Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@ Batches may execute concurrently by default; the cap is controlled by the contra

The machine-decidable members of condition 2's page set come from metadata that already exists: a page whose frontmatter carries `type: overview`, `runtime-card`, or `card-index`; a page carrying `type: term` with `scope: shared`; and any page the selected profile's `Expression Layer Entry` registers as a canonical dependency map. "Other profile-registered hub roles" remains in force, but no profile slot registers one today, so that clause currently contributes no page and is a future extension point. Editing and creating differ: an existing hub page in the frozen manifest blocks concurrent activation and takes the exclusive or `serial-integrator` route below, while a hub page this batch creates does not block it and is reported as a candidate for that batch's hub synchronization step.

Condition 2 is time-invariant, and its reporting reflects that. Whether a queued batch's manifest edits an existing hub page does not depend on which batch is being admitted, how many batches are active, or what has closed since — the answer is the same at every moment until a structural Amendment changes the batch's `execution_mode`. Reported only through readiness, such a batch looks merely "not yet turn" and the defect stays invisible until it reaches the head of the Queue, so a whole mis-specified family is rediscovered one batch at a time. The consistency mode therefore reports every queued batch that fails condition 2 as a candidate over the whole Queue, naming the batch, the hub pages, and the current mode. It stays a candidate rather than an error because the repair is an Amendment and the Amendment tools refuse to run against a runtime carrying errors: a hard failure would wedge the instance out of its own repair path. The other admission conditions stay readiness-only, because each of them is a statement about *now* — a dependency not yet closed, a cap currently reached, a manifest currently overlapping active work — and becomes true or false as the run proceeds.
Condition 2 is time-invariant, and its reporting reflects that. Whether a queued batch's manifest edits an existing hub page does not depend on which batch is being admitted, how many batches are active, or what has closed since — the answer is the same at every moment until a structural Amendment changes the batch's `execution_mode`. Reported only through readiness, such a batch looks merely "not yet its turn" and the defect stays invisible until it reaches the head of the Queue, so a whole mis-specified family is rediscovered one batch at a time. The consistency mode therefore reports every queued batch that fails condition 2 as a candidate over the whole Queue, naming the batch, the hub pages, and the current mode. It stays a candidate rather than an error because the repair is an Amendment and the Amendment tools refuse to run against a runtime carrying errors: a hard failure would wedge the instance out of its own repair path. The other admission conditions stay readiness-only, because each of them is a statement about *now* — a dependency not yet closed, a cap currently reached, a manifest currently overlapping active work — and becomes true or false as the run proceeds.

For a complex batch, readiness additionally requires a current Work Spec
path/hash whose batch ID and ordered manifest equal B. The worker reads that
Expand Down
Loading