The Block Size War: How Bitcoin Split Over One Megabyte (2015-2017)

The block size war was a three-year conflict over whether to raise Bitcoin's 1 MB block size limit through a hard fork. It ended in August 2017: SegWit activated on the main chain at block 481,824 on August 24, and large-block supporters split off on August 1 at block 478,558 to create Bitcoin Cash. The dispute settled who controls Bitcoin's rules — and the answer was economic nodes, not miners.

1. Where the 1 MB Limit Came From

• Event Background: Bitcoin launched with no explicit block size cap. In September 2010, with blocks nearly empty and the network worth almost nothing, Satoshi Nakamoto quietly added a 1 MB limit as an anti-spam measure, deployed with a flag day at block 79,400 on September 12, 2010. It was intended as a temporary guardrail, and Satoshi wrote in forum posts that it could be raised later by a phased-in change.
• Event Details: The limit was irrelevant until usage caught up with it. Through 2013 and 2014, average block size climbed steadily; by 2015 blocks were regularly hitting the cap during busy periods, transaction confirmation times became unpredictable, and a fee market appeared for the first time. What had been a theoretical parameter became a binding constraint on capacity of roughly three to seven transactions per second.
• Technical Architecture Innovation:
  • A consensus rule, not a setting: Because every node enforces the limit, changing it means every node must change simultaneously. A larger limit accepts blocks that old nodes reject, which makes it a hard fork by definition — and hard forks in a system with no central authority risk splitting the chain.
  • The trade-off nobody could measure: Bigger blocks mean more throughput and cheaper transactions, but also higher costs to run a full node, faster chain growth, and slower propagation. How much decentralization a given block size costs was, and remains, genuinely hard to quantify.
• Direct Impact: By mid-2015 the technical question had become a political one, because the answer determined what Bitcoin was for: a payment network for everyday transactions or a settlement layer optimized for verifiability.
• Long-term Significance:
  • The first real governance crisis: Bitcoin had never faced a decision where reasonable people disagreed and no mechanism existed to resolve it.
  • Two visions, both defensible: The "peer-to-peer electronic cash" reading of the whitepaper and the "digital gold with a settlement layer" reading both have textual support. The war was, at bottom, an argument about which one the network would serve.
• Lessons Learned: Temporary parameters in a consensus system become permanent political objects the moment they start to bind.
• Subsequent Development: Every later blockchain has treated capacity limits as a first-order design question, and most launched with explicit governance procedures for changing them, specifically to avoid what happened next.

2. Failed Proposals and Broken Agreements, 2015-2016

• Event Background: Between 2015 and 2016 the large-block side made three attempts to raise the limit by deploying alternative software, and the two sides negotiated two compromise agreements. All five efforts failed.
• Event Details: On August 15, 2015, Gavin Andresen and Mike Hearn released Bitcoin XT implementing BIP101, which would have raised the cap to 8 MB and doubled it every two years, activating at 75% miner support. It peaked around a thousand nodes and never came close to the threshold. Hearn declared Bitcoin a failed experiment in January 2016 and left. Bitcoin Classic, proposing a simple 2 MB increase, launched in February 2016, as did the Hong Kong Roundtable Agreement, in which several Core contributors and major miners agreed to pair SegWit with a 2 MB hard fork. Bitcoin Unlimited followed in October 2016, letting miners signal their own preferred size. On May 23, 2017, at the Consensus conference in New York, 58 companies representing a large majority of hash rate signed the New York Agreement, committing to SegWit followed by a 2 MB hard fork three months later.
• Technical Architecture Innovation:
  • Why miner-activated thresholds failed: All these proposals assumed that miner hash rate constituted a vote. Miners produce blocks, but nodes decide which blocks are valid. A proposal supported by miners and rejected by economically significant nodes produces a chain miners cannot sell into.
• Direct Impact: Nothing changed on the main chain for two years while transaction fees rose and the conflict escalated across forums, conferences, and social media, with heavy moderation disputes and sustained denial-of-service attacks against nodes signaling for particular proposals.
• Long-term Significance:
  • Agreements without enforcement: The Hong Kong agreement was signed by individuals who could not bind the Core project, which has no membership. The New York Agreement was signed by companies that could not bind users. Both failed for the same structural reason.
  • Miners learned their limits: The war established empirically that hash rate does not confer the power to change consensus rules.
• Lessons Learned: In a system where anyone can run a validating node, agreements among large firms are proposals, not decisions.
• Subsequent Development: The negotiating history became the subject of Jonathan Bier's 2021 book "The Blocksize War," now the standard reference account of the period.

