-enrichment pass, but is not reproduced in what's shipping now.
• The admin layer (parish/vingtaine names) uses the clean OGL-Jersey census source and/or Wikipedia, not the unclear-licence ArcGIS FeatureServer this note flags — that server's polygons were never needed for a name-only build.

What is NOT avoided, and remains the real disclosed gap: the postal-code layer itself (freemaptools UK dump + the britishredcrosssociety GitHub table with no LICENSE file). licence_status.decision is explicitly "unresolved" — a deliberate, disclosed exception to the provenance gate's hard-block rule. Full disclosure ships in ATTRIBUTION.txt and the README, with an explicit "do not resell without a redistribution right" line. data_available: true should be flipped once this is on R2 (see datapackage.yml).

Still true: no contact has been made with Jersey Post or its data partners. This publish does not resolve the underlying licence question for the postcode layer — it makes an informed decision to sell into that gap, disclosed, rather than wait on it.

Coordinates re-checked, still blank — 2026-09-03

Re-verified the coordinate gap before deciding whether to fill it. GeoNames JE (freshly re-fetched) has only 2 distinct coordinate points across all 4 outward codes (JE1/JE2/JE3 share one identical point, JE4 has the other), with no accuracy tier claimed even by GeoNames itself. Filling all 3,235 rows with 2 stretched points would look like real per-row geocoding when it isn't — worse than the disclosed blank currently shipped. Nominatim/OSM remains ruled out for the same ODbL/policy reasons as before. Jersey Government ArcGIS FeatureServer licence remains unconfirmed, still blocked by the no-contact-Jersey constraint. Decision: leave coordinates blank.

Switched from premium to free — 2026-09-03

Human decision: given the postal-layer licence status is unresolved (not just under-attributed, genuinely unconfirmed redistribution rights), JE is switched from is_premium: true (tier B) to free rather than sold. premium_tier/premium_source_key removed from datapackage.yml. This does not resolve the licence gap — it changes the risk posture from "selling into an unresolved gap" to "distributing for free with the gap disclosed," which is a more defensible position given no confirmed redistribution right exists. Full disclosure stays in ATTRIBUTION.txt either way. If a redistribution right from Jersey Post/its data partners is ever secured, this can be revisited and switched back to premium.