Before you start#
Code and data are shared differently. Each editor has a separate Git clone. Its Git working tree holds that person's branch, staged and uncommitted changes. Shared data instead has a named information owner and approved location.
Mira analyses beam measurements; Leo reviews her report.
| Area | Mira | Leo | Boundary |
|---|---|---|---|
| Main input | Read | Read only if approved for review | Neither replaces it |
| Mira's run output | Write | Read the agreed report | Separate outputs for concurrent runs |
| Code | Own clone | Own clone if editing | Reviewed changes through GitHub |
Permissions state who may read or change files. An access-control list (ACL) expresses detailed permissions. A readable folder is not automatically writable.
Leo cannot edit the main input. Is that a defect if his task is to review the report? Decide the necessary boundary before opening the explanation.
Why read and write roles differ
The reviewer needs the agreed report and provenance, not permission to replace measurements. Diagnose owner, group, ACL and intended boundary before a narrow authorized repair. Preserve conflicting outputs for the owner to resolve; never use shared credentials or make everything writable.
Keep separate Git working trees. Diagnose a failed permission before asking for a narrow owner-approved repair.