5.4 KiB
Native Capella Scenario spacing observation
A bounded primary observation in Capella 7.1.0.202606231255 on Windows, using isolated workspace capella-layout-benchmark-20260912 and project IncidentScenarioBenchmark. The parent integration session performed the authorized Computer/Sky interactions and reports closing Capella normally afterward. This archive's independent reviewer inspected the saved screenshots and files; the reviewer did not perform native input. observations.json retains the operator's actions, timestamps and capture recovery notes.
Scope: one Operational Analysis scenario, two activity lifelines and one interaction. This does not establish chronology-crossing/reordering behavior, activation nesting, or a general Capella gesture policy.
Observed workflow
| Evidence | Observed result |
|---|---|
| 01 empty scenario, 02 lifeline, 03 interaction choice, 04 message | Created two activity lifelines and one Report incident interaction. Completing the pending interaction used an endpoint click and the Create a new Interaction choice. |
| 05 body drag, 06 selected body drag | Two arrow-body drag attempts produced no visible movement, including the attempt after selecting Sequence Message. This does not prove that Capella never supports body movement. |
| 07 receiver execution drag | Dragging the small receiver execution down 79 pixels moved the whole horizontal message from screen y334 to y413, with its receiver execution. |
| 08 lifeline header drag | Moving the left header from screen x430 to x510 moved that lifeline and its attached message endpoint 80 pixels. |
| 09 first Undo | One Undo restored the left lifeline to x430 while the message remained at y413. |
| 10 second Undo | The next Undo restored the message to y334. After Save, both captured files exactly matched the baseline. |
The operator recorded two capture timeouts: after the first activity placement and after the initial interaction drag. Each was recovered by selecting the returned window afresh and taking another screenshot; the input was not repeated. These failures/recoveries are preserved in observations.json rather than omitted or treated as extra interactions.
Saved-file comparison
Before .capella, after receiver-drag .capella, and restored .capella are byte-identical: 26,594 bytes, SHA-256 d8764060c6f3d11ddf24a91e20955fe035725cfe35019fdd514688d07118c056. Thus this saved move did not alter the captured semantic model file.
The before .aird and after receiver-drag .aird differ on exactly five lines, with no line insertion/deletion:
| Before-file line | Field change | Interpretation |
|---|---|---|
| 11 | representation changeId changed | Presentation revision metadata. |
| 100 | receiver execution notation bounds y53→132 | +79 vertical placement; width/height unchanged. |
| 125 | message relative bendpoint y offsets 0→79 | Both ends of the existing horizontal path moved vertically. |
| 183 | receiver execution absolute-bounds y153→232 | +79 corresponding presentation filter. |
| 205 | SequenceMessage absolute-bounds y153→232 | +79 corresponding message filter. |
Object references identify the changed execution and Report incident SequenceMessage; manifest.json preserves the exact five before/after lines. This is a presentation-file change for the recorded receiver-execution move. No separate saved snapshot was taken after the horizontal header move, so its isolated serialized delta is not claimed.
After both Undos and Save, restored .aird equals before byte-for-byte: 29,469 bytes, SHA-256 c24db589f3a551b8ea0f8d00d2dc6729909157e2a3ac4ebc0ca69cd141393c89. The two separate visible Undo steps and complete saved restoration support the observed local history behavior.
Integrity and implication
All 27 original evidence artifacts are preserved without content changes: ten screenshots, their ten metadata files, observations and six source snapshots. The screenshots were originally named .png but contain JPEG/JFIF bytes. Published names use .jpg with explicit originalName→path mapping in manifest.json; no transcoding or duplicate image copies occurred. All ten metadata SHA-256 values and encoded 1010×761 dimensions were independently verified. The manifest includes every artifact's byte length, SHA-256 and Git blob SHA-1; local -text attributes preserve source line endings.
This evidence supports separating view-local message/lifeline spacing from shared semantics for this simple case. A Studio gesture should make its target discoverable and separately verify preview/commit equality and one-step Undo. Those are design implications, not Capella-wide behavior or Studio implementation acceptance. The archive is observation evidence, not a packaged benchmark project or native Studio test.