What the foray is seeking
The published Foray 165 brief puts intended work and threat-bearing processes in one account of changing state and resources. The first essai established how the wagon and the cattle reserve the same lane. Its timed examples gave one party priority by agreement. Here the agreement itself is on trial. Exactly which service promises could the two parties make together?
The trial
The wagon is ready at zero. The cattle arrive at minute 1 or minute 10. The gatekeeper sees an arrival immediately. Absence at minute 1 therefore identifies the late world. In one treatment the farmer also sends a truthful one-bit message before dispatch. Wagon transit takes 10 minutes, cattle passage 15, shuttering 8 and the pour 12. These are the existing model's durations.
For each world enumerate starts with the original token engine. Pair schedules across the two worlds. Without the early message, any wagon departure at zero must happen in both worlds or neither. At minute 1 the branches can diverge. Minimise two separate worst-case numbers, pour finish and farmer waiting, over a common rule. The maxima can come from different worlds.
| Information before dispatch | Undominated guarantees (pour by, farmer waits at most), minutes |
|---|---|
| Observe at the gate | (22, 9), (38, 1), (47, 0) |
| Truthful early message | (22, 9), (38, 0) |
The minute that matters
With the message, in the early world let cattle occupy the lane from 1 to 16. The wagon clears at 26 and the pour ends at 38. In the late world dispatch the wagon at zero. It clears precisely when the cattle arrive, and the pour ends at 22.
Without the message, starting the wagon at zero can leave the early farmer waiting nine minutes. Waiting to observe at minute 1 preserves the early-world plan. In the late world the wagon then clears at 11, one minute after the farmer arrives. Giving the late farmer immediate passage instead delays the pour to 47. The site sees a nine-minute improvement in its guaranteed finish at zero farmer waiting when it learns the answer before dispatch.
Why the search suffices
The engine checked 911 observation-consistent pairs of whole-minute schedules and 972 pairs with advance knowledge. Counts carry no probability. Replaying each witness checks exclusivity and pour completion against its event log.
What if the farmer sometimes misses the call?
At the dispatch deadline, suppose every received message is true, but sometimes none arrives. If either arrival time can produce silence, the gatekeeper must make the same opening decision in those two silent cases. The gate-only rule remains available throughout. These two bounds give exactly the gate-only service guarantees, however rarely the messages go missing.
If only the early arrival can produce silence, then no message by the agreed deadline means early. If only the late arrival can produce silence, it means late. In either case the original 38-minute deadline with zero farmer waiting survives. The report contract must say which arrivals can give rise to silence, and when the gatekeeper can judge a message absent. The four contracts were checked against both arrival histories. A real promise would need evidence for those conditions.
Could a resource table do this?
Yes. For each arrival, write the two possible lane orders. If the wagon goes first, it takes ten minutes and the farmer waits until it clears. If the cattle go first, the wagon starts fifteen minutes after their arrival. The pour finishes twelve minutes after both shuttering and delivery finish. Giving the first user any extra delay can only postpone the pour or prolong the farmer's wait.
| Gatekeeper's information | Earliest choices the table must cover |
|---|---|
| Gate only | A departure before minute 1 must be the same in both worlds. Departing at zero dominates a later departure before 1. Otherwise observe at minute 1, then choose the first user separately for each arrival. |
| Truthful message at opening | Choose the first user separately in each world at zero. Starting that order as early as its users allow attains its best pair of times. |
These cases also exhaust real-valued dispatch times. An independent calculation from the table recovers both exact frontiers above. The full model earns a narrower role here: it checks that starts and releases remain legal while the two stakeholder measures are extracted from the same execution. It gains no additional service promise over the table.
The choice of columns matters. Dispatching the wagon at opening gives identical project task times in both worlds: the pour finishes at 22. The farmer waits nine minutes in one world and zero in the other. Dropping the farmer's objective therefore merges two executions that give different answers to a real question inside this example. The same counterexample works in an ordinary table if the farmer's column is omitted.
The answer depends on exactly two arrival possibilities, an instantaneous reliable message, a single non-preemptible lane and a shared clock. Neither a frequency of disruption nor anyone's willingness to make the agreement follows from the calculation. The intervention worth investigating in an actual negotiation is the timing and reliability of the specific signal.
Original End, Ways and Means · Resource interpretation and source review · A mathematical theory of resources
Reproduce
Run node research/lane-agreement/run.mjs from the project root. The recorded frontier retains early- and late-world dispatch times for each guarantee.