What the foray is seeking
The catalogue places Gimmer under Processes to plans (110). This is App 20 in the main Gimmer collection, where a chosen execution projects to a timed schedule and a stage-based WBS. The separate mountain-refuge lab investigates forward generation of lawful work alternatives. This trial uses the main collection’s existing scheduler. Here the schedule is evidence for choosing among three stipulated process hypotheses. This is a conditional lift; reconstructing a process from an arbitrary WBS remains open. One should be able to say which choices a schedule preserves, and what extra evidence a planner needs to recover possible continuations. App 20 already produced two different processes with an identical complete 14-task baseline. This essai changes the direction of enquiry. Given rival explanations of the observed sequence, design a trial that discriminates between them.
The trial
Keep the existing net, durations, scheduler and plan projection. Keep its ordinary model and its alternative in which the roof acquires and returns the walls-complete token. Add one rival mechanism. An exclusive shell permit can be held by either trade and must be returned on completion. The permit prevents overlap while allowing either order. It is a new assumption about this fictional project.
Cross one or two units of each of crew, tools and lift with two deterministic priorities. Those 16 interventions produce 48 complete simulations. Compare the three models under the same intervention, by their entire timed task lists and by their finish dates. All fourteen tasks must finish in every run.
| Probe | Ordinary | Walls must finish first | One shell permit |
|---|---|---|---|
| Baseline, walls first | Walls 19, roof 23 | Walls 19, roof 23 | Walls 19, roof 23 |
| Double all three, roof first | Walls 19, roof 19 | Walls 19, roof 23 | Walls 22, roof 19 |
| Finish under that probe | Day 31 | Day 34 | Day 34 |
What distinguishes the explanations
The second probe is the only one of sixteen that separates all three complete task lists in one run. Giving the roof priority tests the direction of a dependency. Increasing all three capacities tests whether activities can overlap. The two start times together answer both questions. Either timestamp alone leaves two candidates tied in that probe.
Even the vector of finish dates across all sixteen interventions leaves the handoff and permit indistinguishable. Both serialize four days of wall work and three of roof work, and utilities needs both completed. The finish date cannot answer whether the walls had to go first.
One visit suffices
With all three capacities doubled and the roof given first call, inspect the work once on model day 19. Both trades working identifies the ordinary net. Walls alone identifies the handoff. Roof alone identifies the permit. The same three answers remain distinct from day 19 up to, but excluding, day 22. At day 22 all three show walls alone. From day 23 up to day 26 the three answers are again distinct: neither trade, roof alone and walls alone respectively. Intervals include starts and exclude finishes.
The field record can therefore be one three-way observation. It need not contain two timestamps or the whole schedule. One yes/no answer alone cannot identify three candidates. The later window also explains why an observation needs a time attached to it. “Walls alone” has a different meaning at day 19 and day 23.
Which decision needs the full trial?
If a planner only needs to know whether the roof can start before the walls, give the roof first call with one crew, one tool unit and one lift. A day-19 visit separates the handoff (walls working) from both other candidates (roof working). This trial needs no extra capacity. It leaves open whether another crew could work alongside the roof.
If the question is whether hiring enough resources could bring the finish forward through overlap, compare priorities with doubled capacities. Only the ordinary mechanism reaches day 31. The handoff and permit both finish on day 34 whichever trade gets first call. A finish alone therefore answers this narrower question. The handbook question “Can both trades overlap?” and the question “Must the walls go first?” call for different evidence. Those statements concern the three proposed mechanisms and the stipulated scheduler.
Interpretation and bounds
A planner could use this as a question for the method owner. What would prevent the roof starting first, and what would prevent both trades starting together? Changing the model is cheap. A physical trial requires its own authority and evidence. The three mechanisms are candidates supplied by this essai. Other explanations may fit all its observations. The deterministic priority rules select executions and the durations are invented.
The route contract supplies the distinction between a process and its projection. The new comparison uses the existing executable Petri core directly. A second diagram language or scheduler would add little to this discrimination test. Baez and Master's distinction between operational and reachability semantics is relevant background. This trial compares timed projections of selected executions.
Route boundary (repository access) · Original route contract (repository access) · Earlier two-model witness (repository access) · Open Petri Nets
Reproduce
Download the complete check and unpack it. From the unpacked folder, run node research/which-rule/run.cjs. The program uses the supplied scheduler and prints the results of all 48 simulations. Compare its output with the included recorded results (JSON).