The evaluator is a deliberate PoC, not the target implementation shape

LIMIT-EVALUATION-004

The tree-walking evaluator is a proof of concept expected to be replaced, not a stable target shape.

Lifecycle

known

Owner

metel-interpreter

Scope

architecture/spec/evaluation.md#evaluation

Origin

this session's architecture-spec inventory (direct source reading), 2026-09-17

Affects · linked records

arch.evaluation.requirement-1architecture requirement

arch.evaluation.requirement-2architecture requirement

limitationKnown limitations ↗

Limitation

evaluator/mod.rs opens with its own verbatim statement of intent: "PoC evaluator — this implementation will almost certainly be rewritten. Implement the simplest correct thing; do not over-engineer." This is a deliberate, currently-real architectural posture, not an incidental comment — evaluator.md's own "Rewrite" section names the intended eventual replacement (a tagged-pointer or NaN-boxing value representation, carrying the v0.13.0 closure-capture model forward, RFC-0122 borrow checking replacing the in-call reentrancy flag, dyn Aspect's real fat-pointer/vtable representation).

Impact

None of arch.evaluation.*'s requirements should be read as describing a stable target shape for future extension — they describe the current, real, correct-for-now PoC, which the project's own docs say will almost certainly be rewritten. A future implementer should not treat this section's requirements as constraints on the rewrite; they describe what exists, not what must persist.

Affects

• arch.evaluation.requirement-1
• arch.evaluation.requirement-2

Resolution

None yet — accepted as the current, deliberate state. evaluator.md's "Rewrite" section is the forward-looking plan, not this record's job to restate in full.