Next Gen STEP AP242 Native-ish BREP Geometric Modeling Kernel
docs: add EDGE-A3 edge-finish taxonomy
Yuechen Li6a183978c057
diff --git a/docs/aetheris-v2-sweep-first-architecture.md b/docs/aetheris-v2-sweep-first-architecture.mdindex d0d16cd1..a42a5adf 100644--- a/docs/aetheris-v2-sweep-first-architecture.md+++ b/docs/aetheris-v2-sweep-first-architecture.md@@ -88,7 +88,7 @@ V2 directive:- Prefer subtractive intent realization in profile/AIR space (e.g., profile-stack intervals, layered radii, resolved profile region normalization).- Use 3D subtract only when intent cannot be declared from bounded constructive relations.-EDGE-A2 constructive chamfer note: chamfers follow the same rule. When construction context is available, a chamfer may be a profile modification, a profile-stack/section transition, or a direct final-topology construction manifest. Do not trim a BRep, edge-sweep a local patch, or mutate an already-emitted sharp body when the chamfered final topology can be emitted from the constructive profile/stack/corner context from the start. EDGE-PROFILE-X1 validates the simplest vertical-edge case as `rectangle -> chamfered pentagon -> extrude`, with the bevel face emitted as an ordinary side face and no AirEdgeSweep, BRep chamfer mutation, topology graft, or 3D Boolean fallback.+EDGE-A2 constructive chamfer note: chamfers follow the same rule. When construction context is available, a chamfer may be a profile modification, a profile-stack/section transition, or a direct final-topology construction manifest. Do not trim a BRep, edge-sweep a local patch, or mutate an already-emitted sharp body when the chamfered final topology can be emitted from the constructive profile/stack/corner context from the start. EDGE-PROFILE-X1 validates the simplest vertical-edge case as `rectangle -> chamfered pentagon -> extrude`, with the bevel face emitted as an ordinary side face and no AirEdgeSweep, BRep chamfer mutation, topology graft, or 3D Boolean fallback. EDGE-A3 adds that edge-finish lowering should also be selection-pattern driven: single-edge, face-boundary loop, and whole-body canonical edge-set cases with uniform rules are the baseline targets; arbitrary edge graphs, unequal-distance chamfers, mixed per-edge rules, and variable-radius fillets remain deferred unless explicitly bounded.This direction is consistent with:diff --git a/docs/edge-a1-chamfer-fillet-support-compatibility-matrix.md b/docs/edge-a1-chamfer-fillet-support-compatibility-matrix.mdindex 879e7c7b..135efb99 100644--- a/docs/edge-a1-chamfer-fillet-support-compatibility-matrix.md+++ b/docs/edge-a1-chamfer-fillet-support-compatibility-matrix.md@@ -10,6 +10,8 @@ Legacy bounded chamfer/fillet behavior remains authoritative wherever productionAIR-CIR-A0 adds an analyzer caveat for these lanes: map/volume/containment support should be read as backend- and mirror-scoped, not as general chamfer/fillet topology proof. `analyze map` may use BRep raycast where the explicit BRep query backend admits the body, or CIR/FRep only when an admitted mirror exists; AirChamfer and fillet routes remain `mirror-unavailable` by default until a separate mirror milestone proves a bounded reduction.+EDGE-A3 adds a selection-pattern taxonomy over the matrix: baseline edge-finish support should be interpreted as selection-pattern driven, with Tier 1 single-edge, Tier 2 face-boundary loop, and Tier 3 whole-body canonical edge-set cases prioritized under uniform symmetric chamfer or constant-radius fillet rules. Arbitrary edge graphs, mixed per-edge rules, unequal-distance chamfers, and variable-radius fillets remain deferred/advanced unless a future bounded policy admits them. See `docs/edge-a3-edge-finish-selection-taxonomy.md`.+EDGE-A2 reclassifies some chamfer cases as profile/profile-stack candidates rather than AirEdgeSweep candidates: vertical edges of history-known extrusions should be tested as profile modifications, and top/bottom extrusion edges should be tested as profile-stack or section-transition constructions. EDGE-PROFILE-X1 now provides lab evidence for the vertical-edge/profile-authored subset, EDGE-PROFILE-V1 packages that proof into an internal production-adjacent emitter, and EDGE-PROFILE-X2 provides lab-only section-transition evidence for one top horizontal +X side chamfer while documenting current profile-stack generalization blockers. EDGE-PRISMATIC-A0 classifies those