Request Resolution

Working out which records a user's request actually names — before anything irreversible happens.

Why it matters in a memory layer

The user's word for a thing is not the store's word for it. Asked to "forget my old address", this course's store contains zero records matching address: it holds "Priya lives at 47 Halloway Road, Bristol". The obvious implementation returns nothing and the request looks already satisfied.

Resolution therefore runs on the labels assigned at write time, which is the strongest argument for keeping classification as a durable label rather than a decision made and discarded — its most important consumer is a request that had not been made yet.

And actionable is not unambiguous. Here exactly one record is labelled address, and the request says old while that record is the address the user gave as new — the old one was never stored. A literal reading deletes nothing; a helpful reading deletes where they live. One candidate makes deletion mechanically possible and says nothing about intent, so the two are reported separately.

Search generously. A wide candidate set gets reviewed; a narrow one gets acted on, and deleting the wrong record is unrecoverable.

Connections

Taught in: Deletion That Actually Deletes

Used in: Proving You Forgot

Do not confuse with: Cascade Deletion