ECSS Commercial EEE Components -- Class 3 Alert Handling (space-systems/ecss/q6013-class-3-alert-handling)
Use when the task is clause 6.5.3 of ECSS-Q-ST-60-13C at the lowest assurance class: an alert or errata notice has arrived, and it has to become a screened, actioned and time-stamped response over parts the project may not be able to trace -- starting with whether the watch that caught it was a watch at all.
Domain quick reference
- The watch is graded before any single alert is. An alert handled well proves nothing about the alerts nobody subscribed to, so the source coverage and the review interval are read first and reported beside the screening result.
- Two channels are not optional even here: the manufacturer notice service and the agency alert system. A watch missing one of them is not a thin watch, it is an absent one, and the remaining sources do not substitute.
- An alert that arrives through a source the watch does not subscribe to is a finding in its own right, however well it was handled. It was caught by luck, and luck does not have a coverage figure.
- What the project can show about a holding decides what a screening result is worth. A lot-traced or date-code-bearing holding screens definitively; a holding known only by part number screens presumptively; an untraced holding does not screen at all.
- The presumption runs toward affected, not away from it. A holding whose part number matches and whose date code cannot be shown has no evidence excluding it, and nothing at this class licenses an exclusion nobody can demonstrate. It is carried as presumed affected and counted separately from the confirmed ones, so the presumption stays visible.
- An untraced holding is a different failure from a presumed-affected one. It cannot even be matched to the alert by part number, so the response is traceability recovery rather than quarantine, and the alert cannot be closed while one exists.
- Criticality decides whether a presumption is good enough to act on. At or below the criticality floor the holding is escalated for traceability recovery instead of taking the routine response, because a critical item should not be retained or scrapped on a guess in either direction.
- Where the parts are decides what is owed. Stock and kits can be quarantined immediately; assembled hardware needs a retrofit assessment; delivered hardware needs an in-service assessment owed to the customer, which is the most expensive branch and the one discovered late most often.
- Date codes are four digits, a two-digit year and a week, and they do not compare as strings once the year rolls over. Week 53 of one year is numerically larger than week 02 of the next, so each code is resolved into an ordinal before the inclusive range test.
- The response has two clocks, both in working days from receipt: acknowledgement, which is short and set by the alert category, and corrective action, which is longer. An alert with no corrective action recorded is still running against its clock.
Workflow
- Grade the watch: the subscribed sources against the mandatory set and the coverage floor under a named tolerance, and the calendar days since the last review against the review interval, with an age equal to the interval still inside it.
- Validate the alert: a recognised category and source, at least one affected part number, a date-code range that is not inverted, and a receipt date. An unknown key is refused rather than ignored.
- Resolve every holding's traceability grade into a screening resolution before testing anything, because the resolution decides which of the verdicts the holding can reach.
- Screen each holding. An untraced holding is unscreenable; a non-matching part number ends it; a match with a readable date code takes the inclusive range test; a match with no readable or resolvable date code is presumed affected.
- Derive each holding's corrective response from its verdict, its state and its criticality, escalating an unscreenable holding and a presumed affected critical item for traceability recovery.
- Count working days from receipt to acknowledgement and to corrective action, excluding weekends, counting to the assessment date where a step has not happened so an open item keeps accruing.
- Compare both counts with the deadlines the category earns, and report the affected, presumed-affected and unscreenable quantities separately.
- Close on the first blocking condition in order -- watch inadequate, holdings untraceable, acknowledgement late, corrective action late -- then on presumption where any holding was presumed affected, and only otherwise on a clean close.
Pitfalls
- Grading the alert and not the watch. An alert response is only as good as the channel that delivered it, and a programme subscribed to one source has no way to know what it missed.
- Reading a missing date code as outside the affected range. An absent value is unknown, not clear, and at this class it means the lot cannot be excluded rather than that it was.
- Folding the presumed-affected lots into the confirmed count. They are the ones a later traceability recovery can still move, and merging the two figures hides both the exposure and the way to reduce it.
- Treating an untraced holding as merely a presumed-affected one. It cannot be matched to the alert at all, so quarantining it answers a question nobody has asked yet.
- Acting on a presumption for a critical item. The criticality floor exists because scrapping good parts and retaining bad ones are both expensive there, and the recovery is cheaper than either.
- Comparing date codes as strings. The comparison only works inside one year; across a year boundary it silently inverts and clears exactly the lots at the edge of the window.
- Counting calendar days against a working-day deadline. It makes an on-time response look late over a long weekend and hides a genuinely late one.
- Stopping the corrective clock because the alert was acknowledged. Acknowledgement closes only the first clock; an alert with no recorded corrective action is still open and still accruing.
- Forgetting the delivered branch. Parts already with the customer are the ones an alert is most expensive to miss, and they are the least visible in a stores-oriented inventory extract.
Behavior contract (gate 3)
The watch grading across source coverage and review interval, the alert validation, date-code parsing and ordinal range test, traceability resolution, holding screening including the presumed-affected and unscreenable branches, corrective response derivation, the acknowledgement and corrective working-day counts against their deadlines and the closing verdict are exercised by the gate 3 contract test: scripts/test_q6013_class_3_alert_handling.py against scripts/q6013_class_3_alert_handling_logic.py (stdlib unittest, offline). Run: python3 scripts/test_q6013_class_3_alert_handling.py
Compliance
- ECSS standards are freely downloadable (ESA); cite the source and paraphrase per standards-map.yaml.
- compliance: STANDARDS-REF, gated: false.