top/bottom/horizontal history-known chamfers under the **prismatic section-transition** roadmap, EDGE-PRISMATIC-X1 introduces the first reusable lab-only prismatic emitter, EDGE-PRISMATIC-X2 re-expresses the top +X witness through that emitter, EDGE-PRISMATIC-X3 proves equal-count line-only generic polygon cases, EDGE-PRISMATIC-V1 packages that capability behind an internal production-adjacent seam, and EDGE-PRISMATIC-V2 adds an internal controlled top `+X` horizontal chamfer route that consumes that seam for the rectangular-prism case. Production statuses remain unchanged until a separately gated production route is admitted.The matrix separates five concerns that were previously easy to conflate:@@ -279,6 +281,20 @@ This milestone does not include:- sketch solver or clipping engine work.++## EDGE-A3 selection taxonomy update++EDGE-A3 reframes the selection-scope rows in this matrix as practical support tiers rather than a single march toward arbitrary edge-graph mutation:++- **Class A / Tier 1 — single edge, uniform rule:** the immediate baseline for plane-plane or simple profile/prismatic contexts.+- **Class B / Tier 2 — face-boundary loop or ordered chain around one face, uniform rule:** the next bounded target because loop order and corner sequence are known.+- **Class C / Tier 3 — whole-body canonical edge set, uniform rule:** a body-family operation over simple boxes/prisms, not `N` unrelated edge selections.+- **Class D / Tier 4 — arbitrary edge graph, mixed rules, unequal-distance chamfer, or variable-radius fillet:** deferred/advanced unless reducible to Classes A/B/C or admitted by an explicit bounded policy.+- **Tier 5 — non-analytic/freeform/surfacing-heavy edge finishes:** deferred outside current bounded analytic scope.++Consequences for existing rows: CH-03 should be read as a single-edge case whose route authority depends on whether history is known; top/horizontal history-known chamfer belongs to the prismatic section-transition route; chain rows should distinguish face-boundary loops from arbitrary edge graphs; and whole-body box/prism uniform chamfer/fillet support should be called out as canonical body-family support rather than broad graph support. Baseline Aetheris work should prioritize uniform symmetric chamfer and constant-radius fillet rules.++## EDGE-PRISMATIC-V1 status noteThe prismatic section-transition row is now **internal production-adjacent / not routed** for Z-axis two/three-section, line-only, one-outer-loop, no-hole, equal-count identity-correspondence transitions. Rectangle, pentagon, hexagon, asymmetric pentagon, and stable+transition rectangle cases validate the split-preserving topology formula and STEP smoke through the existing exporter. EDGE-PRISMATIC-X4 records split preservation as the current prismatic contract: coplanar section-boundary faces are semantic output by default, not merge candidates inside the emitter. EDGE-PRISMATIC-X5 adds artifact-corpus evidence for the same split-preserving contract by writing deterministic STEP files and JSON summaries for successful rectangle, polygon, asymmetric polygon, and top-edge chamfer cases plus JSON-only invalid/deferred diagnostics. No production chamfer/fillet behavior, ProfileStack behavior, STEP exporter/importer behavior, Boolean core behavior, AirEdgeSweep behavior, public API, triangle migration, sketch solver, clipping engine, or NURBS/freeform support changed.diff --git a/docs/edge-a2-constructive-chamfer-reframing-audit.md b/docs/edge-a2-constructive-chamfer-reframing-audit.mdindex 4e17bb51..39eab663 100644--- a/docs/edge-a2-constructive-chamfer-reframing-audit.md+++ b/docs/edge-a2-constructive-chamfer-reframing-audit.md@@ -19,6 +19,9 @@ This audit concludes:The recommended convergence state is **success** for a docs/design audit: no production route changed, the theory is classified, and the smallest falsifiable lab sequence is now explicit.++EDGE-A3 follow-up: `docs/edge-a3-edge-finish-selection-taxonomy.md` adds the selection-pattern layer above this constructive reframing. It collapses baseline edge-finish support around single-edge, face-boundary loop, and whole-body canonical edge-set selections with uniform symmetric chamfer or constant-radius fillet rules. This strengthens the A2 conclusion that history-known chamfers should lower through profile/prismatic/manifest construction where possible, while arbitrary edge graphs, unequal-distance chamfers, mixed rules, and variable-radius fillets remain deferred unless separately bounded.