How Memnox remembers

One word, four different things underneath. What an organization needs to remember is not one kind of record, and pretending it is produces a search box where an answer should be.

Each kind answers a different question, so each is stored in the shape that question needs.

Why the shapes matter

A question like what is our refund policy wants a verified fact. Why is it our policy wants the events behind it. Who is affected if we change it wants the graph. What happens if an agent ignores it wants decision memory, because that is the copy the gate consults.

A single store, whatever its shape, answers one of those well and the rest badly. The failure of the flat approach is familiar to anyone who has asked a search index a why question and been handed nine documents.

Nothing is overwritten

Organizations change their minds, and the old answer stays load bearing long after it stops being current. So a fact or a decision that changes is superseded, never edited. The chain keeps both questions answerable: what is our policy walks to the end of the chain, and why did we do it the other way in March stops at the link that was current in March.

This is also what makes the audit trail mean something. A verdict from last quarter is explained by last quarter's rules, which still exist, unmodified, exactly as the gate read them.

Memory ages, and says so

A decision whose evidence is old, whose owner has left, or which nothing has touched in a long time is surfaced for confirmation. A practice that has drifted from the written rule is reported. Neither is resolved automatically, because in both cases only a person knows whether the memory or the behaviour is the one that is wrong.

Ageing is a reason to ask, not a reason to act. Nothing decays into deletion.

What a reader is entitled to

Memory is filtered before retrieval, not after. Need to know decides which evidence a person or an agent is entitled to reach, then the search runs over what is left. An agent never out-reads the person it works for, and every answer carries a count of what was withheld, because an agent that does not know it was filtered reports a partial answer as a complete one.