docs/map-boundaries-research.md
# A fairer emergence map
Research and first implementation: 2 October 2026. The map now uses this policy, with an explicit distinction between checked source notes and reference-only features. This is not an exhaustive review of present-day territorial disputes.
## Recommendation
Use Natural Earth's physical geography and administrative linework, projected with Equal Earth, with an independently reviewed contested-boundary layer. Reuse reliable geometry rather than drawing a new world dataset. Fairness comes from a consistent, visible editorial policy; no provider's default political map establishes impartiality.
Equal Earth preserves relative areas, unlike the current equirectangular projection. It still distorts shapes and distances. Projection fairness and territorial-status fairness are separate decisions. See the [projection authors](https://shadedrelief.com/ee_proj/) and [PROJ documentation](https://proj.org/en/stable/operations/projections/eqearth.html).
Natural Earth is [public domain and permits modification](https://www.naturalearthdata.com/about/terms-of-use/). Its [default boundary policy](https://www.naturalearthdata.com/about/disputed-boundaries-policy/) favours physical control and supports alternative national viewpoints. We should not adopt a national viewpoint preset or call the default neutral. Its [boundary documentation](https://www.naturalearthdata.com/downloads/10m-cultural-vectors/10m-admin-0-boundary-lines/) explicitly warns that disputed and undefined boundaries are not exhaustively classified.
## Alternatives checked
- **geoBoundaries:** individual country downloads represent countries as they represent themselves; its [global composite](https://www.geoboundaries.org/globalDownloads.html) replaces disputed areas using US Department of State definitions. Useful comparison material, not a neutral default for this purpose.
- **OpenStreetMap:** its [foundation policy](https://osmfoundation.org/w/images/d/d8/DisputedTerritoriesInformation.pdf) prioritises realities on the ground, generally physical control. This does not by itself resolve our distinction between control, claims and legal recognition.
- **UN maps:** useful dated evidence for specific boundary treatments. A [UN Kashmir map](https://digitallibrary.un.org/record/490695/files/3828_Rev-_6.pdf) distinguishes the Line of Control from final territorial status. Consult relevant UN material in individual reviews; do not infer one exhaustive, neutral live dataset from a map disclaimer.
## Problems in the previous map
The previous `assets/emergence-world.json` contained 174 country paths and a merged path assembled from those same administrative polygons. It had no disputed-boundary or disputed-area metadata. Stroking that merged path drew administrative divisions like ordinary geographic outlines. Country attention fills inherited the provider's territory assignments. A dashed overlay alone would have left those assignments implied by the fill beneath it.
The previous generator fetched an unpinned upstream `master` file when its local cache was absent; its asset recorded no exact revision or review date. Geometry and attention dots also assumed a rectangular projection.
## Rendering policy
1. Draw physical land and coastlines independently from political polygons. Quiet land, no ownership colours.
2. Use dashed lines for **Disputed boundary**. Use separately explained patterns for **Line of control** and **Boundary uncertain** where dated evidence supports those categories. A dispute does not necessarily mean active armed conflict.
3. Where an area needs marking, use one neutral hatch that masks country attention fills beneath it. Do not invent one authoritative polygon when the extent itself is unresolved. Separate competing claim outlines when needed, with equal visual treatment and explicit attribution.
4. Keep control, claimed sovereignty and international/legal recognition separate in the source notes. Equal visual access to claims does not mean equal legal standing. Do not label a place as belonging to one claimant by default, or group all disputed recognition into one undifferentiated legal category.
5. Keep newsroom attention in the existing dots and country-linked records. Those country codes remain reporting categories, not sovereignty findings. Disputed territory must not acquire an owner's colour when a country is selected. No redistribution or invention of report counts.
6. Make disputed features reachable by tap and keyboard, with a short region name, category, evidence links and last-review date. Tiny islands need accessible region targets or a list entry. Hover must not be the only explanation.
The compact legend can say **“Dashed: disputed boundary”**. Put the full policy, sources and corrections in an expandable section so mobile working space is preserved. Solid lines mean ordinary reference boundaries in the reviewed layer, not a certification that every boundary is legally settled. Apply one published treatment to all visitors.
## Data and review plan
Start with the [v5.1.2 release](https://github.com/nvkelso/natural-earth-vector/releases/tag/v5.1.2), pinning its exact commit and file hashes. Inspect higher-detail 1:10m source data before simplifying for world and regional views. This scale is 1:10 million, not ten-metre accuracy.
