The Document sentences are retired by ADR 0016 (2026-09-10). The library holds no save format, so there is no Document to keep progress out of, and no reader to drop a schema: 1 key. The decision itself stands, and is untouched: progress is not on Entry and not in the Field registry. The body stays as written, because an ADR records the reasoning of its day.

Progress is scheduling-plugin data, not a core Field

% complete is a project-planning measure. Core names what an Entry is (dated, kinded, spanning). A shift roster, a booking calendar, and machine uptime have no honest home for progress, the same way they have no home for Dependency or the pin flag (ADR 0002, ADR 0003). Shipping progress as a core Field would put a scheduling opinion on every Dataset, including ones that never install a scheduling plugin.

progress is not on Entry, not in the Field registry, and not in the Document. A schema: 1 key named progress is dropped on read, like any other unknown top-level entry key. S7's scheduling plugin stores Progress in plugin-owned data and registers it as a Field. weightedMeanByDuration still ships as a named Aggregator, because a consumer Field and that plugin both need it.

S4 landed this in source: Entry no longer carries progress; schema: 2 Documents omit it; fromJSON drops legacy progress keys (ADR 0005's core Field list and ADR 0008 align).

Considered options

• Keep progress as a core Field with rollUp: 'weightedMeanByDuration'. Rejected: it privileges percent-complete as universal, and D-S4-6 would overwrite authored group progress on every rollUpKinds parent with no per-Field escape.
• Leave Entry.progress as a non-Field passenger until S7. Rejected: no released consumers, and a dead key that update() refuses while toJSON still wrote it is worse than one removal.