Segwit split signatures from transaction ids in 2017

Inherited archive entry · retained from the earlier edition. Its factual, legal and policy claims have not all been reverified for 19 September 2026. Research scope.

A signature format change with a hard date

Segregated Witness activated on the Bitcoin network on 24 August 2017, at block 481,824, according to the block record. The change had been specified nearly two years earlier: BIP141, created 21 December 2015 by Eric Lombrozo, Johnson Lau and Pieter Wuille, set a signalling window running from 15 November 2016 to 15 November 2017 during which miners could indicate readiness through a block version bit. Activation came roughly nine months into that window, after separate signalling arrangements, covered elsewhere in this archive, shifted enough hashpower to meet the threshold.

What moved, and what that fixed

BIP141's core change is structural: it removes the witness field, the signature data proving the right to spend an input, from the part of the transaction used to calculate the transaction's id, storing it instead in a separate witness structure. BIP143, created 3 January 2016, then defines a new signature hash algorithm for spending these outputs, replacing an older method whose verification cost grew quadratically with transaction size, and adding the spent amount directly into what gets signed, which lets an offline signer verify a payment without holding the entire previous transaction. A developer post dated 26 January 2016 lists the resulting benefits: transactions can no longer be altered in transit without invalidating their id, which fixes malleability; blocks can carry more data under a new weight-based limit of four million units; and full nodes can discard old witness data without weakening their ability to validate new blocks.

The boundary: capacity is not privacy, and deployment is not the code

The weight limit raises effective capacity to roughly two megabytes of typical transactions rather than a literal four, since witness data is discounted rather than free, as the January 2016 post itself notes. Fixing transaction malleability is a narrower claim than making transactions private; BIP141 does not alter what a block explorer can see about who paid whom. The nine months between the start of the signalling window and actual lock-in is itself worth separating from the code: a soft fork can be fully specified and backward-compatible and still take substantially longer to deploy than to write, because deployment depends on independent parties choosing to upgrade.

Questions to carry forward

• Does a stated 'up to 4MB' capacity figure account for the discounted weight of witness data, or does it overstate typical throughput?
• What made miner signalling accelerate in mid-2017 rather than earlier in the sixteen-month window BIP141 allowed?
• Which applications built after 2017 depend specifically on the malleability fix, and would they function without it?

Segregated Witness is a useful case for separating a protocol's specification date from its activation date, and separating what a soft fork changes about transaction structure from what it does not change about who can see a payment.