Direct inspection of that release's GeoJSON found:
- `ne_10m_admin_0_boundary_lines_land`: 515 features, including 39 marked disputed, 10 control lines and 13 indefinite boundaries. Several categories explicitly say “please verify”.
- `ne_10m_admin_0_boundary_lines_disputed_areas`: 75 features, including 41 claim boundaries, 14 reference lines, 18 breakaway lines and two elusive frontiers.
- `ne_10m_admin_0_disputed_areas`: 99 features, including claim areas, overlays, leases and indeterminate areas.
These are feature counts, not counts of current disputes. Do not publish all of them as current conflicts or combine overlapping claim polygons without review. Raw files are in the release's [GeoJSON directory](https://github.com/nvkelso/natural-earth-vector/tree/v5.1.2/geojson).
Create a separate map-review registry with feature IDs, geographic names and aliases, category, source references, evidence dates, review date, geometry provenance and the reason for any override. This is map metadata, not a change to `/emergence/data.json`. Never derive boundary changes from AI news summaries. Supplement omissions only with attributable geometry; uncertain extent should remain uncertain.
Review the initial high-risk cases individually: Kashmir and the China–India border, Ukraine, Israel/Palestine and the Golan Heights, Western Sahara, Cyprus, Kosovo, Taiwan, Georgia's disputed regions and maritime/island claims. This is a review queue, not a claim that all cases have the same legal status. Cover the remaining source features too, including less prominent disputes and features whose status has changed since the dataset was released.
Build-time processing should produce vendored SVG paths, independent coastlines, reviewed boundary classes and disputed-area masks. A shared projection function must position geometry and the existing country attention coordinates consistently. Keep the build offline, runtime dependency-free and no-JavaScript usable. Preserve deep links, both themes and mobile zoom/pan. Publish the review date and an explicit coverage limitation; do not imply live front lines or exhaustive maritime coverage.
Before switching the map, check selected disputed regions, overlapping claims, antimeridian geometry, small islands, mask behaviour under country selection and readable line patterns at mobile zoom. Check 320/375/1280px and both modes once the implementation exists.
## First implementation
- Physical land is independent 1:50m geometry; international, disputed, control, uncertain, claim and buffer-zone lines use the 1:10m reference datasets. Country fills are reporting categories, masked beneath the 67 neutral area groups. The source does not cover every dispute.
- Eight regional groups have attributed source notes: Kashmir; West Bank/Gaza/East Jerusalem; Golan/Shebaa; Western Sahara; Crimea; Cyprus; Kosovo; Taiwan. Other groups are visibly **Reference only**. Source-note checks do not certify current geometry, armed control or equivalent legal standing.
- The correction registry excludes old Hans Island, Tiran/Sanafir, Artsakh, Donetsk and Luhansk claim/control footprints; Mbane awaits re-review after the 2025 judgment. Whole-state diplomatic non-recognition and lease boundaries are not drawn as active territorial disputes. These omissions do not supply replacement settlement geometry.
- World view is Equal Earth. At 3× zoom and above, a cylindrical equal-area view uses the focus latitude as its latitude of true scale, reducing the skew of magnified world-map edges. It recentres after a pan; it still distorts shapes away from that latitude. This follows the [PROJ cylindrical equal-area definition](https://proj.org/en/stable/operations/projections/cea.html). Regions near the poles remain difficult; the local focus latitude is limited to ±75°.
- Gestures move a composited SVG surface, coalesced with animation frames. Geometry is reprojected after a gesture ends. Signals retain relative coverage sizes, with screen size decreasing modestly at close zoom; glow is removed during movement/close zoom. Hatch spacing remains stable. Country selection, keyboard pan, region links and source notes stay accessible.
Exact release commit and SHA-256 hashes live in `scripts/map-sources.json`; rendering overrides live in `scripts/map-review.json`; `scripts/map-audit.json` records every source line's classification and the area groups. Refresh explicitly with `node scripts/gen-world-svg.js --fetch`, or use `--cache /path/to/source/files` for a verified offline refresh. The normal site build reads the vendored asset and makes no map-data requests. Review the registry before upgrading the source release; feature IDs and classifications may change.
Natural Earth geometry remains public domain. Original site code and editorial curation retain their existing license; see `assets/emergence-world.md` for provenance and the projection references.