+EDGE-PROFILE-X1 follow-up: `ProfileChamferExtrudeLab` now exercises that smallest vertical-edge profile-chamfer sequence. It builds a chamfered rectangle/pentagon, emits it with `BrepExtrude.Create`, exports it through `Step242Exporter`, and records `10` vertices, `15` edges, `7` planar faces, `0` cylindrical faces, and one bevel side face without AirEdgeSweep, `BrepBoundedChamfer`, topology grafting, or 3D Boolean fallback. See `docs/frictionlab/edge-profile-x1-vertical-edge-chamfer-profile-extrude-lab.md`.EDGE-PROFILE-V1 packaging note: `ProfileVertexChamferExtrudeEmitter` moves the same proof into an internal production-adjacent seam with deterministic admissibility, topology, STEP, and no-legacy-route diagnostics. It remains non-authoritative and does not replace production chamfer, fillet, primitive, STEP, Boolean, or AirEdgeSweep behavior. See `docs/edge-profile-v1-profile-authored-chamfer-emitter.md`.diff --git a/docs/edge-a3-edge-finish-selection-taxonomy.md b/docs/edge-a3-edge-finish-selection-taxonomy.mdnew file mode 100644index 00000000..63d29458--- /dev/null+++ b/docs/edge-a3-edge-finish-selection-taxonomy.md@@ -0,0 +1,341 @@+# EDGE-A3 — Edge-finish selection taxonomy and support collapse audit++## 1. Executive summary++Edge finish support should be selection-pattern driven. The practical chamfer/fillet problem in Aetheris is not “support arbitrary mutation of arbitrary edge graphs”; it is to classify the user’s intent into bounded selection patterns, admit a uniform rule, and lower that intent through the most construction-aware route available.++Practical chamfer/fillet use collapses into three baseline selection patterns:++- **single edge**;+- **face-boundary loop**;+- **whole-body canonical edge set**.++Arbitrary edge graphs, mixed rules, unequal chamfer distances, and variable fillet radii are not baseline support. They may become future advanced features only when an explicit bounded policy defines admissibility, corner behavior, topology, artifact evidence, and fallback authority.++Aetheris should optimize first for bounded uniform-rule cases: symmetric chamfers with one distance and constant-radius fillets with one radius. This aligns edge-finish work with V2’s sweep-first doctrine: preserve construction intent, construct final topology when possible, and use no-history/local-edge AIR fallback only when construction history is unavailable or intentionally not relied upon.++## 2. Problem reframing++The old/general CAD framing treats chamfer and fillet as a broad post-hoc edge operation over already-materialized topology:++- arbitrary subset of arbitrary edges;+- arbitrary radii or distances;+- arbitrary corner interactions among selected and unselected edges;+- arbitrary adjacent surface families;+- variable radii, unequal-distance chamfers, and mixed per-edge rules in one command.++That framing maximizes generality but also maximizes uncontrolled behavior. It pushes the implementation toward edge-graph mutation, late trim discovery, corner-case explosion, and ambiguous fallback semantics. It also obscures the constructive cases where Aetheris can emit the finished body directly.++The Aetheris framing should be narrower and more explicit:++- common manufacturable selection patterns first;+- uniform rule first;+- construction-history-aware lowering where the source operation is known;+- no-history/local-edge fallback only when the constructive route is unavailable or intentionally out of scope.++This means a chamfer on a vertical extrusion edge is usually a profile-authoring problem, a chamfer on a top/bottom extrusion edge is often a prismatic section-transition problem, and a chamfer on an imported plane-plane edge may be a narrow AirChamfer/AirEdgeSweep problem. The same classification should also shape fillet architecture before implementation begins.++## 3. Selection-pattern taxonomy++### Class A — Single edge++Class A selects exactly one edge.++Examples:++- one outside box edge;+- one slot edge;+- one pocket/rim edge.++Lowering depends on construction context:++- **vertical extrusion edge** -> profile-authored chamfer/fillet candidate;+- **top/bottom/horizontal extrusion edge** -> prismatic section-transition candidate;+- **no-history/imported local edge** -> AirEdgeSweep/AirChamfer/AirFillet candidate.