Money as a record

Money began as a record of who owes whom, not as a thing; the tally stick is a split, tamper-evident ledger entry, and every later form of money refines that record.

Where you are. From the last lesson you can name the five parties behind a card tap and say which of them actually moved money. This lesson drops to the bottom of that stack and asks what the thing being moved is. The answer, that money is a record and always was, is the claim the whole course stands on, so it gets a burning building and seven hundred years of production history as evidence.

Two cartloads of firewood

On the morning of 16 October 1834, workmen at the Palace of Westminster were given a chore: dispose of two cartloads of old wooden sticks. The sticks were notched, some of them centuries old, and the office that had made them no longer existed. The workmen fed them into the two furnaces beneath the House of Lords and kept feeding them all day. By evening the overheated flues had set the chamber above alight; by the next morning most of the Palace was gone. Westminster Hall was saved. Almost nothing else was.

The sticks were not scrap. Each one was a notched willow rod recording a debt: who owed the Crown, how much, and from when. Together they were six centuries of England’s money records, kept by the Exchequer, the office that collected and audited the Crown’s revenue. The system had been retired eight years earlier, and the sticks had become what dead records become: a storage problem. Britain disposed of the master financial records of the medieval state as furnace fuel, and lost its parliament building doing it.

Read that as an engineer. The Exchequer ran a two-party datastore with an integrity check, in willow, for seven hundred years. Each stick was one entry; the basement was the database. What burned in 1834 was an archive whose entries were long closed, the obligations paid or carried into newer books, which is why the fire, for everything it destroyed, made nobody richer or poorer.

The idea in one paragraph

Money began as a record of who owes whom, and it never stopped being one. The tally was not a token standing for value held in a vault somewhere; it was the debt itself, written in the most tamper-resistant medium the 12th century had. Coins, banknotes and the balance in your banking app are later technologies for the same job: keeping the record of obligations, and letting people edit it in controlled ways. When you are paid, no substance flows to you; a record changes. That is not a simplification of what payment is. It is the whole of it, and the rest of this course is the machinery that keeps such records honest, fast and in agreement with each other.

The tamper check

Look closer at the stick, because its two design moves are the ones every ledger since has had to make in some form.

First move: the record is carried twice. The names and the date sit in ink on the flat sides, readable by anyone; the amount is also cut into the wood as notches. The writing is the record; the notches are a check on the part most worth forging.

Second move: the split. Once the stick is split down its length, through every notch, neither side can improve its own half. Add a notch to yours and the other half lacks it; shave one off and the other half still shows it; carve a fresh stick and the grain betrays it. Verification is a merge: bring the halves together and confirm that grain and notches continue across the cut.

Here is the whole design in one drawing.

A tally stick split lengthwise: the stock kept by the creditor above, the foil kept by the debtor below, one notch pattern cut across both halves so that only the true pair lines up

Wider than the screen; scroll it sideways.

Check yourself

1. After the split, neither the Crown nor the debtor could quietly improve their half. What exactly makes a one-sided edit detectable?

Verification compares the halves against each other, not either half against its holder’s claim. A notch added on one side has no counterpart across the cut; a shaved notch leaves an orphan on the other half; a freshly carved stick fails on grain. Each party can edit what it holds, but it cannot edit what the other party holds, and only the pair verifies.

2. Why split one stick rather than give each party a full copy of the record?

Two full copies can drift into two internally consistent records that disagree, and nothing in either copy says which is true. Two halves of one stick are useless alone and verifiable only together, because the grain is a pattern neither party can re-create. The split is a commitment, not a backup: the point is not surviving loss but preventing unilateral edits.

3. The 1834 fire destroyed six centuries of tallies and changed nobody’s wealth. Why not?

The sticks were retired evidence of entries mostly long closed; the obligations they once secured had been paid or carried into newer books. A live debt would have survived its stick anyway, in the other half, in memory and in law. Tokens are evidence of the record, the record is the money, and destroying evidence of a closed entry destroys nothing.

4. This morning your banking app showed your salary had “arrived”. Translate that sentence into this lesson’s vocabulary.

Nothing arrived. Ledgers were edited: your employer’s bank wrote one balance down, your bank wrote yours up, and the two banks square up the difference between themselves afterwards. “Arrival” is what a coordinated multi-ledger edit feels like from the outside. Most of this course is about how those edits are coordinated, and what happens when they half-complete.

5. In the exercise, verify recomputes the SHA-256 digest from the text instead of trusting the stored notches. Why must a check be recomputed rather than compared?

Two stored values agreeing proves only that they were written down together; a forged text travels happily beside a stale checksum. Integrity means recomputing the check from the data actually being claimed and comparing that against what was cut at split time. The wooden version worked the same way: you re-matched notch against notch across the grain in front of you, and took nobody’s word that the notches used to match.

Do this

Implement the tamper check. Open code/tally.py: split_tally already cuts one record into a stock and a foil, each carrying the written text plus a sixteen-character slice of the record’s SHA-256 digest as its notch pattern. Your job is verify, marked # TODO(you): both texts must agree, and each half’s notch must match its slice of a digest recomputed from that text. Return True or False; never raise.

From module-00-orientation/, run:

python code/tally.py

The file is its own test bench. It verifies an untouched pair (nine shillings, due at Michaelmas, the September quarter-day), then attempts the two frauds the Exchequer worried about, a doctored amount and a re-cut notch, and asserts your check catches both. Until you fill in the TODO it dies with NotImplementedError; when your check is right, the last line printed is:

tally verified; both tampering attempts caught

A finished verify is in solutions/tally.py; read it after yours passes, not before.

What you can now do. You can explain why a debt record with a tamper check is already money: the tally was not a claim on money kept elsewhere, it was the ledger entry itself, and you have implemented the check that made it trustworthy. You can also read the fire of 1834 correctly, as Britain losing a building and not a penny, because the token and the record are different things. The next lesson takes the step the tally never took: what happens when the records themselves start changing hands, and a piece of paper pointing at a ledger becomes the thing people pay each other with.