Skip to content

Format boundary — evidence and remaining work ​

The recovered decision separates Format, production descriptions and distribution. No new answer is needed for that separation.

Prior answer and later context ​

On 21 July, Joe approved the card proposing one canonical Format, with this qualification: “assuming confirming this is like image video etc”. The card proposed retiring the overlapping legacy field and treating source-page asset roles as collection information. Its claims that migration would be deterministic and cost nothing remain historical claims, not verified implementation facts.

The later approved channel stack uses Media type for distribution. The 10 August feedback on the whole classification page asked for a clearer presentation; it did not approve the later proposal to rename the mixed legacy list Production Form. That list includes Static Image and Audio Only alongside Live Action and Motion Graphics. A rename alone would preserve the mixed questions.

Selected system checks ​

On 5 October, read-only database queries confirmed that the legacy field remains on the content table and its read view. Selected local code still writes it from analysis and filters it in search. Another selected path dispatches analysis using Format. These checks establish real compatibility dependencies; they do not certify the deployed end-to-end path.

The stored pairings include three text records with production descriptions. Those pairings need source inspection before being called errors: a written page may describe other work. No media example was inspected for this recovery.

What follows ​

Retaining and documenting both fields avoids an immediate change but leaves the mixed list. Renaming the whole list does not resolve its meanings. The recommended engineering direction is selective reconciliation: preserve the approved questions, inspect each legacy value and its evidence, then define the mappings and compatibility changes together.

Before implementation, trace collection and format derivation, composed prompts, validation, writes, views, facets, saved filters and result presentation. Test human corrections and whole-piece versus moment scope. Do not turn stored production descriptions into distribution labels or infer campaign membership from format. No lossless mapping, automatic deletion or zero-cost repair is established by this review.

Technical evidence references ​

  • Exact original reply and card, compared alternatives, local consumer references and scope: docs/reference/2026-10/2026-10-04/format-prior-approval-lead.json.
  • Literal selected SQL and returned schema/value pairs: docs/reference/2026-10/2026-10-04/format-boundary-live-evidence.json.
  • July source: docs/reference/2026-07/2026-07-21/theratificationcnooslejuly21.md, card g-1-a; original recommendation in the adjacent ratification-console.html.
  • Later presentation feedback: docs/reference/2026-08/2026-08-10/schema-presentation-feedback.md.

Last reviewed 5 October 2026.

The BrandTrackers domain model. Source: git markdown, drift-checked against the live DB.