03-DESIGN/01-to-be/20-writing-a-module.md

@@ -5,8 +5,9 @@ code:
   - mesh-catalog modules/showcase
   - mesh-controller internal/builder
   - mesh-sdk src
-updated: 2026-09-15
+updated: 2026-09-21
 decisions:
+  - 02-DECISIONS/0053-a-step-that-runs-on-a-schedule.md
   - 02-DECISIONS/0074-the-wire-is-specified-not-the-types.md
   - 02-DECISIONS/0040-what-a-module-is.md
   - 02-DECISIONS/0039-what-the-sdk-holds-and-refuses.md
@@ -178,3 +179,22 @@ knows. Putting the Plex client anywhere else is the shared-library disease with
   the builder records that as the build edge — but *a mesh that can rebuild a commit and get a
   different library* is a real change from how everything else here works, and it should be a
   decision rather than a consequence.
+
+## A file and the thing that reads it are guarded by a gate
+
+*Written 2026-09-21, from resolving [04-ISSUES/066](../../04-ISSUES/066-a-partly-applied-declaration-leaves-a-mixed-state-with-no-rollback/00-report.md).*
+
+The apply is not a transaction: every resource is attempted, every failure reported, and only a
+failed gate stops what follows it ([ADR 0053](../../02-DECISIONS/0053-a-step-that-runs-on-a-schedule.md)).
+So a push can half-happen, and the pairing that matters is a configuration file beside the
+service that reads it. A module keeps that pair correct by declaring them in this order: the file;
+a `run-once` container that validates it; the service, with `restart-on` naming the file. A file
+the validator refuses reaches the disk and nothing else — the gate fails, the service after it is
+left as it was, and the machine reports the push failed. The service never serves what the
+validator refused. That is the grouping, made from what the manifest already has; no primitive
+withholds a file from the disk, and a service that reads its file live rather than at start is the
+one shape this does not protect, and should not be written.
+
+*How it is checked:* the lab's coupled-pair spike declares exactly this pair, pushes a refused
+file, and asserts the file on disk is the new one, the service serves the old one, and the machine
+reports the push failed.

04-ISSUES/020-a-certificate-is-issued-and-never-collected/00-report.md

@@ -1,8 +1,8 @@
 ---
-status: open
+status: resolved
 opened: 2026-08-31
 located-in: [mesh-control, mesh-lab]
-fixed-by:
+fixed-by: mesh-lab multiple-fixes — the bed, not the proxy: its backend stub never listened and its check read a subject the authority leaves empty; against both Pebble and the catalogue's step-ca the same proxy now orders, answers the challenge, is issued and serves
 amended-design:
 ---
 

04-ISSUES/020-a-certificate-is-issued-and-never-collected/01-diagnosis.md

@@ -0,0 +1,20 @@
+# Diagnosis — 2026-09-21
+
+1. The report's own remedy was taken: the same order against a second ACME implementation. The
+   catalogue's own certificate authority, step-ca, was raised in the bed beside Pebble, pinned as
+   the catalogue pins it, and the same proxy pointed at it over the same challenge path. It issued
+   within a second of the order.
+2. The name was still not served, and the reason was neither authority's: the bed's stand-in
+   backend, a netcat loop behind the route, used flags this machine's netcat has not got and never
+   listened. Every request through the proxy was refused by the backend after a completed
+   handshake, which read as "never served over TLS" — and, against Pebble, had hidden behind
+   whatever Pebble refused first. Replaced with a real small server.
+3. With a backend that answers, Pebble issued too. The last failure was the bed's check that the
+   certificate is for the name: it read the subject, which Pebble — like the public authority it
+   stands in for — leaves empty, putting the name in the alternative names alone.
+4. So the symptom was the bed's, twice over, and the proxy's ACME client is right against both
+   implementations. Ruled out: a Pebble interop detail, and a fault in the client's order or
+   challenge handling.
+
+**Located in:** the certificate bed. Both authorities remain in it: Pebble because it models the
+public authority's strictness, step-ca because it is what a mesh runs.

