Replace-by-fee: opt-in, then full
Let an unconfirmed transaction be replaced by a version that pays a higher fee. BIP 125 (2015) made this opt-in: a transaction signals that it may be replaced. “Full-RBF” drops the signal, so any unconfirmed transaction can be replaced.
Full-RBF camp · Peter Todd and many Core contributors
Miners can always take the higher-fee version, so “first-seen” was never a real guarantee. Reliable replacement helps fee bumping and multi-party protocols, and accepting unconfirmed payments as final had long been discouraged.
Zero-conf services · e.g. Muun, and Mike Hearn in 2015
Businesses that accept unconfirmed payments relied on first-seen relay. Full-RBF turns a miner-only attack into one anyone can run from a wallet, so these services had to switch off instant payments.
It set a promise many merchants relied on (an unconfirmed payment usually sticks) against the incentive argument that miners will take the higher fee. When Bitcoin Core 24.0 shipped a mempoolfullrbf option in 2022, there were calls to pull it before release.
Opt-in RBF (BIP 125) was merged in November 2015 (PR #6871). Bitcoin Core 24.0 added mempoolfullrbf, off by default, and a pull request to disable it on mainnet (#26287) was closed. By March 2024 Peter Todd estimated 95%+ of hash power ran full-RBF, and Bitcoin Core 28.0 turned it on by default (mempoolfullrbf=1).
- 28 Mar 2015Mike Hearn argues against replace-by-fee “scorched earth”; Peter Todd asks for a list of companies that actually rely on first-seen. [1]
- Oct–Nov 2015Opt-in RBF (BIP 125) is proposed and merged into Bitcoin Core. [1] [2]
- 2022PR #25353 adds a -mempoolfullrbf node option; it ships in 24.0, off by default. [1] [2]
- 7 Oct 2022Muun’s Dario Sneidermanis warns zero-conf apps they must switch off zero-conf features. [1] [2]
- 12 Oct 2022A pull request to temporarily disable full-RBF on mainnet draws a NACK from Murch and is later closed. [1]
- 31 Oct 2022Peter Todd tests whether miners can be paid to run full-RBF. [1]
- 28 Mar 2024ViaBTC turns on full-RBF; Todd estimates 95%+ of hash power. [1]
- 2024Bitcoin Core 28.0 turns full-RBF on by default. [1]
“This article lays out the case against RBF-SE and argues it is harmful to Bitcoin.”
bitcoin-dev (quoted in the reply) · 28 Mar 2015 · source
“Would you so us all a favor and make a list of companies actually relying on "first-seen" mempool behaviour.”
bitcoin-dev · 28 Mar 2015 · source
“But this is the kind of proposal only a game theorist could really love.”
bitcoincore.org · 2015 · source
“we realized that zero-conf apps (like Muun) must immediately turn off their zero-conf features.”
bitcoin-dev · 7 Oct 2022 · source
“We've been talking about full-rbf for seven years. It is well-established that zero-conf subsists on nobody deciding to try to sidestep it, and it has been long communicated that anyone using zero-conf should be aware that it is a bad practice.”
GitHub, PR #26287 · 12 Oct 2022 · source
“Let's see if miners can be bribed to run full-RBF, starting small.”
X · 31 Oct 2022 · source
“.@ViaBTC finally turned on full-RBF. They were the last notable pool without full-RBF, which puts us at something like 95%+ hash power doing full-RBF.”
X · 28 Mar 2024 · source