Read the chain · A rule born from a discovered flaw

Two Transactions with One Name: the Rule Bitcoin Did Not Yet Have

A transaction identifier looks like a unique name. Early Bitcoin relied on that property, but did not yet require it from every valid block.

EDITORIAL OVERVIEW · Prepared by the project editors from the sources listed below.

How a duplicate became possible

An ordinary transaction refers to earlier outputs, which makes an exact repeat difficult. A coinbase transaction, used to pay a block's miner, has no ordinary inputs. If its data were repeated, its txid could be repeated too.

Two historical cases remain in Bitcoin's main chain. Blocks 91,842 and 91,880 contain coinbase transactions whose identifiers had appeared before. At the time, no explicit consensus rule prohibited them.

Why the duplicate mattered

The old software used a txid as the key for a record of spendable outputs. A newer transaction with the same identifier could overwrite the earlier state. During a chain reorganization, a node might then restore that state incorrectly.

Researchers showed that this behavior could weaken a confirmed transaction or split nodes over which chain was valid. A rare edge case had become a consensus problem.

The BIP 30 rule

In 2012, Pieter Wuille proposed BIP 30: a block may not contain a transaction whose txid matches an earlier transaction that still has unspent outputs. The two historical blocks were retained as exceptions because they were already part of the chain.

The repair did not rewrite history. It constrained future blocks and formalized an important principle: adding and then removing a block during a reorganization must not silently change the set of spendable coins.

If monetary state depends on a property, ‘almost impossible by accident’ is not a substitute for a consensus rule.

Sources and verification

  1. BIP 30: Duplicate transactions

Historical technical material. The word ‘name’ in the title is a plain-language description of the transaction identifier, or txid.

Unless stated otherwise, the text, conclusions, structure and editorial arrangement were created by the project editors. Facts, quotations and source materials remain attributable to their authors and rights holders.

Back to the rubric