Decide the exact-analysis capability matrix

[HsiangNianian]

Parent

• [RFC-0006] Program Sampling and Exact Analysis Scope #3

What to decide

Review the exact-distribution prototype and make the product decision for which v2 constructs, limits, and failure modes OneRoll will support in v2.1 without silent approximation.

Review inputs

• docs/rfcs/0006-capability-matrix.json
• docs/rfcs/0006-analysis-contract.json
• docs/rfcs/0006-distributions.json
• Prototype measurements and exhaustive-enumeration evidence from Prototype exact probability distributions for the v2 language #32

Required decisions

• Resolve every provisional bounded entry to supported-with-normative-limits or unsupported.
• Accept or revise state, support, operation, byte, and rational-bit ceilings.
• Accept the one-Instruction exact boundary and AnalysisValue materialization.
• Accept sequential weighted unique semantics.
• Confirm that exact mode never samples, truncates, prunes, or returns partial results.
• Move RFC-0006 from Draft to Review only after prototype evidence is complete.

Acceptance criteria

• Each language construct has a final status and reason.
• Complexity ceilings and resource errors are normative and evidence-backed.
• The API never substitutes Monte Carlo output for an exact request.
• The final decision is incorporated into RFC-0006 and user documentation.
• Human review records explicit acceptance.

Blocked by

• Prototype exact probability distributions for the v2 language #32