|
| 1 | +// SPDX-License-Identifier: MPL-2.0 |
| 2 | +// SPDX-FileCopyrightText: 2025-2026 Jonathan D.A. Jewell <j.d.a.jewell@open.ac.uk> |
| 3 | += AffineScript co-development proposals (cross-repo, staged for review) |
| 4 | +:toc: macro |
| 5 | + |
| 6 | +toc::[] |
| 7 | + |
| 8 | +== What this directory is |
| 9 | + |
| 10 | +This tree holds **cross-repo proposals produced during AffineScript |
| 11 | +co-development that target a repo *other than* affinescript**. They are |
| 12 | +staged here — as real, diff-able files — because the working agreement |
| 13 | +for this session is *writes go to the affinescript repo only*; everything |
| 14 | +destined for idaptik, echo-types, standards, etc. is parked here for human |
| 15 | +review before it is applied to its home repo. |
| 16 | + |
| 17 | +Each subdirectory is a self-contained proposal with its own `README.adoc` |
| 18 | +explaining what it is, where it belongs, how it was verified, and how to |
| 19 | +land it. |
| 20 | + |
| 21 | +== Index |
| 22 | + |
| 23 | +[cols="1,2,1,2",options="header"] |
| 24 | +|=== |
| 25 | +| Subtree | Target repo | Status | What |
| 26 | + |
| 27 | +| `echo-types/` |
| 28 | +| `hyperpolymath/echo-types` |
| 29 | +| *Verified* |
| 30 | +| `EchoEncodingFaithfulness.agda` — a machine-checked Agda module |
| 31 | + certifying the soundness of the "the integer IS the X" co-processor |
| 32 | + integerization pattern (lossless = injective; lossy = collision / |
| 33 | + clamp-to-sentinel). Typechecks under `--safe --without-K`, zero |
| 34 | + postulates. |
| 35 | + |
| 36 | +| `idaptik/` |
| 37 | +| `hyperpolymath/idaptik` |
| 38 | +| _pending_ |
| 39 | +| The kernel migration(s) (ReScript -> AffineScript) as a reviewable |
| 40 | + case-study / patch, with their differential-parity harnesses. |
| 41 | + |
| 42 | +| `standards/` |
| 43 | +| `hyperpolymath/standards` |
| 44 | +| _pending_ |
| 45 | +| Estate-standards implications: the co-processor integerization + |
| 46 | + differential-parity convention, the String host-boundary ABI, the |
| 47 | + github-only-network install recipe, the `--deno-esm` / TS-0 |
| 48 | + exemption-table delta. |
| 49 | + |
| 50 | +| `toolchain/` |
| 51 | +| `hyperpolymath/affinescript` (compiler/roadmap) + estate |
| 52 | +| _pending_ |
| 53 | +| Compiler/toolchain implications harvested from dogfooding: the |
| 54 | + variable-string-backend gap, `--deno-esm` to retire the TS harnesses, |
| 55 | + compiler-emitted Echo boundary obligations, release-cadence coupling. |
| 56 | +|=== |
| 57 | + |
| 58 | +== Why staged-not-applied |
| 59 | + |
| 60 | +Three reasons, all from the session's working agreement: |
| 61 | + |
| 62 | +. *Review gate.* The owner wants to review anything touching repos other |
| 63 | + than affinescript before it lands. |
| 64 | +. *Honest provenance.* A staged proposal records exactly how it was |
| 65 | + produced and verified, so the receiving repo's maintainer can re-run |
| 66 | + the checks rather than trust a claim. |
| 67 | +. *No silent cross-repo drift.* Estate policy (see the standards repo's |
| 68 | + license + no-automated-edits guardrails) is firmly against bulk |
| 69 | + cross-repo sweeps; per-repo human acceptance is the rule. |
0 commit comments