++Class A is the smallest practical support target. It isolates adjacent-face families, endpoint policy, rule admissibility, and topology counts before loop and body-wide coordination are attempted.++### Class B — Face-boundary loop / chain around one face++Class B selects a face boundary loop or a connected chain around one face, ordered by topology.++Examples:++- chamfer the top rim of a box;+- fillet the rim of a pocket;+- soften the boundary of a slot face.++Advantages:++- natural ordering;+- known corners;+- known adjacency;+- corner policy derived from loop structure.++Likely routes:++- profile loop modification;+- prismatic section transition;+- loop-level AirEdgeSweep only if no construction history is available or admitted.++Class B should not be conflated with arbitrary edge chains. A face loop has topological order, repeated local context, and an implied corner sequence; an arbitrary chain may not.++### Class C — Whole-body canonical edge set++Class C selects all or classed edges of a simple body.++Examples:++- chamfer all outside edges of a box;+- fillet all outside edges of a box;+- soften all vertical edges of a prism;+- manufacturing roundover for a simple bounded primitive.++This should be treated as a canonical operation over a body family, not as `N` unrelated edge selections. A box-wide chamfer or fillet has known edge classes, known valence, repeated symmetry, and predictable corner interactions. The operation should therefore be represented by a constructive manifest or body-family policy when possible.++Likely routes:++- constructive manifest;+- profile/prismatic generation;+- legacy bounded route while migration/parity is unproven.++### Class D — Arbitrary edge graph++Class D selects arbitrary edge subsets with no guaranteed face-loop order, no simple body-family guarantee, and no bounded corner topology.++Properties:++- no natural loop/body family guarantee;+- mixed corners and termination;+- selected and unselected incident edges can interact unpredictably;+- ordering may be ambiguous or irrelevant to final topology.++Class D should be advanced/deferred unless reducible to Class A, B, or C, or unless a future milestone defines an explicitly supported bounded graph with deterministic admissibility, rejection reasons, corner policy, artifacts, and route authority.++## 4. Rule taxonomy++### Uniform rule++A uniform rule applies the same chamfer distance or fillet radius across the entire admitted selection. This is the first-class support target for Aetheris edge finishes.++### Symmetric chamfer++A symmetric chamfer uses equal offsets or a single distance. It is the first-class chamfer rule because it is easiest to express in profile/prismatic construction, easiest to validate geometrically, and easiest to compare against legacy bounded behavior.++### Constant-radius fillet++A constant-radius fillet uses one radius. It is the first-class fillet rule because it keeps cylindrical patch construction and corner policy bounded for plane-plane and simple prismatic cases.++### Unequal-distance chamfer++An unequal-distance chamfer uses two distances or an asymmetric offset rule. This is deferred/advanced because it increases adjacent-face trim policy and route-specific parameter interpretation.++### Variable-radius fillet++A variable-radius fillet changes radius along an edge or chain. This is deferred/advanced because it moves beyond the baseline cylindrical-patch model and complicates corner and STEP topology evidence.++### Mixed per-edge rules++Mixed per-edge rules assign different radii or distances to different edges in one selection. This is deferred/advanced unless a bounded policy admits it for a known body family or loop form.++Aetheris baseline edge-finish support should prioritize uniform symmetric rules: one chamfer distance or one fillet radius for the selected Class A/B/C target.