Essai 010 · Scope change interfaces

The Tangled Triangle — who must agree after scope changes?

Toy scenario

The control centre’s civil, systems and assurance teams hold local accounts of the same handover. Their owners differ legitimately; their shared revision, gate and evidence facts need review.

Try: split Gate works into two packages. Follow both the copied green and red interfaces, then require evidence agreement on a child’s interface with Controls. More obligations do not necessarily mean more risk.

Change the scope

A split copies the parent’s stated handover facts and incident overlaps to each child. Cost and exposure points are divided equally; the crew allowance is inherited. These are explicit bookkeeping assumptions, not new estimates.

Loading the illustrative accounts…
JSON controls
Declared field checks—
Interfaces needing review—
Δ cost allowance (£k)—
Δ largest package crew—

Declared interfaces

Solid green: declared fields agree · Dashed red: review values · Dotted amber: no fields checked

Baseline → current

Cost is an illustrative allowance in £k. Crew is the largest single package allowance, with no timing or concurrency model. Exposure points are an additive bookkeeping proxy, not a probability or a schedule forecast. Missing inputs leave the affected comparison unknown.

Changes applied

    What exactly agrees?

    Each field selects the same named fact from both accounts. Values retain their JSON types: 2 differs from “2”; object key order is irrelevant and array order matters. Missing, blank or null values are unknown, including inside a structured fact.

    Baseline and current facts on every current interface
    Interface / owner pairFactBaseline valuesCurrent valuesCurrent check
    Inspect all local accounts and the retained baseline
    Original accounts remain in the baseline after a split
    AccountBaselineCurrent

    Integration obligations

    Items follow interface order; they are not ranked by severity. Owners must establish whether the facts or the comparison rule need correction.

    ItemInterfaceOwner pairReview required

    This label records review intent. It performs no repair, optimization or safety assessment and changes no results.

    Expert JSON import and change controls

    Import an object with nodes and overlaps arrays. Importing sets a new baseline and current model. Invalid input leaves both intact. Exported obligation files include these snapshots; import their current object to continue an exported case.

    Changing the action supplies an example payload. Fields are direct property names, not dotted paths. An empty field list is allowed but remains “Not checked”. Null or omitted metric values remain unknown; supplied numbers must be finite and nonnegative. Retired account and interface IDs cannot be reused in the same case; reset or import starts a new case.

    Current importable JSON
    Construction, sources and limits

    This is a finite checker of explicitly declared interfaces. Each local account supplies handover facts; a field name is the projection used to compare two accounts. Passing means equality for the declared, known fields. A disconnected or incomplete interface graph can omit important questions, so passing does not establish project-wide completeness or approval.

    Splitting a package changes the graph and copies incident checks. Where the handover facts and field lists are inherited unchanged, their agreement or disagreement survives the split. New disagreement needs changed facts or additional tests. Editing the field list changes what is tested. Removing a troublesome field can make a check pass without resolving the original disagreement; the baseline and change log preserve that distinction. The split does not establish a topological cover refinement because this model does not define geometric regions or a fixed cover.

    Rosiak’s Sheaf Theory through Examples, chapter 5 (pp.148–149), requires compatible local sections and a unique gluing that recovers them. Ghrist’s cellular-sheaf examples compare data after the declared maps. Those ideas motivate explicit interface comparisons here; this page performs no sheafification, diffusion, global-section search or feasibility proof. See the foray’s public source trail.

    This essai studies how scope edits change review obligations. Essai 009 instead applies a stated policy to competing delivery estimates; the Triangle’s spatial lab checks selected location constraints.