Two independently ready refuge tasks need lifting equipment. Does giving them one lift or two change only the timing—or which actions are possible together?
Circles hold the current token counts; boxes are work events. Solid green arrows consume tokens; dashed clay arrows produce them. A shared boundary is drawn as one place. This is the actual checked net, with the lab’s once-only event restriction.
● Green box: an admitted next event✓ Completed: cannot fire again here○ Pale box: waiting / not returned by the search
One state. Every admitted next step.
These are atomic decisions with no work already in progress. A joint step consumes all its inputs before producing any output.
THE SAME STEP, READ AS AN SMC ARROW
From tokens to compositional execution
In a symmetric monoidal category (SMC), executions are arrows between resource markings. Inspect a possible step before taking it: its task arrows are tensored with the identity on any resources left untouched. The step buttons above actually advance the marking.
Then, or alongside? Two consecutive step arrows compose as g ∘ f when the first output marking matches the second input marking. A joint step tensors the task arrows, but is admitted only if all input tokens are available at its start. With one shared lift, Frame then Roof and Roof then Frame can both be legal while Frame ⊗ Roof cannot start together.
These are count-token arrows in a commutative monoidal category. Occurrence identities and the joint-start check are retained separately: the category alone can forget token-reuse distinctions. See the interpretation and its limits.
Generate the lawful orders
Every selected task must finish, including tasks with no output. This essay composes a fixed work scope; the forward generator separately explores alternative work branches.
A representative schedule
Joint steps become batches; outputs are released at each task’s own finish. This is one replay-checked schedule under toy durations, not an optimum or proof of timed equivalence.
Retain the requirements, including blocked work
A standalone history records what happened with one supply. An open contract must also retain work waiting for its boundary inputs. We retain those demands, outputs and internally feasible states, then close the connection under its declared supplies.
What the equality does and does not establish
The two routes compare completed-event states, full count markings and jointly enabled steps for the declared finite fragments. The product and source-gluing constructions share a traversal routine; independent oracle tests separately check gluing, markings, labelled-token steps and complete orders. A cutoff means unknown. Equality is neither a theorem about arbitrary open nets nor evidence of engineering or human value.
Each admitted step interprets its event signatures through tensor, composition and untouched token context. Ownership and count semantics are explicit; no canonical token ancestry or exact elementary work tree is inferred.
Local tokens and unconnected input supplies belong to this scenario. A connected input’s old external supply is discharged; changing it cannot create a second token.
CURRENT FINDING
Preparing the example…
Inspect or edit the full construction
Change the fragments, event signatures, durations or named port connections. Checking validates the whole construction. An invalid edit keeps the last valid example available; it cannot silently become a result.
Keep and reopen a construction
Save one copy in this browser or download a portable JSON file. Opening a copy recomputes the result; imported findings are never trusted. No file, repository or remote model is overwritten.
No browser copy saved in this session.
EXTENDING THE SAME BOUNDARY CONTRACT
Can we store less without forgetting which work is done?
Two suppliers, A and B, each produce one support token. C consumes one. Every named task must occur once. This comparison extends the boundary contract above: store exact resource rules instead of expanded state tables. Choose an initial token, then try C first.
Why merely keeping the token count fails
After A and after B there is one token, but different work remains. If those situations are merged, the A-step available after B is also offered after A. The apparent sequence A → A repeats a once-only occurrence. Individual edges are insufficient: a continuation must work from every situation its summary represents. It already fails our promise to report named completed work.
What this attempt found: named completed-work observations prevent state merging in this model. An exact rule encoding can nevertheless avoid storing the expanded tables. Added supply changes the possible work while the receiver’s rule stays adequate. This is a bounded construction and an elementary structural argument, not a new general compression theorem.
The controls affect only this demonstration; reload resets it. They do not change the main workbench or save a project record.
PURPOSE & FOUNDATIONS
From reusable work packages to lawful plans.
The aim is to compose recurring pieces of work without imposing an unnecessary order or inventing resources. Petri nets make the mechanism explicit; monoidal structure describes how its executions combine. This lab proposes a boundary contract that retains resource demands, partial states, outputs and joint steps. The papers supply the foundations; the restricted contract and its checks are our construction.
Petri nets, open boundaries and monoidal structure — the three layers
1
Petri nets: what can happen?
A circle holds tokens representing resources or conditions. A task box consumes its input tokens and produces its outputs. Borrowing and returning a lift is a pair of ordinary arcs, not an unlimited permission to use it.
Lift : L → L ⊗ R
L is an available lift; R is an installed roof. The token’s return permits later reuse. This lab separately marks selected events as once-only.
2
Open Petri nets: how do pieces meet?
Expose places through named input and output ports. Connecting compatible ports identifies their places: the two packages now refer to the same resource. Two names must not become two lifts.
The lab uses restricted wiring and an explicit supply convention; the paper treats a more general construction.
3
Monoidal structure: how do actions combine?
Markings are objects; executions are arrows. Composition (∘) joins actions through matching intermediate markings. Tensor (⊗) combines resources or actions alongside one another. An identity carries untouched resources through a step.
We use its commutative count-token interpretation, a particular SMC. Equal arrows alone do not establish that tasks can start together.
Connecting open nets assembles a mechanism; composing arrows describes an execution within it. These are related operations at different levels. The practical question comes from Wynn & Clarkson’s ASM2.0 work: can reusable process descriptions earn their maintenance cost when context changes?
Scope: this is a finite, once-only, selected-work construction—not a complete engineering model or general WBS generator. The forward generator explores alternative work branches; the feedback trial handles review and rework. Here, timing is an added illustration, and a lawful execution need not fit one elementary WBS tree. Follow the full source → construction → limitation map.
WHY THIS MAY HELP
A reusable piece needs an explicit boundary.
The benefit is modest and testable: change a refuge method or resource arrangement, and see whether the combined work keeps its lawful alternatives. The source literature’s warning remains: context can make reusable process modules more trouble than they are worth. This example tests one piece of that problem.