04-ISSUES/064-a-mesh-build-cannot-fetch-a-modules-external-dependencies/00-report.md

@@ -1,11 +1,11 @@
 ---
-status: located
+status: resolved
 opened: 2026-09-18
 located-in:
   - mesh-catalog
   - mesh-controller
-fixed-by:
-amended-design:
+fixed-by: ADR 0097 and mesh-controller #38 (a vendor image is a declared build input; a recipe copying out of an undeclared image is refused); the package half by the facts on current code — the one recipe that installs a public package (the catalogue module, pg) built and installed through the mesh's builder in the genesis bed on 2026-09-21, and no recipe copies from a public image any more
+amended-design: 03-DESIGN/01-to-be/18-building-a-module.md
 ---
 
 # 064 — A mesh build cannot fetch a module's external dependencies

04-ISSUES/064-a-mesh-build-cannot-fetch-a-modules-external-dependencies/01-diagnosis.md

@@ -21,3 +21,12 @@ whose Dockerfiles fetch what no manifest names). The image half is decided and b
 is declared under `build.on`, copied in, and a recipe fetching what is undeclared is refused. Still
 open: the package half — a build must be re-run against the mesh's proxying registry to place the
 404 — and the three recipes themselves, which now declare their images or are refused.
+
+*2026-09-21, later.* The package half was placed by a run rather than a reproduction: the catalogue
+was read again, and exactly one recipe installs a public package — the catalogue module's own,
+which installs a postgres driver into an empty directory. That module was built through the mesh's
+builder and installed in the genesis bed run that day, against the mesh's package registry as it
+now proxies the public one. The 404 the report saw is not reproducible on current code. No recipe
+copies out of a public image any more either; the vendor-tool case that opened the image half was
+rewritten before the issue was, and the rule that would catch its return is
+[ADR 0097](../../02-DECISIONS/0097-a-vendor-image-is-a-declared-build-input.md). Resolved.

04-ISSUES/066-a-partly-applied-declaration-leaves-a-mixed-state-with-no-rollback/00-report.md

@@ -1,9 +1,9 @@
 ---
-status: located
+status: resolved
 opened: 2026-09-20
 located-in: [mesh-host internal/apply, mesh-controller internal/inventory (node_report)]
-fixed-by:
-amended-design:
+fixed-by: nothing new — a run-once validator before a service that restarts on the file is the grouping (ADR 0053); the lab's coupled-pair spike proves the service never serves what the validator refused; the visible half was issue 065
+amended-design: 03-DESIGN/01-to-be/20-writing-a-module.md
 ---
 
 # 066 — A partly applied declaration leaves a mixed state, with no rollback

04-ISSUES/066-a-partly-applied-declaration-leaves-a-mixed-state-with-no-rollback/01-diagnosis.md

@@ -18,3 +18,12 @@
 the grouping question alone; it closes when a coupled pair that must not be half-applied is
 found in a module, and the declaration gains a way to say so — or when enough modules have run
 that the absence is evidence. The visibility half is done.
+
+*2026-09-21, later.* The pairing was made on purpose, in a lab spike: a config file, a run-once
+validator, a service that reads the file once at start and restarts on it. The second push changed
+the file and made the validator refuse it. Observed: the file on disk was the new one, the service
+served the old one, and the machine reported the push failed. The half-state exists — and it is
+harmless, because the gate held: the refused configuration never reached the service. The grouping
+the report asked for is this order of three resources, made from what a manifest already has. The
+one shape it does not protect is a service that reads its file live, which a module should not
+write. Resolved; the pattern is in design 20 with the spike as its check.

04-ISSUES/074-a-mesh-test-wears-a-catalogue-modules-name/01-diagnosis.md

@@ -20,3 +20,10 @@
 
 **Located in:** mesh-lab, the ten declared beds, one conversion each with a lab run. Not renamed,
 by decision; converted two at a time as the modules they need are raised beside them.
+
+*2026-09-21, later.* Ten conversions were the wrong shape. Six of the ten beds duplicated coverage
+that exists elsewhere since the catalogue beds and the whole-mesh beds read the catalogue: the four
+sidecar beds proved a sidecar alone, which the module beds prove whole; the minio and postgres
+grant beds proved a grant with a second store beside the foundation's, which the grant bed and
+the vault bed prove against the catalogue. Retired, with their scenarios. Two remain declared:
+route-forwarding, which needs the certificate authority beside the proxy, and the large mesh test.