Issues 011 and 012 — what the manifest review could not fix #12

2026-09-01T23:07:17Z

jschoubben commented

2026-09-01 23:07:17 +00:00

Two issues from the 2026-09-02 review of the mesh-control catalogue examples, opened per playbook 03 because each needs a decision the records don't yet hold:

008 — Reconciling a seed file wipes what grew in it. The cache's ACL file is declared with fixed empty content so the program can start; the provisioner and the running program then legitimately write into it; every reconcile restores it to empty and reports success. The gap is generic: "exist before first start" and "have this content forever" are different intentions with one vocabulary.
009 — Six modules own what they must share. The media stack's manifests each declare the shared library/download directories, and the resolver's duplicate-path rule (correctly strict) refuses the stack's only sensible co-assignment. No test co-resolves any pair, so it passes today. The fix is a fork in what a module is — owning module, shared-resource vocabulary, or one-module stack — hence an issue, not a patch.

The mechanical findings from the same review (gitea SSH port, dnsmasq tcp/53, registry claim scope) are fixed on mesh-control branch claude/four-more-of-the-catalogue.

https://claude.ai/code/session_01LrgweAeERJYBg88c5cKDzF

Two issues from the 2026-09-02 review of the mesh-control catalogue examples, opened per playbook 03 because each needs a decision the records don't yet hold: - **008 — Reconciling a seed file wipes what grew in it.** The cache's ACL file is declared with fixed empty content so the program can start; the provisioner and the running program then legitimately write into it; every reconcile restores it to empty and reports success. The gap is generic: "exist before first start" and "have this content forever" are different intentions with one vocabulary. - **009 — Six modules own what they must share.** The media stack's manifests each declare the shared library/download directories, and the resolver's duplicate-path rule (correctly strict) refuses the stack's only sensible co-assignment. No test co-resolves any pair, so it passes today. The fix is a fork in what a module *is* — owning module, shared-resource vocabulary, or one-module stack — hence an issue, not a patch. The mechanical findings from the same review (gitea SSH port, dnsmasq tcp/53, registry claim scope) are fixed on mesh-control branch `claude/four-more-of-the-catalogue`. https://claude.ai/code/session_01LrgweAeERJYBg88c5cKDzF

jschoubben closed this pull request

2026-09-05 00:54:47 +00:00

jschoubben reopened this pull request

2026-09-05 00:55:39 +00:00

jschoubben added 1 commit 2026-09-05 01:52:08 +00:00

Renumber manifest-review issues 008/009 -> 011/012 2c8616b650

Numbers 008 and 009 were taken on main after this branch was opened
(008-provider-runtime-has-no-seal-key, 009-runtime-config-change-does-not-restart,
both merged). Renumber the seed-file-wipe and shared-directory issues to the next
free numbers so merging records two more issues rather than duplicating two.

Claude-Session: https://claude.ai/code/session_01LrgweAeERJYBg88c5cKDzF

jschoubben changed title from ~~Issues 008 and 009 — what the manifest review could not fix~~ to Issues 011 and 012 — what the manifest review could not fix

2026-09-05 01:52:18 +00:00

jschoubben merged commit 600a57aafb into main

2026-09-05 01:52:32 +00:00

jschoubben deleted branch issues/manifest-review

2026-09-05 01:52:32 +00:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: novox/hq#12