docs(proposals): T-CostLens canvas-pair (Q-Lens-Target-Context β-extended + Q-Cost-Composition-Layering ε)

docs/proposals/q-cost-composition-layering-canvas.md

@@ -0,0 +1,116 @@
+# Canvas — Q-Cost-Composition-Layering (ε substrate-shape question)
+
+**Authority**: Director ratification of α-revised T-CostLens supersession at gunbc#828 #issuecomment-4400920572 — ε path (Rust-side cost-composition wiring; preserves `cost.dag` PROXY without introducing first-precedent .dag-side fold-over-realizations) elevated from slice-tier to canvas-tier.
+
+**Cross-cutting parent**: PR #2171 α-revised pure-docs (cost.dag deferred-discussion + 5-option matrix + 4 findings) merged 2026-05-07.
+
+**Sibling canvas**: `q-lens-target-context-canvas.md` (β-extended path; substrate-shape question on whether `.dag`-side lenses receive target-context).
+
+**Status**: **canvas — Director-RATIFIED 2026-05-07 (ε standalone for cost; canonical-not-transitional)**; PROPOSAL maturation pending. Ratification at gunb-ai/gunbc#2181 #issuecomment-4401584012 — ε ratified as canonical-for-cost; β-extended DEFERRED to N=2 trigger event (when second lens surfaces target-context need); "both sequenced" REJECTED as bridge-with-named-dissolution anti-pattern.
+
+## The factoring claim
+
+**ε path proposes**: cost composition factors as **(target-agnostic-shape) × (target-specific-values)**, where:
+- **Target-agnostic-shape** = the abstract `SymbolicCost` algebra (7-variant per `src/v3/std/algebra.dag:181-188` at HEAD: `ConstantCost(Int)` | `LinearCost(SizeVariable)` | `PolynomialCost { var, degree }` | `ProductCost(NonSingletonList<SymbolicCost>)` | `SumCost(NonSingletonList<SymbolicCost>)` | `LogCost(SizeVariable)` | `UnknownCost(String)`) + `Semiring<SymbolicCost>` composition (sequential via `SumCost` / `ProductCost`; branch / iterate via algebra ops)
+- **Target-specific-values** = per-primitive realization-cost facts (`TypeRealization.cost: Int`, `CallableRealization.cost: Int`, etc.) loaded Rust-side at emit time
+
+If this factoring is **structurally honest**, cost composition can happen Rust-side: the Rust consumer reads the abstract `SymbolicCost` shape from the `.dag`-side lens output, then composes target-specific values via Rust-side data-flow (HashMap-build over realization rows, lookup, multiply by abstract structural coefficients).
+
+If this factoring is **NOT honest**, cost composition NEEDS .dag-side target-context — ε is parallel-representation debt (Rust-side wiring duplicates / bypasses the canonical Lens-framework authority).
+
+## Test against `feedback_abstraction_layering` (Director-named anchor)
+
+Per Director correction at gunbc#828 c#4400921658:
+
+> `feedback_abstraction_layering` — objective concepts pure; LanguageSpec language-agnostic. Cite this as the framework anchor for whether emit-side cost-keyed composition respects abstraction layering.
+
+**Abstraction-layering test**:
+
+1. **Pure objective-concepts layer**: `SymbolicCost` algebra at `src/v3/std/algebra.dag:12+` carries the abstract shape. Per `feedback_abstraction_layering`, objective concepts must be pure — no language/target awareness leaks.
+2. **LanguageSpec layer**: language-agnostic; describes structural per-target realization facts without committing to a specific language's idioms.
+3. **Emit-side composition layer**: Rust-side `emit/{rust,python}_target.rs` reads LanguageSpec data + abstract `SymbolicCost` shape; composes per-target concrete cost via per-primitive HashMap-lookup.
+
+**Layering compliance test**:
+- ✓ Layer 1 (objective): `SymbolicCost` algebra is target-agnostic. Holds at HEAD.
+- ✓ Layer 2 (LanguageSpec): per-primitive realization-cost facts are language-agnostic substrate-data. Holds at HEAD (`src/v3/spec/{rust,python,go}.dag` rows are spec-data, not Rust-specific dispatch).
