Notice copies drifting

Think of it when the same fact is kept in more than one place, such as a price, a setting or a list of users.

In everyday life

Like a library’s stock check: staff walk the shelves, compare them with the catalogue, and list every book that is missing. Nothing rings when a book goes missing; only the check finds it.

One case

A shop keeps its prices in its database and sends a copy to a price comparison site every night. A sale lowers 30 prices, and that night the copy job fails without an error. For 6 days the comparison site shows the old, higher prices, until a customer complains on day 7.

The copy job failed, and nothing compared the copies
The case, cell by cell1 cell = 1 day
Sale prices in the database
30303030303030
Sale prices on the comparison site
00000030
Days the copies differed
6
Days a nightly check needs
1
On day 7 a customer complained, and the team ran the copy job again.

Every case

Copies drift. A job fails, someone edits a copy by hand, a cache keeps an old value. Nothing crashes, so no alarm fires, and the copies grow further apart each day. A drift check compares the copies on a schedule and names the one that differs. The fix always goes in the same order: correct the real copy, copy it again, check again.

The word

Engineers call this

drift detectionA regular check that compares copies of a fact that should match, and names the copy that differs. It catches driftTwo copies that should match have stopped matching. before a customer does.

Where a fact has copies, compare them on a schedule and name the one that differs. Fix the source, never the copy by hand.

Where this idea leads