# Independent boundary-order evaluator contract

Source inspection at research f4fa0c11; controls are frozen SCC run02 and prior old-layered results. No implementation, compiler, router, UI or Gitea action. Evaluate one predeclared boundary-order candidate, not a family selected after viewing scores.

## Exact missing information

`layered.rs` incoming-neighbor pass (approximately137–167) considers an edge whose target lifts to the current child and whose source does not lift to that same child. An external source need not be a sibling. Its world-space center Y influences downstream child order. Three passes carry updated positions between groups; outgoing neighbors are intentionally not chased simultaneously, avoiding a swap of both groups that preserves reversal. This is a useful mechanism, not a proof that its three-pass update converges for arbitrary parent cycles.

`condensed.rs:224–241` requires BOTH endpoints to lift into the same sibling group before retaining a link. Thus an external source contributes no local child-order information. Lines365–390 only order SCC blocks from lower-ranked neighbors already in that local links list, followed by same-rank proximity. Correct root condensation knows Left→Right, but the singleton children inside Right have neither local arcs nor external order hints. Their component/canonical order survives. This omission is separate from the successfully fixed SCC rank problem.

## Existing two-container witness: fixed expected order

Fixture definitions are `auto_layout_probe.rs:580–603`. Placement suffixes uniquely identify this fixture:

- Left parent002: Left A006 then Left B008.
- Right parent004: Right A00a then Right B00c.
- Directed edge00d: Left A006→Right B00c.
- Directed edge010: Left B008→Right A00a.

In frozen SCC run02 both groups retain A above B: child Y136 and300. The concrete downstream order witness is **Left A < Left B and Right B < Right A** (compare projected centerY; names are labels, IDs are identities). Preserve Left's order as the upstream reference; swapping both groups is not success. The mirrored physical arrangement can be equivalent globally, but it is not the predeclared one-sided witness. Every child remains under its original parent and each parent's local SCC membership/rank remains unchanged.

Old layered establishes a feasible target with unchanged semantic endpoints: zero shared length/contacts/label collisions versus SCC164 sharing,2contacts,1label/node; route length372 versus788, area272800 versus339852. Do not assert the new implementation must reproduce every coordinate. Require the ordered pairs above and compare actual final released routes to that zero-sharing witness. Smaller connector inversions alone do not pass if routing still merges the traces or labels become obstructed.

## Three additional topology falsifiers, fixed before implementation

Use equal-size ordinary SA function children, no authored ports initially; add explicit anchors only as the stated separate invariant variant. Stable aliases A/B determine the unconstrained upstream order. Use the same renderer-independent node sizes/gaps as existing fixtures, not dimensions chosen after candidate output.

1. **Opposing cross-parent evidence.** Two parents L/R, children A/B each; arcs L.A→R.B, L.B→R.A, R.A→L.A, R.B→L.B. The first pair asks for opposite L/R order; the second pair asks for matching order. These demands are incompatible. Expected: semantic direction and SCC condensation remain exact, deterministic bounded resolution with explicit ties/diagnostics, no endless joint swapping or silently dropped arcs. Do NOT require zero crossing or a particular arbitrary winner. Repeated full candidate calls must stabilize; route/label costs and any introduced hard violation remain visible.
2. **Three-parent propagation.** Parents L/M/R with A/B children. Arcs L.A→M.B, L.B→M.A, M.B→R.A, M.A→R.B. With upstream L.A<L.B fixed, expected M.B<M.A and R.A<R.B. This catches a one-step only pass and averaging stale absolute coordinates from differently positioned parents. Add no reverse arcs. Verify final relative child orders and actual connection quality; parent ranks must remain L<M<R. Run a storage/relation-ID permutation without changing semantic topology and report any different order explicitly.
3. **Multiple SCCs and local ranks.** Within target parent T, two two-member directed SCCs X=(X1↔X2) and Y=(Y1↔Y2), each feeds singleton Z via X2→Z and Y2→Z. External parent S has upperA/lowerB, with S.A→Y1 and S.B→X1. Expected same-rank block Y above X, both before Z's rank; preserve exact SCC membership and internal perimeter order. This catches implementations that sort all children globally by external neighbor Y, split cyclic blocks, or move Z into the same rank. Boundary ordering must act on SCC blocks within an existing rank, not override rank constraints. If cross-boundary preference is unsupported for multi-member blocks, return/report that scope explicitly rather than falsely claiming this fixture passes.

These are structural tests, not new gap or penalty parameters. One variant of the main witness should retain authored port-side/offset anchors to ensure order is achieved by permissible geometry, not anchor mutation. Undirected edges supply no invented rank direction; reversing their stored endpoint order must not change their meaning or create hidden directed precedence. Preserve the existing compound/folded corpus instead of constructing a grand taxonomy of synthetic cases.

## Acceptance checks for the single candidate

Retain all14 existing cases with full failures, reset/old-layered/SCC controls and raw before/after Projects. Require exact source/model conservation under declared node geometry plus existing route reset/translation; original semantic orientation, containment, port identities, attachments and route-entry ordering remain exact. Parent owns the complete Project oracle. Folded nodes remain opaque; stored expanded sizes and latent descendant offsets survive fold→layout→expand, and out-of-projection placements are untouched according to the existing contract.

Independently reconstruct quotient SCC membership, condensation ranks and rank membership before comparing boundary orders. Record hints with edge identity, actual external owner and destination block, using final projected coordinates or a deterministic canonical order as explicitly specified; avoid alias-only success claims. If ordering is ordinal, compare like-for-like coordinates within a reference parent rather than treating arbitrary parent Y translation as semantic priority. Parallel edges must not be silently deduplicated or weighted accidentally by whichever container pass saw them first.

Final actual projection is authoritative for automatic port sides/positions, node capacity and enclosure. Require exact authored anchors, no hard lead incompatibility, no new unrelated node/header penetration or overlap, and preserved pin/metadata policy. Report final route crossings, proper/T/bend contacts, raw/edge-deduplicated sharing with identified pairs, labels, area and length. Keep folded retrace visible. A better topological order that worsens actual traceability is not accepted automatically.

**Sparse40 exactness is a predeclared preservation target**: it is flat and lacks cross-parent boundary relations, so its whole candidate Project, plan geometry and metrics should remain exact to run02 (except new diagnostic fields). Prefer the same check for every case without applicable cross-boundary hints. If it changes, demand a causal explanation and classify it as scope leakage rather than accepting an accidental improvement. Mixed and reversed-ID cases are also flat: preserve their existing outputs, including known differences, rather than promising to remove router ID sensitivity in this change.

Require same-input determinism and ten full repeated candidate calls, including opposing parents, multilevel propagation and the real folded/compound views. Success cannot depend on starting from the special fixture's authored Y values alone; compare a harmless uniform translation of an entire parent subtree while preserving local child order. With canonical ordering policy, this should not reverse the result merely by changing the parent's global position. An explicitly geometry-sensitive policy must state that tradeoff before testing.

Stop after one implementation and one frozen comparison. A viable result must solve the exact existing two-container order/tracing witness, preserve sparse/mixed flat outputs, survive the contradictory/chained/multi-SCC tests without topology corruption or instability, and retain all real-case regressions transparently. Production command/history integration and native tracing remain separate acceptance work; passing these geometry checks alone is not closure of #32.