Skip to content

story(docs): decide whether a derived anchor is model data #172

Description

@Zaba505

Description

As someone recording a measurement that does not run between two authored corners, I want to
know whether the point it runs to is model data, so that an annotation set is checkable rather
than a list of coordinates whose meaning is a convention somebody remembers.

Do not start this story. It is filed so that the question is visible and so the evidence is
not lost. It is blocked on a consuming repository first attempting the cheaper thing, and the
residue of that attempt is the actual requirement.

The evidence, from a real 62-dimension floor plan. Written as claims on edges, only about half
of its recorded spans anchor to two authored vertices, and the endpoints that do not fall into
a small closed set: a point at some distance along an edge (roughly four fifths of them), the
intersection of two edge lines extended beyond both segments, and the projection of one vertex
onto another edge. ISO 10303 AP242 named exactly these three in 2008 as derived shape aspects,
which is a strong signal the taxonomy is real and not an artefact of one project.

That standard also carries a boolean this format has no equivalent of: whether a referenced
element is an integral feature of the product or a derived one — a face or a centreline. Its
absence is what allowed the same house to be dimensioned face to face on one level and
centreline to centreline on another, with a 13% area discrepancy in one room and nothing able
to detect it.

The cheaper thing to try first: four fifths of that residue is a measured location. A
point 13'-7" along a wall can be authored as an ordinary vertex with a position claim carrying
its source, method and accuracy, and then the span is an ordinary edge between two vertices and
this story is unnecessary for it. What survives that attempt is what needs deciding.

And the distinction to preserve whatever is decided: what a measurement runs between is a
fact and is checkable. Whether it appears on a drawing is editorial and stays with the
consumer. Only the first is a candidate for this format.

Acceptance Criteria

  • Not worked until a consuming repository has attempted to author the points it measures to
    and reported what remains
  • An ADR deciding whether a derived anchor is representable, and if so listing the closed
    set — the record closes a set, so the members are listed
  • The decision addresses whether the integral-versus-derived distinction is carried, since
    that is the one with demonstrated failure behind it
  • It is argued on whether the thing stated is falsifiable, not on domain vocabulary — an
    anchor passes 0010's test, so 0010 is not the record that decides it
  • If the answer is no, SPEC §11 gains a line so the absence reads as a decision

Related Issues

  • 0007 — the pattern by which something obvious and unfalsifiable is excluded rather than
    deferred
  • 0011 — the same pattern for an expression language

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

gateIts outcome changes what comes afterstoryA unit of user-facing work

Type

No type

Projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions