Record that the fleet diff cannot see through the secrets input

allod-agent commented

2026-07-31 00:36:58 +01:00

Member

The check the brainstorm proposes — "does this change any machine I actually run?" — has an input it cannot see through, and it turned up the first time the check met a real change. Every machine holding an encrypted secret embeds the secrets repo's own store path in its configuration, so any commit to that repo changes every such machine's derivation. A comment-only commit does it; that was measured as a control while reviewing the Forge opt-out work.

The consequence matters more than the mechanism: for changes to secrets the check always reports a change and can never distinguish a real effect from a no-op. A gate that always fires is one you learn to ignore, which is exactly the rot the brainstorm warns about elsewhere.

This records the limit and the two ways out — normalise the secrets store path away before comparing, or keep the noise and require the expectation line to say what it believes changed — without choosing between them, because nothing has adopted the process yet.

Refs allod/strategy#20

Risk

R0. Brainstorm text.

Validation

Measured during the review of allod/archetypes#17: with the secrets branch substituted, three of four machines were byte-identical and the dev machine shifted; a comment-only commit to the same repo shifted it identically, which isolates the cause to the input's store path rather than to the change.

The check the brainstorm proposes — "does this change any machine I actually run?" — has an input it cannot see through, and it turned up the first time the check met a real change. Every machine holding an encrypted secret embeds the secrets repo's own store path in its configuration, so any commit to that repo changes every such machine's derivation. A comment-only commit does it; that was measured as a control while reviewing the Forge opt-out work. The consequence matters more than the mechanism: for changes to `secrets` the check always reports a change and can never distinguish a real effect from a no-op. A gate that always fires is one you learn to ignore, which is exactly the rot the brainstorm warns about elsewhere. This records the limit and the two ways out — normalise the secrets store path away before comparing, or keep the noise and require the expectation line to say what it believes changed — without choosing between them, because nothing has adopted the process yet. Refs allod/strategy#20 ## Risk R0. Brainstorm text. ## Validation Measured during the review of allod/archetypes#17: with the secrets branch substituted, three of four machines were byte-identical and the dev machine shifted; a comment-only commit to the same repo shifted it identically, which isolates the cause to the input's store path rather than to the change.

allod-agent added 1 commit

2026-07-31 00:36:58 +01:00

Record that the fleet diff cannot see through the secrets input 2baaca7d10

allod-agent referenced this pull request from allod/tools

2026-07-31 01:02:59 +01:00

Add fleet-diff: check which machines a change rebuilds against a declared expectation #136

allod-agent closed this pull request

2026-07-31 02:21:41 +01:00

Pull request closed

This pull request cannot be reopened because the branch was deleted.

Rows
Columns

Record that the fleet diff cannot see through the secrets input #25

Risk

Validation

Pull request closed