++## 5. Lowering strategy matrix++| Selection class | Rule class | Construction history known? | Preferred route | Fallback route | Current evidence | Missing evidence | Production readiness |+| --- | --- | --- | --- | --- | --- | --- | --- |+| Class A: single vertical extrusion edge | Uniform symmetric chamfer | Yes | `ProfileVertexChamferExtrudeEmitter` | Legacy bounded route if production case exists; no-history AirChamfer only when history is unavailable | EDGE-PROFILE-X1 lab; EDGE-PROFILE-V1 internal production-adjacent emitter | Production authority, selector integration, parity against legacy route | Not production-authoritative |+| Class A: single top/horizontal extrusion edge | Uniform symmetric chamfer | Yes | `PrismaticTopEdgeChamferPrototype` / `PrismaticSectionTransitionEmitter` | Legacy bounded route if production case exists; controlled no-history AirChamfer only if admitted | EDGE-PRISMATIC-X1/X2/X3/V1/V2, X5 corpus, X6 analyzer confirmation, X8/X9 map evidence | Production selector policy, broader edge classes, loop/corner policy | Internal controlled route / not production-authoritative |+| Class A: single no-history plane-plane edge | Uniform symmetric chamfer | No | AirChamfer / AirEdgeSweep | Legacy bounded route where existing production behavior already owns the case | EDGE-X AirChamfer policy, topology, artifact, corpus, shadow diagnostics | Production route replacement gates, broader plane-plane cases, fallback authority | Experimental/non-authoritative except legacy-owned cases |+| Class B: face-boundary loop on profile/extruded face | Uniform symmetric chamfer | Yes | Profile loop modification or prismatic transition sequence | Legacy bounded route if it already owns the case | Profile/prismatic single-edge evidence implies likely route family | Loop admissibility, ordered corner policy, topology/corpus evidence | Deferred; next audit target |+| Class B: face-boundary loop no-history | Uniform symmetric chamfer | No | Future AirEdgeSweep loop policy | Legacy bounded route if existing behavior owns a narrow case | AirChamfer single-edge evidence only | Loop-level topology plan, corner policy, artifacts, rejection diagnostics | Deferred/future |+| Class C: whole-body box/prism | Uniform symmetric chamfer | Usually yes for generated primitives; no for imported bodies unless recognized | Constructive manifest / future profile-prismatic synthesis | Legacy bounded route while migration/parity is unproven | EDGE-A2 corner-manifest framing; legacy bounded chamfer exists for current supported cases; prismatic/profile evidence | Body-family manifest, selector classing, corner/corpus parity | Legacy remains authoritative where supported; constructive route deferred |+| Class A: single plane-plane edge | Uniform constant-radius fillet | Yes or no | Future AirFillet, or profile/prismatic lowering where history is known | Legacy bounded fillet where production behavior exists | EDGE-A1 scoped fillet rows; no EDGE-A3 implementation | AIR-FILLET/EDGE-FILLET architecture, cylindrical witness, STEP/corpus evidence | Deferred except legacy-owned cases |+| Class B: face-boundary loop | Uniform constant-radius fillet | Prefer yes; no-history possible later | Future loop-level AirFillet with corner policy, or construction-history loop rewrite | Legacy bounded fillet where existing behavior owns a narrow case | Taxonomy only | Ordered cylinders, corner patches, loop artifacts, analyzer/mirror evidence as applicable | Deferred |+| Class C: whole-body box | Uniform constant-radius fillet | Usually yes for generated primitives | Legacy bounded route / future constructive manifest | Legacy bounded fillet where supported | Existing bounded legacy behavior; taxonomy only for future constructive route | Structured corner patches, manifest, corpus, parity | Legacy remains authoritative where supported; constructive route deferred |+| Class D: arbitrary edge graph | Mixed chamfer/fillet | Irrelevant or mixed | Deferred | Deterministic rejection / split into Classes A/B/C where possible | EDGE-A1 marks broad graph support as gap | Explicit bounded graph policy if ever pursued | Not baseline support |+| Any class | Unequal-distance chamfer | Yes/no | Deferred/advanced | Deterministic rejection or legacy-only if existing route owns a specific case | EDGE-A1 rule taxonomy | Asymmetric rule semantics, per-route topology, artifacts | Not baseline support |+| Any class | Variable-radius fillet | Yes/no | Deferred/advanced | Deterministic rejection or legacy-only if existing route owns a specific case | EDGE-A1 rule taxonomy | Non-constant radius geometry, corner policy, STEP evidence | Not baseline support |++## 6. Support tiers++### Tier 1++- single-edge;+- uniform rule;+- plane-plane or simple profile/prismatic context.++### Tier 2++- face-boundary loop;+- uniform rule;+- construction-history known or explicitly recoverable.