docs: D-HF honesty — public DHD still sees chords; fail() APEX not h=3

EPIC_SFA_CURVE_AWARENESS.md

@@ -61,15 +61,15 @@ Make JTS preserve OGC SFA / ISO 19125-2 curve geometries — `CIRCULARSTRING`, `
 | `feat:` | **R-PR DE-9IM** for disc vs Point, LineString, hole-free Polygon, and two discs. Point: interior `0F2FF1FF2`, boundary `FF20F1FF2`, exterior `FF2FF10F2`. Line: crossing `1F20F1102`, tangent `FF20F1102`, miss `FF2FF1102`, endpoint-interior `1020F1102`. Polygon: disjoint `FF2FF1212`, nested `212FF1FF2`, disc-in-square `2FF1FF212`, crossing `212101212`. Two discs: crossing `212101212`, disjoint `FF2FF1212`, nested `212FF1FF2` / `2FF1FF212`, ext tangent `FF2F01212`, int tangent `212F01FF2`, equal `2FFF1FFF2`. Finish slice: single-member MultiSurface both orders; `equalsTopo` on equal/crossing/rotated-control discs; full SFS table; `crosses` of two areas always false. Half-disc / CompoundCurve miss still null → linearise. | **R-PR partial**. Not “any combination of curved/flat”. Do not delete `test_R_PR_*` / `test_R_CONT_*`. |
 | `feat:` | Distance closed-form helpers in `CurveExact` (arc-to-segment, overlapping discs `nearestPoints` 0, `decideTolerance` arc-aware). | Helpers only. **D-PT / D-AA / D-OP** still describe the public `DistanceOp` TAG — keep those spec methods red. |
 | `feat:` | Disc / arc convex hull closed forms. | **H-CV partial**. **H-CC** still open. |
-| `feat:` | TestBuilder `CurveExactFns` Hausdorff laser. Public `DiscreteHausdorffDistance` still sees chords. | **D-HF half-red**. Keep the spec method. |
+| `feat:` | TestBuilder `CurveExactFns` Hausdorff laser. Public `DiscreteHausdorffDistance` still sees chords in general. `DiscreteHausdorffDistance` on #7 has closed-form for two pairs only: single-arc `CircularString` → single-segment `LineString` (apex √949/6 − 7/6 ≈ 3.967641), and two circular discs. A single-member `MultiSurface` of one disc is the same pair, not a third. Exact path skips densify; densify 0.05 is not the laser. Keep the spec `fail()`. | **D-HF half-red**. Keep the spec method. |
 | `feat:` | Disc `CurvePolygon` area 25π. | **M-AREA-CP partial** (circular discs). Keep the spec method. |
 | `arch:` | PERF-GATE: identity/chord-path rows use `assertChordPath`; slack stays **15%** (`1.15`). | Contract, not a TAG. |
 
 These closed-form subsets of **OV** and **R-PR** did **not** wait for N-SS. Line–circle clip is used *inside* R1.6; there is still no public arc-arc / arc-line utility and no arc `SegmentString`.
 
 **Still open** (say so, do not close as “future work” a TAG that already has a named subset):
 
-- Public `DiscreteHausdorffDistance` (D-HF) — TestBuilder laser only
+- Public `DiscreteHausdorffDistance` (D-HF) — still sees chords in general; TestBuilder laser only. `DiscreteHausdorffDistance` on #7 has closed-form for two pairs only: single-arc `CircularString` → single-segment `LineString` (apex √949/6 − 7/6 ≈ 3.967641), and two circular discs. A single-member `MultiSurface` of one disc is the same pair, not a third.
 - General circular noding / CompoundCurve overlay (N-AA, N-AL, N-SS, rest of OV)
 - Open-arc buffer, CompoundCurve hull, Fréchet, LEC
 - OFF = concentric arc (refactor)
@@ -199,7 +199,7 @@ Mirror the `CompoundCurve` work onto the remaining composite types.
 
 - **D-PT / D-AA** — analytical point-arc and arc-arc distance. Closed-form helpers exist in `CurveExact`; the public `DistanceOp` TAG is still red.
 - **D-OP** — `DistanceOp` accepts curved inputs without forced densification. Still the public TAG; keep the spec method.
-- **D-HF** — `DiscreteHausdorffDistance` / `DiscreteFrechetDistance` parameterise by arc length. **Partial (half-red):** TestBuilder `CurveExactFns` laser only. Public `DiscreteHausdorffDistance` still sees chords. Keep the spec method. Fréchet is still open.
+- **D-HF** — **Partial (half-red):** public DHD still sees chords in general; TestBuilder `CurveExactFns` laser only. `DiscreteHausdorffDistance` on #7 has closed-form for two pairs only: single-arc `CircularString` → single-segment `LineString` (apex √949/6 − 7/6 ≈ 3.967641), and two circular discs. A single-member `MultiSurface` of one disc is the same pair, not a third. Exact path skips densify; densify 0.05 is not the laser. Keep the spec `fail()`.
 - **C-LIN / C-AREA / C-IP** — centroids and interior point.
 
