Aglet

Learn from recurring image orientation drift

Repeated orientation defects usually reflect an unstated agreement about whether metadata is honored, normalized, or removed. Learning should preserve the marked images that reveal the difference and document the expected visible result at each workflow boundary, including edits and final delivery.

Keep the lesson for the next incident

  1. Cluster the rotation cases

    Group findings by metadata loss, double application, preview-only rendering, processor transform, mirrored content, and downstream reader differences. Record landmark orientation, dimensions, and the first boundary where views diverged.

  2. Write the orientation contract

    State whether previews and outputs honor or normalize metadata, how pixels are transformed, and which consumer defines final appearance. Include sideways, upright, and mirrored fixtures with expected landmarks and dimensions.

  3. Trigger fixture review

    Require the marked images when changing preview rendering, image processing, cropping, metadata handling, storage, or download paths. Assign an owner for the contract and record any consumer-specific assumption that must remain visible.

What to carry forward

The learning record is useful when the orientation rule covers both pixels and metadata, landmark fixtures remain available, and final consumers are named. If consumers intentionally differ, document the boundary and avoid presenting preview appearance as a universal guarantee.

Technical background: MDN reference.

Keep the decision with the work.

Use a Work Item in Aglet to record the problem, the evidence you have, and the next decision. Add an owner and priority, then keep updates in the discussion so the next person can follow the reasoning.

Create an account See the product workflow