ADR 003a v0.1 Ships Traditional Analysis Only

ADR-003a: v0.1 Ships Traditional Analysis Only

Status: Accepted Date: 2026-05-16 Deciders: Michael Zargham Related: ADR-003 Topological Framework Documented as Future Work; ADR-021 Three Attestation Subclasses Ship in v0.1; Traditional Forward and Backward Analysis; Design Spec

Context

Once the topological framework is deferred (see ADR-003 Topological Framework Documented as Future Work), a natural follow-up question is: how much of the framework should still ship, and how much of v0.1's analysis should be traditional (forward / backward trace coverage, gap reports, coverage percentages) versus an early subset of the framework? Earlier drafts of the design spec used "Level 1" and "Level 2" framing — Level 1 being traditional analysis, Level 2 being the topological audit. That framing fragmented the deferral story and implied a partial topological shipment that the research had not actually validated. The question is whether to simplify to a single coherent v0.1 analysis story. See Design Spec §3.3 and Traditional Forward and Backward Analysis.

Decision

flexo-rtm v0.1 ships traditional bidirectional analysis only: forward trace (requirements → design → verification), backward trace (verification → design → requirements), gap reports, and quantitative coverage metrics (see ADR-004 Quantitative Certification Outcome). Named-approver attestation infrastructure (see ADR-021 Three Attestation Subclasses Ship in v0.1) ships alongside as a separate v0.1 capability — but the closed-triangle audit (the topological audit that consumes attestations) does not ship in v0.1. The "Level 1 / Level 2" framing from earlier drafts is collapsed into "v0.1's analysis" (traditional) and "the topological framework" (future).

Consequences

Positive

• Single coherent v0.1 analysis story: traditional bidirectional analysis with quantitative coverage; named-approver attestation as a separate, well-scoped v0.1 capability
• Eliminates the half-shipped Level 2 problem — no partial topological audit that depends on registry decisions deferred to future work
• Traditional analysis is trusted and familiar to RTM practitioners; v0.1 adoption is not gated on understanding the topological framework
• Forward-compatible interop (per ADR-032 Methodology Agnosticism as Foundational Axiom): today's attestations (see ADR-021 Three Attestation Subclasses Ship in v0.1) are independently meaningful as flexo-rtm's named-signer accountability, and equally readable by any downstream-analysis path adopters may choose (topological closed-triangle audit, SLSA, GSN, ARP4754A, in-house)

Negative / Tradeoffs

• v0.1 does not deliver the topological framework's headline claim ("the RTM is closed iff every face is approved by a registered type") — adopters expecting that from the Zargham 2026 paper get traditional analysis plus accountability infrastructure instead
• Some sophistication of the topological research line is invisible to v0.1 users — rtm:AssuranceFace instances may exist in their graph (via vocabulary alignment per ADR-020 Vocabulary Alignment with Zargham 2026) without any v0.1 tooling that consumes them. That is by design: flexo-rtm does not commit to the topological audit, and adopters who choose to run topological analysis as a downstream-analysis mode read the data without translation

Neutral

• The collapse simplifies the design spec's framing: flexo-rtm v0.1's analysis is traditional bidirectional + named-signer attestation, and the topological framework is a related research line, not flexo-rtm's destination (per ADR-032 Methodology Agnosticism as Foundational Axiom). Vocabulary alignment ships as forward-compatible interop (see ADR-020 Vocabulary Alignment with Zargham 2026)

Alternatives Considered

• Ship traditional analysis + named-approver attestation + closed-triangle audit (some subset of the topological framework): A "Level 2-lite" that ships the triangle audit but not the full registry-of-pre-approved-types recursive completeness check. Rejected: the closed-triangle audit is the topological audit; what makes it meaningful is the recursive completeness check, which requires the registry. Shipping the triangle audit without that check produces a misleading certification outcome — every triangle "closes" trivially because the recursion termination is hand-waved. The chosen decoupling — ship attestation (named-approver SHACL) but not the triangle audit — is what makes v0.1's analysis coherent: traditional analysis + accountability infrastructure, with the topological audit deferred as a single unit along with the framework it belongs to.

Implementation Notes

v0.1's analysis layer in flexo-rtm (oracle/src/oracle/analysis/) implements: forward-trace SPARQL queries (requirements with no satisfying design), backward-trace SPARQL queries (verifications with no covered requirement), aspect-coverage reports (see ADR-005 Adequacy and Sufficiency as Guidance Subtypes and Aspect Coverage with Adequacy and Sufficiency), and quantitative coverage metric computation (see ADR-004 Quantitative Certification Outcome and ADR-019 Derived Binary View from Quantitative Metrics). Named-approver SHACL profiles (see ADR-021 Three Attestation Subclasses Ship in v0.1) enforce attestation integrity but do not aggregate into closed-triangle audits — that's deferred.

References

• Design Spec §3.3 (Analysis Scope), §4.1 (Traditional Forward/Backward)
• Traditional Forward and Backward Analysis — the v0.1 analysis vocabulary and SPARQL recipes
• ADR-003 Topological Framework Documented as Future Work — the framework-level deferral this ADR refines
• ADR-021 Three Attestation Subclasses Ship in v0.1 — the attestation infrastructure that ships separately