Chapter at a Glance

Engineering knowledge can be reused when versions, configurations and evidence applicability remain clear.

Learn / Chapter 11

Engineering Knowledge at Scale

Which definitions belong together, and when can their evidence be reused?

Learning objective: identify a coherent configuration, distinguish identity from revision, and review evidence applicability before reuse.

Prerequisite: Governance and Assurance.

Manual configuration and reuse review using the small V0 demonstrator. This lesson adds no configuration platform, approved product baseline or scale benchmark.

The same identifier is not the same configuration

A colleague finds a passing result and attaches it to a new monitoring proposal. The requirement ID matches, so they assume the result is reusable. Before accepting that reasoning, ask which criterion, geometry and method produced the result. A stable name lets us follow a definition through change; it does not say that its meaning or evaluated inputs stayed unchanged.

The problem becomes harder as more people reuse sources and results. We can practice the necessary reasoning with this small demonstrator: identify the definitions that belong together, inspect their differences and qualify the evidence attached to them. More records would increase the review burden, but no large dataset or infrastructure is needed to expose the mistake.

Here, configuration means an identified, coherent set of definitions used for a particular analysis. Baseline means the named starting set for comparison. These local teaching terms do not designate an approved product baseline. The canonical sources remain the reference for this lesson; temporary proposals are separate candidates, not replacements for those sources.

Identify the set, not just one member

Start with the need, criterion and scenario, then the model and test registration. The need supplies intent; the other records define what the calculation and checks mean. Their typed relationships provide a route through the sources. Follow the actual relationship fields rather than assuming every arrow denotes the same dependency.

The table reads current identifiers and values from validated canonical sources at build time. It identifies their roles, not a historical execution. To interpret an older record, inspect its evaluated definitions and resolve surrounding sources at the recorded revision. A source link to main can later show a different state.

Members of the current reference context
SourceRoleWhat else must be identified?
NEED-001Intent motivating the criterionIts wording and revision, not only its ID
REQ-COV-001Criterion: 50%Statement, basis and evaluated scenario/model links
SCN-001Area 100 × 100; center (50, 50); radius 35 model unitsExact inputs used by the candidate
MODEL-COV-001Model version 0.1.0Implementation revision, method and assumptions
TEST-COV-001Registered canonical assertionsActual assertion execution, separate from analysis generation

Identity, revision and execution answer different questions

A requirement ID identifies the record. A source commit identifies committed repository state. The model version labels the model, while execution metadata identifies a particular run and its context. None of these alone identifies every evaluated input. A branch name can move; a model version does not describe the scenario or threshold.

A dirty checkout adds another qualification: HEAD alone does not identify its uncommitted changes. Generated records report working-tree state and embed evaluated definitions, but that flag is not a complete snapshot of every modified source. Record the actual differences needed for reproduction and do not attribute the whole working tree to the commit alone.

Coherence also requires meaning, not just matching references. A source bundle can contain resolvable IDs while combining a criterion from one candidate with inputs from another. Validation detects the contracts it implements; it cannot decide that this combination was intended, justified or covered by a historical result.

Reuse implementation deliberately

The browser and automated checks reuse one pure coverage function and evaluator. That avoids separately maintained formulas. It does not transfer the outcome of one execution to every consumer of that implementation. A new input still needs its own calculation, and a broader claim still needs evidence that the model can address it.

Evidence reuse is therefore a separate review question. Compare evaluated inputs, criterion, method, assumptions and intended claim. An unchanged number may have a different meaning after a threshold change. An unchanged criterion may have a different outcome after a geometry change. A shared function supports consistent computation, not universal applicability of its earlier results.

Review a proposed reuse

  1. Inspect the reference context above. Identify what its build-time source view can tell you and what historical context it cannot supply.
  2. Try the radius proposal. Inspect its exact field difference and proposed definition. Reset before trying the threshold proposal.
  3. Explain why unchanged coverage in the threshold example does not preserve the evaluation's meaning. Keep both candidates separate.
  4. Break a reference, observe withheld analysis, then restore/reset. Explain why passing structural checks would still leave an applicability review.

A definition change, not an approval

Temporary, unapproved browser proposal. This workspace changes a copy of the definitions. It does not edit YAML, execute TEST-COV-001, retain evidence or approve a requirement. Reset restores the baseline; leaving or reloading discards edits.

