karti
e5b9339015
Three changes, and the first is the one that mattered. `orbitMinDistance` was `boardSpan * 0.02`. On the Bay Area that is 20.1 units of 94.3 m — 1.89 km, the range every close photograph of this product was taken at. On the merged board the same rule is 11.1 units of 1,919 m, which is 21.3 km: eleven times further out, on the one board that carries both metros' districts, landmarks and bridges. The floor is a fraction of board *size*, and board size is exactly what `unify.ts` changed without changing what stands on it — so the merged board could never resolve a city no matter what the level of detail did, because the camera was not allowed to arrive. A board that carries `detail` districts now floors on true metres instead. Boards without them keep the span rule, which is right for an atlas. The metro chapters come back. They were merged, photographed and reverted at `64618cb`, and the recorded reason was specific and true: the corridor's 4.7 km atlas glyph filled every close frame. `23e4bb8` — the next commit — added `overviewLayers`, which withdraws that glyph inside `DETAIL_STANDOFF_M`, naming that same photograph. The blocker was gone and nobody came back for the rungs it had blocked. Each rung is restated in this board's units: `distance` on `latScale` alone, `height` carrying the exaggeration ratio as well, or the camera arrives at the right altitude over relief drawn 2.6x taller and every authored angle flattens. And with `?one=1` there is one entry in `CITIES`. That is what makes the merge a board rather than a nicer tier on a three-board product: every ladder rung's `board` is `california`, so `goToPlace` never switches; the two doors stop intercepting clicks and fly their authored poses; and `DEFAULT_CITY_ID` is the one board that exists. Without the flag nothing moves, which keeps the capture guards and every budget cell aiming at the boards they were written against. Measured, all twelve budget cells PASS. `california-one` reads 355,859 triangles and 415 draws against 400,000 / 460 — unchanged from the 352,927 / 420 recorded when the board landed. No cap was raised. What this does NOT fix, photographed rather than assumed: at FiDi's 7.7 km the camera is inside the terrain, and at the Bay's 45 km the ground is visibly faceted. Both are one cause — `verticalExaggeration` is a per-board constant, correctly derived at 15x for framing the whole state, and a board you can traverse from 1,551 km to 7.7 km cannot hold it constant. See TODO.md.
Symbolic link
1 line
43 B
Plaintext
Symbolic link
1 line
43 B
Plaintext
/home/kartios/repos/gitea/tera/node_modules |