+- ⚠️ Layer 3 (emit-side composition): IF cost composition factors honestly, this layer is the natural home for (target-agnostic × target-specific) → concrete cost. IF factoring fails, this layer carries load-bearing logic that should have been substrate.
+
+## Honest-factoring evidence
+
+### Pro factoring (ε is structurally honest)
+
+**Argument**: `SymbolicCost` is already target-agnostic. Per fierce-ram-21's grep at gunbc#2153 c#4399898495 (Slice 1 finding):
+
+> `lenses/cost.dag` lens output is already `Lookup<SymbolicCost>`-typed; `Lookup<SymbolicCost>` MissingCost/HitCost machinery all already substrate. #38/#39 substrate-already-aligned.
+
+Cost composition via `Semiring::sum`/`::product` is structural-fact at HEAD. The abstract algebra IS the load-bearing substrate. Per-primitive realization values are downstream data-flow that doesn't need .dag-side lens-fold authority — Rust consumer reads the abstract shape + multiplies per-primitive concrete values.
+
+**Empirical test**: at HEAD, `lens_cost_symbolic_generated.rs` (or analogous Rust consumer) IS the natural home for emit-time cost composition. No substrate-shape change needed; ε is "make Rust-side composition explicit + document as the canonical path".
+
+### Con factoring (ε is parallel-representation debt)
+
+**Argument**: cost composition that reads target-realization-rows via Rust-side HashMap-build IS load-bearing logic. If Rust-side does the composition, the lens-framework abstraction stops at "abstract `SymbolicCost` shape from structural fold"; concrete cost values flow OUTSIDE the Lens<C> framework via Rust-side data-flow.
+
+This means: **the lens-framework authority for cost composition is incomplete at the substrate layer** — Rust-side carries load-bearing semantics that should have been substrate. Per `feedback_parallel_representation_debt`, parallel-author-Rust-side-composition violates canonical-authority-consumption (Lens<C> exists; the cost-specific composition should consume it through, not bypass it).
+
+**Empirical test**: future lenses needing target-context (e.g., emission_provenance per Q-Lens-Target-Context canvas) face the same dilemma — does each lens-with-target-context get its own Rust-side wiring? That's N parallel Rust-side compositions, each violating the lens-framework abstraction.
+
+## Mgr-tier provisional reading
+
+The factoring claim is **probably honest for cost specifically** but **probably NOT generalizable**:
+
+- **Cost-specific honest**: `SymbolicCost` algebra is genuinely target-agnostic; per-primitive realization-cost is genuinely target-specific data; Rust-side composition is the natural home per `feedback_abstraction_layering`.
+- **Not generalizable**: emission_provenance (and possibly other future lenses) may have target-keyed structural-needs that DON'T factor as cleanly. ε for cost works; ε for emission_provenance may not.
+
+**Mgr provisional reading**: ε is the right framing **IF** cost is the only target-context-needing lens at HEAD AND the factoring honestly holds for cost specifically. Per Q-Lens-Target-Context canvas's cross-cutting analysis, cost.dag IS load-bearing target-context-needing; emission_provenance.dag is only secondary. If only cost has the need + the factoring is structurally honest for cost, ε is the more constrained path.
+
+If multiple lenses surface target-context needs over time, ε's per-lens Rust-side wiring becomes parallel-representation debt → β-extended (Q-Lens-Target-Context option (i)) becomes the right path.
+
+## Director ratification (RECEIVED 2026-05-07)
+
+Director ratified at gunb-ai/gunbc#2181 #issuecomment-4401584012:
+
+1. **Factoring test**: ratified honest for cost specifically per `feedback_abstraction_layering` (Layer 1 objective concepts pure; Layer 2 LanguageSpec language-agnostic; Layer 3 emit-side composition reads abstract shape × concrete values).
+2. **Generalization scope**: per-lens Rust-side wiring is acceptable bounded to N=1 at HEAD; reopens at N=2 trigger event (sibling Q-Lens-Target-Context canvas DEFERRED until then).
+3. **Path ratified**:
+   - **(ε) RATIFIED** as canonical-not-transitional for cost composition. Cost-specific Rust-side composition; preserves `cost.dag` PROXY; future lenses re-evaluate independently per N=2 trigger.