Canonical baseline from this site build

Requirement REQ-COV-001: 50%. Scenario SCN-001: radius 35, center (50, 50), area 100 × 100 model units.

Under reference scenario SCN-001 and the V0 model assumptions, the monitoring configuration shall cover at least 50 percent of the defined area geometrically.

FAIL — 38.52% estimated coverage. Margin: -11.48 percentage points.

This is a build-time calculation, not a displayed CI verification artifact. The baseline stays unchanged while you edit.

Temporary proposal controls

Threshold: 0–100%. Radius: 0–200 model units. Choosing a different change restores that field’s baseline value.

This edit applies only to the temporary copy. Its source_need must resolve to the existing need.

Enable JavaScript to validate and evaluate a temporary proposal.

Exact field differences

The requirement statement and basis change with the threshold so the proposed definition stays consistent. These differences describe the current trial, not a committed revision.

Canonical value → temporary proposed value
FieldBaselineProposal
Inspect the validated proposed definition
No proposal evaluated yet.

Both calculations use MODEL-COV-001 v0.1.0: a 100 × 100 cell-center area estimate. No physical sensor calibration, terrain, occlusion, directional effects, time, uncertainty, faults or communications are modeled. A PASS establishes neither operational suitability nor an optimal design.

Make change impact inspectable

Begin a reuse decision with the actual difference. A radius-only proposal changes scenario geometry while retaining the criterion. The threshold proposal changes criterion fields while preserving that geometry. Neither is the unmodified canonical configuration, and their separate results must not be combined into a third candidate that nobody evaluated.

Follow affected source relationships to identify what needs checking again. A broken need reference blocks analysis; a valid changed criterion may still lack an engineering basis. Existing canonical assertions do not claim reference-test execution against every proposal. Use the actual command logs and record semantics rather than interpreting the test ID as universal verification.

Leave the next engineer a reviewable handover: the intended claim, identified definitions, actual differences, evidence inspected and unresolved questions. This is more useful than a list of unexplained links. State whether evidence is unavailable or outside scope, and identify the next check or responsible review instead of silently carrying forward a selected PASS.

Review hypothetical reuse cases

These are designed cases, not reports of actual corrupted bundles or executions.

An old PASS has the same ID but a different criterion

Compare the evaluated criterion and inputs. The ID alone does not establish applicability; inspect or rerun the relevant evaluation and justify the intended claim.

A bundle mixes definitions from different revisions

Identify the intended configuration and actual differences. Resolvable IDs do not prove that this combination was evaluated. Do not invent a combined result.

A dirty checkout is attributed only to HEAD

HEAD identifies committed state. Inspect uncommitted differences and actual record inputs; a dirty-state flag alone cannot reconstruct all changed sources.

The same model is reused for a new scenario

Reuse the implementation, validate new inputs and execute the relevant calculation/checks. Review assumptions and evidence applicability; the old result does not automatically verify the new scenario.

Inspect actual records before reusing their conclusions

Follow the repository reuse review to inspect canonical and proposed definitions, differences, calculations and provenance at an identified checkout. Keep assertion execution, reference evaluation generation and proposal analysis distinct. The written template records reasoning, not an approved baseline or a new engineering evidence format.

No historical CI record is loaded here. Browser proposals are temporary; retained artifacts expire after 30 days and may require sign-in. Local reproduction is a new execution. Even a coherent configuration and successful checks cannot establish physical detection, operational suitability or an optimal design under V0's omitted behavior.

Check your reasoning

1. Does an unchanged ID establish unchanged meaning?

No. Inspect the definition and revision, including criterion, inputs and relationships.

2. Is shared implementation the same as reusable evidence?

No. A shared function computes consistently; evidence applicability depends on the evaluated configuration, method, assumptions and claim.

3. Does a dirty-state flag fully identify an execution's sources?

No. It warns that HEAD alone is insufficient. Inspect the actual inputs and relevant uncommitted differences.

4. Can two passing candidates be combined without another evaluation?

No. Their separate results do not establish a result for an unevaluated combination.

Next: connect the whole review

We can now identify what belongs together and explain why a reuse decision needs evidence of its own applicability. Continue with Chapter 12: The Integrated Engineering Environment for an integrated exercise connecting operational intent, definitions, execution, evidence interpretation and responsible review without losing those distinctions.

← Chapter 10 · Learning path · Inspect engineering sources →