Scenario movement: source evidence and discriminating workflows
This extends the earlier anchor-plan experiment and the layout contract candidate. It informs the user's #20 request for ordinary diagram manipulation and the application specification's requirement to learn from Capella. It does not select a final gesture policy or claim native Capella evaluation.
Verified primary-source behavior
Capella7.1.0, commit3153b930a4284f6c82952658f1e8d8d6d8c8f340, connects movement of messages, Create/Delete, returns and executions to DoReorder. Its service moves sending and receiving interaction fragments separately, then moves the existing message in ownedMessages, retaining identities. See tool mapping and implementation.
Sirius documents separate blank-space insertion/removal using Ctrl-drag on the vertical ruler or Ctrl+Shift-drag on the canvas. Arrange All preserves order while reducing spacing. Create moves a lifeline's birth/header and Delete its death/bottom. The user guide describes these operations. Sirius7.6.1, commit0a7288dda7c0069659b54091ba811abab0d247c2, spacing assertions demonstrate that inserting space within an execution leaves the call fixed but moves the return and extends the lifeline; inserting below leaves both fixed.
Sirius destruction movement tests permit ordinary drag from y400 to230 across an unrelated message at300, preserving that message's coordinates, but reject death before earlier use of the receiver. Crossing is constrained by affected lifelines, not universally forbidden.
Call/execution/return movement can be coupled, including self calls. Sirius has distinct predecessor values for starting and finishing event ends (reorder tool contract); synchronous call assertions check coordinated movement and birth/death limits.
Messages and executions must keep their endpoints within the same operand. Frame coverage tests reject enclosing a fragment within one covering fewer required participants. Whole-frame movement assertions translate enclosed calls and can enlarge a containing execution. This does not establish unrestricted semantic reparenting.
Interpretation and alternatives still to evaluate
The evidence rules out assuming every sequence drag is cosmetic. It does not justify copying Capella event-end reorder directly: ARCADIA Studio currently orders complete occurrences in Scenario/Branch sequences, and validates reply dominance, lifetime and operand relations at that granularity. Independent Capella event-end ordering is a material semantic difference.
Candidate A uses direct dragging for presentation spacing and a separately discoverable semantic reorder action. It reduces accidental semantic changes but may make the common order-correction task less discoverable.
Candidate B uses one drag with explicit spacing or semantic-reorder preview according to the prospective position, preflighted against actual chronology/operand constraints. It may better support direct manipulation but must make a semantic transition clear before commit and prevent accidental changes near a boundary. No coordinate-only finite check authorizes that semantic operation.
Both candidates need actual geometry feasibility, clear rejected-move feedback, cancellation, one transaction for model and presentation, exact Undo and saved/reopened geometry. Existing floor-clamp counterexamples remain rejected; silently snapping to a different chronology does not satisfy an exact authored coordinate request.
| Engineer workflow | Observable discriminator |
|---|---|
| Make room for a commissioning review note without changing behavior | Predictable spacing; no unintended semantic order mutation near another row |
| Correct vent-then-isolate to isolate-then-vent within an operand | Discoverable reorder, comprehensible semantic preview, one coherent Undo |
| Extend service duration around a self call | Coupled reply/activation effects visible before acceptance; start and duration operations distinguishable |
| Move shutdown earlier across unrelated telemetry | Valid crossing allowed; death before last receiver use rejected clearly; exact lifecycle anchors retained |
| Adjust a fault-handling message near an alt divider, then move its frame | No accidental operand change; whole-frame translation distinguishable from semantic reparenting |
Record task completion, unintended semantic mutations, preview comprehension, cancellation, Undo and reopen behavior. Pointer distance alone is not an adequate evaluator. The user request includes resizing and ordinary controls beyond this initial sparse lane/message storage slice.
Evidence limits
This was source inspection, not executed native Capella/Sirius validation. Exact modifiers, feedback and persistence/Undo behavior have not been observed locally. Capella and Sirius revisions were inspected separately; their precise bundled dependency relationship was not established. Findings follow assertions rather than misleading test names. Real GUI acceptance remains required for the chosen product behavior.