++### Tier 3++- whole-body canonical edge set;+- simple primitives/prisms;+- uniform rule.++### Tier 4++- arbitrary edge graph;+- mixed rules;+- variable radius;+- unequal-distance.++### Tier 5++- non-analytic/freeform/surfacing-heavy edge finishes.++Aetheris should target Tier 1–3 first. Tier 4–5 remain deferred unless bounded by explicit policy.++## 7. Chamfer implications++Chamfer support collapses into a small route decision tree:++- vertical history-known edge -> profile-authored chamfer;+- top/bottom history-known edge -> prismatic transition;+- face loop -> profile/prismatic loop rewrite;+- whole-body simple box/prism -> constructive manifest or legacy bounded route while parity is unproven;+- no-history edge -> AirChamfer;+- arbitrary graph or mixed/unequal distances -> deferred.++This is the direct continuation of EDGE-A2, EDGE-PROFILE, and EDGE-PRISMATIC results. EDGE-A2 reframed many chamfers as construction of final topology rather than mutation of an already-sharp BRep. EDGE-PROFILE work demonstrates the vertical profile-authored branch. EDGE-PRISMATIC work demonstrates the top/horizontal section-transition branch. AirChamfer remains valuable, but its baseline role is no-history/local-edge fallback, not universal chamfer authority.++The important collapse is conceptual: the question is no longer “can the chamfer engine trim arbitrary edges?” The first question is “which selection class and uniform rule did the author ask for, and which construction-aware route can emit the final topology directly?”++## 8. Fillet implications++Fillet support should use the same selection taxonomy before adding geometry:++- single plane-plane edge -> cylindrical patch / future AirFillet;+- face-boundary loop -> ordered cylinders plus corner policy;+- whole-body box -> cylinders plus corner patches in a structured pattern;+- uniform radius only for baseline;+- mixed/variable radius deferred.++This should become input to a future `EDGE-FILLET-A0` or `AIR-FILLET-A0`. The future fillet audit should avoid starting from universal arbitrary edge graphs. Instead, it should classify fillet selection intent by Class A/B/C/D, admit constant-radius plane-plane or simple prismatic cases first, and document deterministic rejection for variable/mixed/freeform cases until explicitly bounded.++## 9. Corner policy simplification++Corners are much easier when the selection class and uniform rule are known.++Face loops provide an ordered corner sequence. The implementation can reason over adjacent selected edges, loop orientation, face ownership, and repeated endpoint transitions rather than discovering arbitrary graph neighborhoods.++Whole-body canonical sets provide known corner valence and symmetry. A box has structured three-edge convex corners; a simple prism has repeated top/bottom/vertical edge classes. A canonical manifest can name these interactions directly instead of pretending the command is just many unrelated single-edge operations.++Arbitrary graphs create uncontrolled corner policy: mixed selected/unselected incident edges, ambiguous termination, non-uniform rules, and topology that may depend on graph traversal order. Such support should be deferred unless a future milestone admits one bounded graph family with explicit policy.++## 10. Production-readiness implications++Readiness gates should be selection-class specific.++### Single-edge gate++A Class A route must define:++- admissibility policy for adjacent face family, edge orientation, rule size, and endpoint context;+- constructive route preference: profile, prismatic, AirChamfer/AirFillet, or legacy;+- explicit BRep topology and STEP expectations;+- feature-recognition parity where imported/no-history cases are claimed;+- fallback/legacy authority when the new route is not authoritative;+- artifact/corpus evidence for representative admitted and rejected cases;+- CIR mirror/analyzer evidence only where relevant to analysis claims, not as topology authority.++### Face-loop gate++A Class B route must define:++- loop selection contract and ordered topology;+- admissibility for open chains vs closed loops;+- uniform rule policy;+- corner policy for every loop vertex;+- constructive lowering route and fallback route;+- BRep topology/STEP artifacts for successful loops;+- deterministic rejection diagnostics for unsupported loop forms;+- feature-recognition parity if STEP/imported loops are admitted;+- CIR/analyzer evidence where occupancy or containment is claimed.