6. Measure the right tasks and evaluate exposure
SYSE.36 is applied twice, to different software services. The application owner supplies the address-form acceptance meaning; the platform cannot substitute “deployment completed.”
| Subject and users | Eligible attempt and good result | Missing observation or decision |
|---|---|---|
| Supported deployment path used by developers, with its path/configuration edition named | One authorized request for an already verified artifact/configuration in a supported class. Good means actual requested runtime identity and bounded deployment test returned within that class’s agreed response limit. A pre-backend platform failure remains a failed eligible attempt. | Missing user-entry observation, unknown runtime or absent test prevents the good claim. Platform users/providers agree the objective and response policy. Without an agreed response bound, timeliness is not qualified. |
| Address application h2/c2, compared with the named h1/c1 control, used by parcel operators | One authorized view/edit/save of a supported address. The new form displays the correct components and preserves submitted text under the correspondence within the application’s own agreed bound. A version/snapshot or controlled test separates later legitimate edits from corruption. | Missing form observation, value/identity correspondence or candidate/control attribution leaves application success unestablished. HTTP 200 is insufficient. The application owner supplies meaning and its release holder decides reliance. |
The common comparison concerns correct address use under each form’s supported contract, not identical layout. The instrumentation must observe that meaning or expose the remaining gap. Each error budget and SYSE.37 alert, if used, belongs to its own service and task population.
With permission for the bounded exposure and a qualified stopping action, SYSE.35 evaluates application h2/c2, not the platform path edition. In one invented interval, nine general candidate requests occur but none uses the changed form. The relevant evidence is insufficient: the application-exposure result is inconclusive.
Meanwhile, all ten observed platform deployment attempts may have completed correctly within their own agreed and observed response bound. Those successes do not alter the application’s inconclusive result.
In an adverse form observation, the new screen hides a non-empty tail although the backend preserves D/L/T and the platform deployed exactly h2/c2. Application acceptance fails, so exposure stops or returns within the qualified recovery boundary. This does not prove deployment-path failure. Conversely, failed CI workers can interrupt developer tasks while the old application continues serving parcel operators correctly.