+1 draw / +894 tris with the bird on screen, +0 with it off — measured same page, same frame, by
toggling the InstancedMesh count. ?noassets keeps the 182-tri procedural bird (+1 / +182), so this is
furniture.js's use-if-ready SECOND path, not a substitution. state.asset reports which bird is bound.
THREE THINGS MEASURED BEFORE THIS WAS SAFE, AND ALL THREE WOULD HAVE SHIPPED WRONG:
· E's bird faces +Z and this file flies -Z (Matrix4.lookAt's -Z column is the travel direction).
Sliced the GLB's 894 tris along Z in ten bins: +Z end mean y 0.31, half-width 0.009 (head + bill);
-Z end mean y 0.028 (tail). Without rotateY(PI) the magpie swoops TAIL-FIRST. This is the FIFTH
-Z/+Z instance and the first caught BEFORE it shipped — because R39's lineage table said to point
a measurement at it. Verified in pictures: perched head-leading, and a swoop frame taken head-on.
· Its origin is at the FEET (GLB law minY=0) while PERCH_Y is a body-centre height: dropped 0.20 m
so it stands on the lamp arm (4.12) instead of hovering. `folded` is 1 for the GLB — it IS the perch.
· THE FETCH IS ARMED BY createMagpie(), NOT BY THE IMPORT. index.html imports this module
unconditionally and gates only the CONSTRUCTION, so a module-scope loadGLB (furniture.js's shape)
would have put a brand-new GLB request on the ?classic=1 boot — a fetch delta on the one boot that
is not allowed one. Asserted on the network log, not on the code: default 1 request (asset 'glb',
894 tris) · ?classic=1 0 · ?magpie=0 0 · ?noassets=1 0 (asset 'primitive', 182).
AND ONE MY EDIT INTRODUCED INTO ANOTHER LANE'S TOOL: a module-scope loadGLB also fires GLTFLoader's
fetch path under node, so Lane E's pipeline/dump_bird.mjs (which imports this module unmodified) died
with "ProgressEvent is not defined". Gated on IN_BROWSER as well as ?noassets; E's tool then ran
against my edit and reported 182 triangles / 216 verts — the placement fixes are exactly +0 tris.
THE FALLBACK, FIXED AND THEN MEASURED — a half-win, stated as one. E is right that the white bars sat
3 mm ABOVE horizontal wing planes, so from below (the player is always below: perch 4.32 m, EYE 1.62)
the black wing occluded its own flash: bar y 0.048 -> 0.042. Same class of fix, my own arithmetic, for
the other two marks: the nape was BURIED IN THE HEAD (head-sphere surface at z=-0.10 is y 0.0728, the
box topped out at 0.067) so it moved behind the head to z -0.07; nape and rump thickened 0.014->0.030
and 0.02->0.034, because from the flank a 14 mm slab on a 166 mm bird is 1.3 px at 64 px. All +0 tris.
Measured in the same crop of the same pose: pale area 180 -> 399 px, dark 3,017 -> 2,755 — but TRUE
white 33 -> 28, because a downward-facing surface gets no sun. E's GLB reads 755. The fallback cannot
be made pied with primitives at zero cost; the problem is photometric as well as geometric.
Stale comments corrected: magpie.js:9 and :123 said 154 triangles; it is 182 (E's ask 1).
FILED, NOT FIXED: the arcade's awning posts land +/-0.50 m from the centreline — two rows 1.00 m
apart, 34 posts down a 5.00 m lane (my own measurement from the plan, confirming E's). AWNING_DEPTH
2.2 is tuned for a 9 m footpath. One clamp fixes it, but it changes default-boot geometry for a
district whose whole treatment is a scheduled item, and the post line should be designed with the roof.
Lane E's transmission recall checked, not assumed: longbench, streetlight and magpie as staged in
web/assets/models/ carry no KHR_materials_transmission. My lane wires neither of the first two today.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
f0b9a1d30f