 **Depends on:** Phase 1 (F-CP for `CurvePolygon` cases).

modules/curve/src/test/java/org/locationtech/jts/spec/curveawareness/CurveAwarenessSpecTest.java

@@ -261,39 +261,39 @@ public void test_D_OP_distanceOpForCurveInputs() throws Exception {
    *   B = LINESTRING (0 0, 10 0)   // the chord / diameter baseline
    * </pre>
    * Oriented Hausdorff {@code h(A,B) = max_{a in A} min_{b in B} d(a,b)} is the
-   * max height of A above B. On the true circular arc the apex sits near
-   * {@code (5, 3.968)} (circle centre {@code (5, -7/6)}, r ≈ 5.134), so
-   * continuous {@code h(A,B) ≈ 3.968}. The control-point-only discrete reading
-   * attains only the mid control {@code (2, 3)} → {@code h = 3}. Chord-fraction
-   * densify cannot close the gap: it densifies the straight control chords,
-   * which lie <em>inside</em> the arc and never reach the true apex.
-   * <p>
-   * Spec: densify / sample along <em>arc length</em> (uniform sweep), so the
-   * discrete max-min approaches the continuous directed Hausdorff. Same gap
-   * applies to {@code DiscreteFrechetDistance}. Delete on green (epic D-HF).
+   * max height of A above B. {@code DiscreteHausdorffDistance} on #7 has
+   * closed-form for two pairs only: single-arc {@code CircularString} →
+   * single-segment {@code LineString} (apex {@code √949/6 − 7/6} ≈ 3.967641),
+   * and two circular discs. A single-member {@code MultiSurface} of one disc
+   * is the same pair, not a third. The exact path owns APEX and skips densify;
+   * densify 0.05 is not the laser. Stale mid-control {@code h = 3} is retired.
+   * That exception is not the full TAG: public DHD still sees chords in
+   * general. Keep this {@code fail()}.
    */
   public void test_D_HF_hausdorffFrechetCurveAware() throws Exception {
     Geometry arc = read("CIRCULARSTRING (0 0, 2 3, 10 0)");
     Geometry baseline = read("LINESTRING (0 0, 10 0)");
 
-    // Continuous directed Hausdorff (analytic apex of the circle through the
-    // three control points, relative to the x-axis baseline): ≈ 3.96764.
-    final double expectedContinuous = 3.96764;
+    // Apex of the locked pair: √949/6 − 7/6. Exact path owns this; densifyFrac
+    // is skipped on this pair (not the laser).
+    final double expectedContinuous = 3.967641;
     final double tol = 1e-3;
 
-    double controlOnly =
+    double exactPath =
         DiscreteHausdorffDistance.orientedDistance(arc, baseline);
-    double chordDensified =
+    double densifyCall =
         DiscreteHausdorffDistance.orientedDistance(arc, baseline, 0.05);
 
-    // Red ratchet: always fail with measured numbers. When arc-length densify
-    // (or a curve-native path) reaches expectedContinuous within tol, delete
-    // this method (do not edit it green) — epic D-HF.
-    fail("D-HF: oriented DiscreteHausdorffDistance on CIRCULARSTRING(0 0, 2 3, 10 0) "
-        + "vs LINESTRING(0 0, 10 0) should approach continuous h≈" + expectedContinuous
-        + " (±" + tol + ") by sampling along arc length; control-only got " + controlOnly
-        + ", chord-densify(frac=0.05) got " + chordDensified
-        + " (today both ≈ mid-control height 3 — densify walks control chords, not the arc).");
+    // Full-TAG ratchet: always fail. Exact path owns APEX; densify is skipped
+    // on this pair, not the laser. Stale h=3 retired. Keep this fail().
+    fail("D-HF: full TAG still open. Public DiscreteHausdorffDistance still sees chords in general. "
+        + "DiscreteHausdorffDistance on #7 has closed-form for two pairs only: "
+        + "single-arc CircularString → single-segment LineString "
+        + "(apex √949/6 − 7/6 ≈ " + expectedContinuous + "), and two circular discs. "
+        + "A single-member MultiSurface of one disc is the same pair, not a third. "
+        + "Exact path owns APEX (orientedDistance got " + exactPath
+        + "); densifyFrac is skipped on this pair (call returned " + densifyCall
+        + ", not the laser). Stale h=3 retired. Keep this fail() (±" + tol + ").");
   }
 
   // ============================================================