3. 2017: BIP148, BIP91, and Activation

• Event Background: SegWit had been deployed as BIP141 with signaling opening on November 18, 2016, requiring 95% of blocks in a retarget period to signal support. Miners aligned with the large-block position withheld signaling, and by spring 2017 support sat far below the threshold with no path forward.
• Event Details: A pseudonymous developer known as shaolinfry proposed BIP148, a user-activated soft fork: from August 1, 2017, participating nodes would reject any block that did not signal for SegWit. This inverted the usual dynamic by threatening to orphan non-signaling miners' blocks. As the flag day approached, James Hilliard's BIP91 offered a compromise path, forcing SegWit signaling at a lower 80% threshold; it locked in on July 21 at block 476,784 and activated on July 23 at block 477,120. With BIP91 enforced, SegWit reached its lock-in threshold on August 9, 2017 at block 479,808 and activated on August 24, 2017 at block 481,824.
• Technical Architecture Innovation:
  • How a user-activated soft fork works: Because a soft fork only tightens rules, nodes enforcing the stricter rule will reject blocks that violate it. If enough economically important nodes — exchanges, payment processors, custodians — enforce it, miners producing non-compliant blocks earn rewards they cannot sell. No miner vote is required.
  • The risk that was accepted: BIP148 could have caused a chain split on August 1 if a large minority of hash rate ignored it. Businesses published contingency plans, and several exchanges announced how they would label the resulting chains.
• Direct Impact: SegWit activated without a chain split on the main chain. The bitcoin price rose sharply in the weeks around activation, and the fee market cooled temporarily as capacity increased.
• Long-term Significance:
  • A new upgrade mechanism: The UASF precedent showed that node operators can activate a soft fork over miner objection. Taproot's 2021 activation used a negotiated variant of the same idea.
  • Miner signaling reinterpreted: After 2017, miner signaling has been understood as coordination rather than consent.
• Lessons Learned: The credible threat of activation without miners was more effective than three years of negotiation with them.
• Subsequent Development: The SegWit2x hard fork planned for block 494,784 in November 2017 was cancelled on November 8 by its organizers, who cited insufficient consensus. That cancellation ended the war.

4. What SegWit Actually Changed

• Event Background: SegWit is usually described as a capacity increase, which understates it. It was a restructuring of how transactions are stored and identified.
• Event Details: Segregated Witness moves signature data — the "witness" — out of the transaction body into a separate structure committed to in the block. The 1 MB byte limit was replaced by a 4 million weight unit limit, where base transaction data counts as 4 weight units per byte and witness data as 1. In practice this raised effective capacity to roughly 1.7 to 2 MB of typical transaction data per block, and up to 4 MB in extreme cases.
• Technical Architecture Innovation:
  • Fixing transaction malleability: Because the transaction ID is now computed without witness data, third parties can no longer alter a transaction's ID by tweaking its signature encoding. This closed the bug that Mt. Gox had blamed for its collapse and, far more importantly, made second-layer payment channels safe to build.
  • Enabling the Lightning Network: Payment channels depend on pre-signed transactions that reference an unconfirmed parent by its ID. Without malleability protection, that construction is unsafe. SegWit is the precondition for Lightning, which launched on mainnet in 2018.
  • Script versioning: SegWit introduced a versioning scheme for scripts, which is how Taproot was later deployed in November 2021 as a clean soft fork rather than a disruptive change.
• Direct Impact: Capacity roughly doubled for adopters, and the witness discount created a lasting fee incentive to use SegWit addresses. Adoption was gradual, passing half of all transactions in 2019 and rising further after Taproot.
• Long-term Significance:
  • Layered scaling became the strategy: Bitcoin's answer to throughput settled on keeping the base layer small and verifiable, with volume handled above it. Lightning, and later sidechains and rollup-style proposals, all follow from that decision.
  • Unexpected consequences: The witness discount later made it economical to embed arbitrary data in witness space, which is the mechanism behind the Ordinals inscriptions that filled blocks from 2023 onward — an outcome no one anticipated in 2017.
• Lessons Learned: A change designed as a compromise on one axis often turns out to matter most for something else entirely.
• Subsequent Development: The base block weight limit has not changed since 2017, and no serious proposal to raise it has gained traction on the main chain in the years since.

5. Who Won, and What People Get Wrong

• Event Background: Both sides claim vindication, and the popular retelling flattens a complicated outcome into a simple one.
• Event Details: Bitcoin Cash forked on August 1, 2017 at block 478,558, mined by ViaBTC at approximately 12:37 UTC, running Bitcoin ABC with an 8 MB limit, no SegWit, an adjusted difficulty algorithm to survive on minority hash rate, and replay protection via SIGHASH_FORKID. It was driven by Jihan Wu on the mining side, Roger Ver on promotion, and Amaury Séchet on implementation. Bitcoin Cash itself split again on November 15, 2018, producing Bitcoin SV.
• Technical Architecture Innovation:
  • SegWit was not a block size increase in disguise: It increased effective capacity, but the large-block side's core demand — a hard fork raising the base limit — was never adopted. Describing SegWit as a compromise that gave both sides something is generous to the outcome; the small-block position prevailed on the substance.
  • The fork was not a 51% attack or a hack: Bitcoin Cash was a voluntary, replay-protected chain split. Holders of BTC at the fork height received an equal balance on the new chain, which is how every subsequent contentious fork has been structured.
• Direct Impact: The main chain kept its rules and its network effect. Bitcoin Cash retained a functioning chain and a much smaller share of market value and hash rate.
• Long-term Significance:
  • A durable answer about control: The war demonstrated that Bitcoin's rules are set by the users and businesses who validate them, not by miners, developers, or company coalitions. That finding has shaped every governance discussion since.
• Lessons Learned: A system designed to be hard to change is also hard to fix, and the war showed both halves of that trade-off in the same event.
• Subsequent Development: Taproot's 2021 activation was deliberately conducted differently — a long, low-conflict review process with a negotiated activation method — precisely because everyone involved had lived through 2017.