4.0 KiB
PR#3: The staircase tells frontend and backend apart by authorship
WARNING: This document describes only the change applied in this PR. It may already be outdated once the next PR is merged. Historic PR documentation is not maintained along with new PRs; treat it as a snapshot.
Problem
The staircase view "Engineering Activities by Software Criticality" gives a human the production code review of a frontend and of a backend. With AI review agents, both are shared work of AI and human; only critical backends and systems keep a human review.
Changed alone, that would leave frontend and backend apart only by how binding their activities are: in authorship they would differ in "Performance / Resilience Review" only.
Non-Goals
- No new engineering activity, no new software type.
- No change of the other views; the profiles view does not model the activities per software type.
Specification
Feature: authorship of frontend and backend in the staircase
Scenario#3.01: the production code review of a frontend and a backend is shared work
… so that the staircase shows the review agents next to the human review.
- Given the staircase view is shown
- When the visitor looks at "Production Code Review"
- Then the frontend shows "Optional · AI + Human" and the backend "Suggested · AI + Human", and the critical backend and the critical systems still show "Required · Human"
Verified by
- the browser check below; there are no automated tests yet
Scenario#3.02: a frontend has no API or contract review of its own
… so that the staircase does not ask a frontend for a contract that its backend defines.
- Given the staircase view is shown
- When the visitor looks at "API / Contract Review" for the frontend
- Then the cell shows "n/a", and its tooltip says that the backend reviews the contract
Verified by
- the browser check below; there are no automated tests yet
Scenario#3.03: AI writes the behavior tests of a frontend
… so that frontend and backend differ in who writes the scenarios.
- Given the staircase view is shown
- When the visitor looks at "Behavior / Scenario Tests"
- Then the frontend shows "Required · AI" and the backend still "Required · AI + Human"
Verified by
- the browser check below; there are no automated tests yet
Solution
The cells of the frontend and the backend column in the staircase, inside the srcdoc of its iframe:
| Activity | Frontend before | Frontend after | Backend before | Backend after |
|---|---|---|---|---|
| Production Code Review | Optional · Human | Optional · AI + Human | Suggested · Human | Suggested · AI + Human |
| API / Contract Review | Suggested · AI + Human | n/a | Required · AI + Human | unchanged |
| Behavior / Scenario Tests | Required · AI + Human | Required · AI | Required · AI + Human | unchanged |
- "n/a" uses the grey style of "Not normally used"; its tooltip says why the activity does not apply.
- The description of "Production Code Review" loses "Human", since the activity is no longer human-only: "Inspection of implementation choices and maintainability".
- The cells were changed on the unescaped document; escaping it again gives back the rest byte for byte.
- Version 1.1.0, since the content changes.
So frontend and backend now differ in authorship in "Performance / Resilience Review" and "Behavior / Scenario Tests", and in how binding they are in five more activities.
Browser Check
Checked on a local HTTP server in Chromium on 2026-10-14: the three rows show the cells of the table above, the tooltips the new texts, and the other cells and rows are unchanged.
After the merge, the human installs the new version.
Open Questions
- Which existing activities tell frontend and backend apart besides the code review? Decided by the human: "API / Contract Review" does not apply to a frontend, and AI writes its behavior tests.