++### Whole-body gate++A Class C route must define:++- body-family recognition or construction manifest;+- canonical edge classes such as outside, vertical, top, bottom, rim, or primitive-specific classes;+- uniform rule policy;+- structured corner manifest;+- BRep topology and STEP artifact expectations;+- parity against legacy bounded behavior where legacy owns existing production support;+- fallback/legacy authority during migration;+- artifact/corpus coverage for each admitted body family;+- CIR mirror/analyzer evidence only for admitted mirror requests.++### Arbitrary graph gate++A Class D route is not a baseline goal. If pursued, it must define:++- graph admissibility and reduction to A/B/C where possible;+- deterministic traversal or graph-independent semantics;+- mixed selected/unselected endpoint behavior;+- corner and termination policy for all valence patterns;+- rule compatibility policy, including whether mixed rules are admitted;+- BRep topology/STEP artifacts and rejection fixtures;+- fallback authority and no-silent-partial-success behavior.++Until these gates are proven, arbitrary graph support should be documented as deferred rather than implied by single-edge or loop evidence.++## 11. Updates to EDGE-A1 matrix++EDGE-A1 rows should be interpreted through this taxonomy:++- CH-03 is a single-edge case whose route is either controlled/history-known or no-history depending on the path under test.+- Top/horizontal chamfer belongs to the prismatic section-transition route when construction history is available.+- Chain rows should be split into face-boundary loop/chain around one face versus arbitrary edge graph.+- Whole-body simple box/prism rows should be added or called out as canonical body-family operations rather than `N` unrelated edge selections.+- Constant-distance chamfer and constant-radius fillet rows are baseline rule candidates; unequal-distance, mixed-rule, and variable-radius rows are deferred/advanced.++The EDGE-A1 compatibility matrix now has an EDGE-A3 selection taxonomy update note to make this interpretation explicit.++## 12. Recommended roadmap++Recommended next milestones:++1. `EDGE-LOOP-A0` — face-boundary loop edge-finish audit.+2. `EDGE-FILLET-A0` or `AIR-FILLET-A0` — fillet architecture audit using selection taxonomy.+3. `EDGE-CORNER-X1` — constructive manifest for whole-body/simple corner chamfer/fillet cases.+4. `EDGE-LOOP-X1` — face-boundary uniform chamfer policy lab.+5. `EDGE-FILLET-X1` — single-edge plane-plane constant-radius fillet witness.+6. Optional: whole-body box uniform chamfer/fillet corpus.++Arbitrary edge graph support should not be the next target. The next work should prove Tier 1–3 support gates, especially face-boundary loops and constant-radius fillet architecture, before broad graph features are considered.++## 13. Relationship to CIR/AIR/BRep authority++The selection taxonomy is AIR/Firmament intent. It names what the author meant to select and which rule class applies before a topology route is chosen.++BRep emits explicit topology once the route is selected. BRep is authoritative for the resulting vertices, edges, loops, faces, surface bindings, curve bindings, and STEP-exportable explicit topology, but it should not be treated as the default place to recover high-level edge-finish selection intent.++CIR mirrors/analyzers may validate occupancy, containment, volume, or map behavior for admitted mirrors. They do not define selection topology. A prismatic or edge-finish CIR mirror can provide analysis evidence only under its own admission policy.++STEP does not recover selection intent by default. STEP import can provide BRep topology and, with separate recognizers, may recover some constructive semantics, but imported topology should not be assumed to carry Class A/B/C intent unless a recognizer admits that interpretation.++## 14. Non-goals++EDGE-A3 explicitly does not include:++- implementation;+- production route changes;+- public API changes;+- new fillet/chamfer geometry;+- broad arbitrary edge graph support;+- variable-radius or unequal-distance implementation;+- STEP exporter/importer changes;+- Boolean core changes;+- CIR node or analyzer behavior changes;+- AIR emitter behavior changes;+- BRep topology changes;+- production chamfer/fillet route replacement;+- test weakening;+- triangle migration retry;+- NURBS/freeform expansion.