Skip to content

READINESS.md

Latest commit

 

History

History
72 lines (57 loc) · 4.71 KB

READINESS.md

File metadata and controls

72 lines (57 loc) · 4.71 KB
Raw

proven-servers Component Readiness Assessment

Standard: Component Readiness Grades (CRG) v1.0 Assessed: 2026-03-02 Assessor: Jonathan D.A. Jewell

Current Grade: C

Grade Reference

Grade Name Release Stage Meaning
X Untested — No testing performed. Status unknown.
F Harmful / Wasteful — Reject, deprecate, or delegate.
E Minimal / Salvageable Pre-alpha Barely functional. Needs redesign or major work.
D Partial / Inconsistent Alpha Works on some things but not systematically.
C Self-Validated Beta Dogfooded and reliable in home context.
B Broadly Validated Release Candidate Tested on 6+ diverse external targets.
A Field-Proven Stable Real-world feedback confirms value. No harm in wild.

Component Assessment

Assessed by component group. Evidence is tool-backed (see audits/proof-panic-attack-2026-06-23.md): idris2 --build type-checks the proofs; zig build test exercises the engines.

Component group Count Grade Release Stage Evidence Summary Last Assessed
Connector ABI-FFI (connectors/proven-*conn) 6 B Release Candidate Full RSR stack: Idris2 ABI proofs + generated C headers + Zig FFI; exercised via Rust/Gleam/Elixir bindings. 2026-06-23
Protocol cores (protocols/proven-*) 88 C Beta Idris2 types + proofs compile under %default total (96/96 packages); Zig engines pass FFI tests (98/98). In-memory skeletons, dogfooded; not externally validated. 2026-06-23
Reference servers (proven-timestamp, proven-quic, proven-http3) 3 C Beta As above, plus deeper proofs (validator decidability, exhaustive conformance, universal safety theorems) and reproducible RFC/test vectors. 2026-06-23
Core primitives (core/proven-*) 8 C Beta Idris2 cores compile; Zig engines pass where present. 2026-06-23
Language bindings (bindings/) 20 langs D Alpha 20 bindings over the generated C ABI. Ada (pragma Import) and Java (JNI) are fully FFI-wired (reference); most others (Rust/Go/Python/OCaml/Julia/…) call libproven_* via FFI; Elixir/Gleam/ReScript are constants-only scaffolds after unproven reimplementations were removed (ADR 0003). Binding constants not yet cross-tested against the C ABI. 2026-06-24

Detailed Assessment

Protocol cores — Grade C (Beta)

  • Evidence: All 96 Idris2 packages type-check (proofs verified); all 98 Zig FFI engines pass zig build test; no proof-escape hatches and no @panic/unreachable in engine code.
  • Known limitations: the proofs constrain an Idris model; the Zig engine that runs is linked to it by hand + tests, not by a machine-checked proof (the model–implementation gap). Engines are in-memory state machines — no real networking, TLS/crypto, or persistence. 16 protocols still carry an unverified (orphaned) ABI pending migration.
  • Promotion path (→ B): close the model–implementation gap (exhaustive conformance / single-sourced tags), make every transition relation Dec, and validate on 6+ external targets via the bindings.
  • Demotion risk: Low — the build + test gates are green and reproducible.

Connector ABI-FFI — Grade B (Release Candidate)

  • Evidence: the 6 connectors are the only components with the complete, documented ABI-FFI layer and multi-language binding coverage.
  • Promotion path (→ A): real-world deployment feedback.
  • Demotion risk: Low.

Notes

  • Grades are per-component, not per-project.
  • Grade A does not mean perfection — it means demonstrated value in the field.
  • Grade F includes opportunity cost — maintaining something when a better tool exists.
  • Grades can be skipped if evidence supports it (e.g., X → C if dogfooded immediately).
  • Review all grades before each release and at least once per release cycle.
  • See the full CRG standard for complete definitions, evidence requirements, and transition criteria.