+   - **(β-extended option (i)) DEFERRED** to N=2 trigger event (sibling canvas).
+   - **(both, sequenced) REJECTED** as bridge-with-named-dissolution anti-pattern per `feedback_same_slice_dissolution_discipline`.
+
+Closed-system disposition: if N=2 condition never fires, β-extended never fires — structurally correct under closed-system design.
+
+## On ratification — sequencing
+
+**If ε ratified**:
+1. **Q-Cost-Composition-Layering PROPOSAL doc** matures (this canvas).
+2. **T-CostLens follow-on slice** Rust-side wiring: `lens_cost_symbolic_generated.rs` (or analog) authors per-primitive realization-cost composition; `cost.dag` lens output stays abstract `SymbolicCost`-typed; PROXY status either retained or refined to "BEHAVIORALLY COMPLETE for abstract-shape; concrete composition Rust-side per ε ratification".
+3. **§10.3 row refresh** + capability-register row updated per ε disposition.
+4. **Cementing receipts**: gates #37 + #40 + #70 advance via Rust-side composition demonstrating end-to-end cost reading. #70 demonstration fixture lands per ε framing.
+5. **Future lenses with target-context need**: re-ratify per-lens (re-open canvas if N>1).
+
+**(Historical alternatives — REJECTED at ratification):** β-extended (option (i)) ratified-instead path and "both, sequenced" path both rejected per Director ratification (see Director ratification section above). Sequencing under those alternatives recorded only in the canvas-historical record at this point; not live.
+
+## Framework discipline anchors
+
+- **`feedback_abstraction_layering`** (Director-named for this canvas): objective concepts pure; LanguageSpec language-agnostic. Test: does ε's emit-side composition respect the layer separation, or does it leak target-specific logic into objective-concepts-layer? Mgr provisional reading: ε respects the layering IF the factoring is honest for cost.
+- **`feedback_parallel_representation_debt`**: canonical-authority-consumption rule. Test: does ε's Rust-side wiring duplicate or bypass Lens<C> framework? Mgr provisional reading: ε is bounded to one consumer (cost) at HEAD; doesn't yet violate. Re-evaluates if N>1.
+- **`feedback_same_slice_dissolution_discipline`**: bridge-with-named-dissolution anti-pattern. Test: does ε ship a bridge with a named-future-dissolution to β-extended? Mgr provisional reading: NO if Director ratifies ε as canonical-for-cost (not as "transitional until β-extended"). YES if Director ratifies "both sequenced" — that becomes a named-future-dissolution.
+
+## Sibling canvas
+
+**Q-Lens-Target-Context** addresses the broader cross-cutting question: should `.dag`-side lenses receive target-context generally? This canvas (Q-Cost-Composition-Layering) addresses the cost-specific factoring claim. The two canvases together give Director ratification surface across both substrate-shape decisions.
+
+**Deeper cross-cutting axis**: target-context belongs `.dag`-side (β-extended) or emit-side (ε)? The two canvases articulate the choices; Director ratifies the load-bearing shape.
+
+## Worker pin (post-ratification)
+
+If ε ratified: T-CostLens follow-on slice — fierce-ram-21 (continuity from α-revised); Rust-side wiring is implementation-tier, not substrate-fact-introduction.
+
+If β-extended ratified: Lens<C> refactor + downstream slices — see Q-Lens-Target-Context.
+
+## Cross-Mgr coordination
+
+- **Verification Mgr (#2075)**: ratchet authoring on §1.6 NYI → executable transitions sequences post-canvas-ratification.
+- **Grounding Mgr (#1944)**: target-realization data is Grounding lane authority. Under ε, Rust-side consumer reads the data; under β-extended, lens fold reads via `LanguageSpec` parameter. Cross-Mgr coordination on data-availability shape.
+
+— Authored by warm-wolf-698 (Substrate Mgr) 2026-05-07 per Director α-revised supersession at gunbc#828 #issuecomment-4400920572 + framework anchor correction at c#4400921658. Sibling to `q-lens-target-context-canvas.md` (β-extended path).