[jsmith45]

My understanding was that historically double entry accounting very much had each entry stored separately (although tied together by some transaction identifier). You would normally add both at the same time, so one could consider it "one record" on that basis, but physically they were separate entries often in separate physical books.

In computerized systems, having a single record in one place, and virtually constructing the ledgers from it is feasible. Under the constraints that all transactions debit from just one ledger, and credit to just one ledger, storing each transaction as tuple (id, amount, debited ledger, credited ledger) is feasible.

Double entry does have the slight advantage of being possible to generalize to allow one transaction to debit from multiple ledgers or credit to multiple ledgers, as long as all the debited amounts sum to the same as all the credited amounts.

It has a disadvantage in that the storage format does not guarantee validity, unlike the tuple method.

[ElevenLathe]

You are right that there would traditionally be separate books. The first, a "day book" would be a list of transactions that a merchant (or his servant/employee) would write down as business took place. This would say something like "I sold 10 spools of yarn for 8 florins apiece to Mr. Lucio" (remember that this all started a system to keep track of the affairs of Florentine Renaissance merchants). Later, after business had concluded for the day, the merchant (or his bookkeeper) would "post" the transaction to the ledger, which was a book that had a page for each "account." He would go through the daybook and, for each transaction, post it to (at least) two accounts, debiting and crediting equal amounts so that the books "balanced." This left him with a single number for each account which represented the total net activity for that account. This is the account's "balance" and is a very interesting piece of information for a merchant, so much so that it warrants all the associated ceremony.

Today, the "posting" can usually be done by a computer. There is some human creativity/regulation involved in deciding which specific accounts to credit/debit, but this can be noted in the "daybook entry" (more likely a database record somewhere than something literally written in a book) instead of being put off until the posting step.

The important characteristic is what the grandparent noted, that each transaction (in the daybook) tracks completely where each cent (or florin) came from and where it went. If a system of bookkeeping has this characteristic, I think it can be called double-entry, at least in spirit. A hard requirement that a transaction occur between exactly two accounts makes some kinds of events difficult (or impossible) to model, but is still conceptually the same.