Out-of-band fees are payments made directly to a specific miner (or group of miners) in exchange for confirming one or more transactions. They can be contrasted with standard in-band fees that are paid using the fee implied by the difference in a transaction’s input and output value.
For example, Alice broadcasts a transaction at a feerate that is low relative to other transactions in typical miner mempools. Alice wants to increase its feerate but is unable to use either RBF or CPFP fee bumping. Instead, she contacts a miner directly and pays them to include the transaction in their candidate blocks, which will eventually lead to confirmation (unless the miner gives up). Alice’s payment can be completely independent of her transaction; she may even pay using a non-bitcoin form of currency.
Consistent use of out-of-band fees weakens Bitcoin’s censorship
resistance. Miners controlling a large amount of hash rate produce
blocks more consistently than smaller miners, meaning someone such as
Alice who wants a transaction confirmed quickly will put more effort
into paying large miners than paying small miners. For example, if
Alice pays miners controlling 55% of hashrate to include her transaction
in their block candidates, there’s a 99% chance that her transaction
will be confirmed within 6 blocks:
1 - (1 - 0.55)**6
The advantage to Alice of paying small miners out of band is minuscule, likely meaning they will not receive the same opportunity to earn fees as large miners. If large miners earn a significantly higher percentage of profit than small miners for a long period of time, we would expect large miners to control a majority of total network hash rate. The fewer entities that control a majority of hash rate, the fewer entities there are that need to be compromised to censor which transactions get included in blocks.
Optech newsletter and website mentions
- Frequent use of exogenous fees may risk mining decentralization due to out-of-band fees
- Discussion about the effect of out-of-band fees on proposed fee-dependent timelocks