Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion PROOF-STATUS.adoc
Original file line number Diff line number Diff line change
Expand Up @@ -142,7 +142,9 @@ its historical success is not review or a receipt for the default profile.
The current working-tree entrypoint uses the source's existing separate
`--safe --cubical` options because it constructs propositional truncation
as a higher inductive type; profile review remains *OPEN*, and this island
is not part of the `--without-K` aggregate. The postulated shadow remains
is not part of the `--without-K` aggregate. A draft review record, awaiting
the owner's review and sign-off, is
link:docs/proof-profiles/cubical-island.adoc[`docs/proof-profiles/cubical-island.adoc`]. The postulated shadow remains
quarantined as documented in `docs/proof-debt.adoc`.

=== Local cold-check record (2026-11-01)
Expand Down
97 changes: 97 additions & 0 deletions docs/proof-profiles/cubical-island.adoc
Original file line number Diff line number Diff line change
@@ -0,0 +1,97 @@
// SPDX-License-Identifier: CC-BY-SA-4.0
// SPDX-FileCopyrightText: 2025-2026 Jonathan D.A. Jewell <j.d.a.jewell@open.ac.uk>
= Proof profile: the full-Cubical island
:toc: macro

This is the separately reviewed proof profile that
https://github.com/hyperpolymath/standards/blob/main/3-practice/PROOF-POLICY.adoc[PROOF-POLICY]
§2.1 requires for a theory proof that genuinely needs full Cubical Agda.

*Status: DRAFT, not reviewed.* An agent drafted this record. It is not a
review until the owner fills in the review block below and signs it off.
Until then the profile review stays *OPEN*, as `PROOF-STATUS.adoc` says.

toc::[]

== Scope

[cols="1,3"]
|===
| Module | `proofs/agda/EchoImageFactorizationPropCubical.agda` (the only module in this profile)
| SHA-256 at drafting | `20f77384844fcf4e73e41877caa3e0742f4a9d3121ac91806c3141fbe793f6bf` (echo-types `main` at `140002f`)
| Options | `{-# OPTIONS --cubical --safe #-}`, in the file
| CI entrypoint | `.github/workflows/agda.yml`, job `cubical`: `agda --no-libraries -i "$AGDA_STDLIB" -i "$ABSOLUTE_ZERO_AGDA" --ignore-interfaces --safe --cubical -i proofs/agda proofs/agda/EchoImageFactorizationPropCubical.agda`
| Toolchain pins (CI) | Agda 2.7.0.1, release tarball SHA-256 `975b515a…5db8a`; agda-stdlib `8326c7475af1fa456ca04a0203c168d5d7ff842b`; absolute-zero `8a3162a24c8524735b464bedb9747517dc8b0467` (all from `scripts/setup-agda.sh`)
|===

== Why this module departs from the default profile

The default profile is `--safe --without-K --cubical-compatible`. Under it,
propositional truncation `∥_∥` cannot be constructed, because it is a higher
inductive type. `EchoImageFactorizationPropPostulated` therefore assumes it
through four postulates. Those postulates are a necessary axiom under the
Trusted-Base Reduction Policy (`docs/proof-debt.adoc`).

This module constructs `∥_∥` as a real higher inductive type under
`--cubical`. It proves the four interface obligations, so the (epi, mono)
image factorisation becomes a proof instead of a consequence of assumed laws.

Agda forbids a `--cubical` module from importing a `--without-K` module that is
not `--cubical-compatible`. The module therefore restates `TruncInterface`,
`Echo`, `echo-intro` and the `ImageProp` content locally.

== What the profile establishes, and what it does not

Established when the CI job is green:

* The module type-checks under `--cubical --safe`, so there are no postulates
and no unsafe pragmas in it. `--safe` rejects a `postulate` under `--cubical`:
a one-line control module with a postulate exits 42.
* The four truncation obligations and the factorisation are proved in Cubical
Agda.

Not established:

* *Definitional identity with the `--without-K` originals.* The module's
header comment says the restated pieces are "definitionally identical" to
their `--safe --without-K` originals. No tool checks this. It is an argued
claim, and a reviewer should compare the restated definitions by hand.
* *Membership of the `--without-K` aggregate.* This module is outside that
aggregate, and nothing proved here transfers to the default-profile modules.
* *Warning-free compatibility.* Agda reports warnings on this module (17 with
Agda 2.8.0.2 locally). They are `UnsupportedIndexedMatch` notes on transports
in Cubical Agda, not errors. Triage remains open in `PROOF-STATUS.adoc`.

== Evidence at drafting (2026-11-18)

The CI pin is Agda 2.7.0.1. This local run used a different compiler, so it
supports the CI receipt and does not replace it.

Run from the repository root, with `AGDA_STDLIB_DIR` set to the absolute path of an agda-stdlib
checkout at commit `8326c7475af1fa456ca04a0203c168d5d7ff842b`:

[source,shell]
----
cd proofs/agda
agda --no-libraries -i "$AGDA_STDLIB_DIR/src" --ignore-interfaces \
--safe --cubical -i . EchoImageFactorizationPropCubical.agda
----

* Agda 2.8.0.2, agda-stdlib `8326c74`: exit 0, with 17 warnings.
* The control (`--cubical --safe` module containing a `postulate`): exit 42.

== Review

Fill this block in to complete the profile review. Leaving any field blank
keeps the review OPEN.

[cols="1,3"]
|===
| Reviewer |
| Date |
| Module SHA-256 reviewed |
| CI run inspected (link) |
| Restated definitions compared with the originals (yes/no, notes) |
| Decision (accept / accept with conditions / reject) |
| Conditions or follow-ups |
|===
Loading