Shared boundary data across systems, electrical, electronics, mechanical and software, with Teamcenter as the system of record
Which formal semantic should carry the data between engineering domains?
Four candidates model the same test slice, connector J1 seen by five domains, and are scored against criteria that were committed before any trial ran. Every number below says whether it came from running the candidate's own tool (exercised) or from analysis (derived).
Verdict under the pre-registered rule
Offline phase: provisionalLoading results.
If the open amendments are accepted
Scoreboard
G0 openness and G1 offline fidelity are pass or fail. C1 and C3: higher is better. C2, C4 and C5: lower is better. Teamcenter native is the baseline: it is scored so the contenders have something to beat, and it cannot win because it fails G0.
C4 Concepts invented
N native to the language, L from its standard library, E invented inside the standard's extension mechanism, X not expressible at all. Score: E + 2 × X. Select a cell to see the construct used.
C3 Validation: six planted defects
Each defect is one mutation of the clean model. Builtin: the standard validator caught it with no user rules (2 points). Authored: caught by a rule we wrote in the candidate's own constraint language (1 point). The clean model is the false-positive control.
C1 Fidelity: the 50 slice facts read back
Each candidate's own parser read its model back and a query returned the facts. A mutated copy (J1.pin3 carries S_PWR28) is the control: it must produce a difference.
C2 Mapping effort (derived)
27 native concepts from Cameo, Capital, NX, Xpedition, the software signal dictionary and Teamcenter, classified per candidate: direct 1, composite 2, extension 3, lossy 5. Phase 3 replaces this with measured connector code.
Trial notes
What the tooling did, including what broke. Recorded, not scored.
Amendments
Questions the trials raised about the frozen criteria. They are shown, scored as sensitivity cases, and decided by a person, never quietly applied.