Per-vertex bone indices into joints, grouped in fours.
The bones any vertex may be bound to (the skin's joint set).
OptionalmethodHow a vertex blends its joints' transforms; omission means "linear".
"linear" sums the weighted joint matrices. The sum is not rigid, so a strongly bent joint collapses and a twisted one pinches. "dual-quaternion" sums the weighted unit dual quaternions, each aligned to the hemisphere of the vertex's largest-weight joint, and normalizes the result into one rigid transform per vertex, which keeps the volume but bulges slightly on the outside of a bent joint and flips once two influences differ by more than 180 degrees.
skinAutoMovieMesh in @automovie/engine evaluates the same formula the viewer's shader draws. A dual quaternion skin needs its part transform to scale uniformly, because a non-uniform scale makes the blended transform non-rigid; admission refuses that combination.
Finite per-vertex weights in [0,1], grouped in fours. Admission accepts absolute sum error at most 1e-6 from one; viewer construction subsequently normalizes its skin buffer.
Per-vertex skeletal binding: which bones influence each vertex and by how much. Drives mesh deformation when the skeleton poses.
Both arrays are grouped in fours per vertex (glTF's 4-influences-per-vertex convention): vertex
iis influenced byboneIndices[4i .. 4i+3]with normalizedweights[4i .. 4i+3].The joints' transforms blend by
method. Every bone transform is rigid, because rig rest scale is ignored by convention, so both methods receive rotations and translations only.