Skip to content

Next Gen STEP AP242 Native-ish BREP Geometric Modeling Kernel

docs: record ANALYZE-P1d-impl-2 honest-stop blocker

Yuechen Li05f8edf2be5d
diff --git a/docs/analyze-p1d-impl-2-blocker.md b/docs/analyze-p1d-impl-2-blocker.md
new file mode 100644
index 00000000..554199a8
--- /dev/null
+++ b/docs/analyze-p1d-impl-2-blocker.md
@@ -0,0 +1,28 @@
+# ANALYZE-P1d-impl-2 blocker report (honest stop)
+
+Date: 2026-05-05
+
+## Outcome
+
+Honest stop.
+
+## Why implementation was stopped
+
+A bounded, trustworthy `multi_axis_ray_consensus` implementation requires a generic face-hit provider that can prove trim inclusion for arbitrary imported faces. Current core topology/geometry bindings do not persist face-level 2D trim loops or robust surface-domain evaluators needed to classify ambiguous seam/vertex/edge/tangent hits deterministically for non-primitive BRep bodies.
+
+## Evidence from code inspection
+
+- `BrepSpatialQueries.Raycast` currently hard-gates to primitive recognition (`TryResolvePrimitive`) and has no non-primitive per-face hit contract.
+- `ClassifyPoint` has judgment candidates scaffolded, but only `primitive_analytic` is executable; `multi_axis_ray_consensus` is intentionally disabled (`GenericRayHitProviderAvailable: false`).
+- Face bindings map only `FaceId -> SurfaceGeometryId`; no persisted per-face UV boundary domain/trim representation is available in query-layer APIs.
+- Topology loops/coedges encode adjacency but not canonical per-face 2D trim-space loops needed for robust in-trim tests after analytic ray/surface intersections.
+
+## Narrow next blocker
+
+Implement/queryable face-trim projection contract (planar/cylindrical/spherical first) that can answer:
+
+1. hit point projection validity in surface parameter space,
+2. inside/on/outside trim classification with tolerance,
+3. edge/vertex ambiguity detection and seam duplicate resolution.
+
+Once available, deterministic six-axis parity can be safely activated without guessing through tangent/boundary cases.