mg/FIELDNOTES

Maintenance guide / FROM THE WORKBENCH

Build, activate, verify, recover

A practical maintenance sequence that keeps source review, builds, activation and real runtime evidence distinct.

Review before changing the system

Inspect Git state and the relevant diff. Back up files being changed privately. Preserve unrelated work, the current lock file and a known working boot generation. New files referenced by a Git-backed flake must be deliberately included in its source.

Use a Git-filtered flake source. An explicit unfiltered path: source may include ignored private files before a build-time check can reject them.

Match the check to the change

ChangeRelevant evidence
DocumentationGenerated pages, links, keyboard and browser checks
User-profile applicationPackage build, executable, launch and a real file operation
NixOS serviceOption evaluation, system build, activation and actual service behavior
Graphics/kernel updateSupported hardware branch, complete build, rollback route and controlled boot acceptance
Backup procedureRetained-data fixtures, successful real copy and a separate restore test

For this machine, the nr shortcut runs sudo nixos-rebuild switch against the NixOS flake. It does not decrypt secrets, stage files or commit changes. Administrative authentication remains an operating-system boundary.

The verified maintenance example

The 16 October security pass enabled the stateful firewall, restricted local DNS, separated root’s editor from mutable personal Lua, removed the global browser decoder-sandbox override, and removed unattended Git publication. The non-deleting HDD copy was repaired and actually completed. Syncthing received HDD mount guards and private-file exclusions; Unicode filename conflicts were corrected while preserving both original byte copies.

The system build also runs a source-hygiene check and eight backup-copy tests with disposable data. A successful build, an activated configuration and a working service are recorded separately.

Recovery remains explicit

The same-machine HDD copies and legacy Borg repository are not an independent encrypted recovery plan. Credential recovery, backup encryption and a controlled system/driver upgrade remain separate work. Existing browser processes inherit their startup environment until restarted. An offline paired phone prevents an end-to-end sync observation even when local checks pass.

The original public configuration predates this local maintenance pass. Publishing a reviewed source revision is a separate step; the guides do not represent the baseline repository as already containing unpublished changes.

For software delivery, the same discipline applies: local tests, real persistence, browser journeys, CI, deployed checks and recovery each answer a different question. See how to read the evidence .

← All The Workbench documentsBack to top ↑