Abstract
A method is disclosed for keeping a multi-source monitor truthful when some of its inputs cannot be read. Assertions over non-partitioned inputs sit in a dependency table and declare the input sources they need; assertions over a partitioned input run only over the partitions that were read; a capacity check always runs and reports an unusable reading itself. Each cycle derives the unread sources from recorded collection errors, runs every table entry whose declared inputs were all read, and withholds only the rest, so one lost input neither suppresses unrelated alerts nor downgrades a live critical finding. Every finding is stamped, when it is produced, with the source keys it was derived from; a partition's findings carry the same key a failed read of that partition produces. An alert dispatcher stores the stamp with each raised finding and declares an absent finding recovered only if every source in its stamp was read in that cycle. Otherwise, or when the whole cycle failed, the finding is held with its alert state unchanged, listed as unverified, and audited as not recovered. Because an absent finding cannot be inspected, stamping at production time is what separates "the breach is gone" from "the monitor could not look". The monitor also reports each unread source as its own degraded finding, reports each backup-controller list's counts as null rather than zero when that list was not read, treats an internally impossible reading as unread, and writes a clean-cycle record only when nothing is held.
Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Assuncao, gustavo matthew, "Source-Dependency-Stamped Monitor Findings with Surgical Assertion Withholding and Unverified-Never-Recovered Alert Reconciliation", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/13197