Bitcoin’s 2017 Fork Summer: Three Deadlines, Two Chains, No Final Judge

One megabyte became a proxy for two scaling strategies

By 2015, Bitcoin’s maximum block size was no longer an obscure software parameter. As blocks filled more often, users competed for limited space and proposals diverged over where future transactions should go. BIP 101, assigned to Gavin Andresen in June 2015, proposed replacing the fixed one-megabyte maximum with an 8 MB limit that doubled every two years. It documented a clear larger-block path, but its later closure also showed that publishing code did not make the network adopt it.^[1]

Segregated Witness offered a different package. BIP 141 moved signatures and scripts into a separate witness structure, fixed transaction malleability for non-intentional cases, and measured blocks by weight rather than simply raising the old byte limit. That created more effective capacity and made later payment layers easier to build. It was a soft fork: older nodes could continue to follow the chain, although they would not validate the new witness rules themselves.^[2]

The disagreement was therefore broader than “large” against “small.” Participants weighed immediate base-layer room, the cost of validating larger blocks, backward compatibility, miner coordination and the prospect of off-chain payments differently. Developers could propose rules, miners could signal and order transactions, businesses could choose software and tickers, and users could decide which ledger they treated as valuable. None of those groups had a constitutional switch that bound all the others.^[1][2][3][5]

The title of Bitcoin’s white paper supplied common language—peer-to-peer electronic cash—but not a single engineering schedule. By early 2017, the unresolved question was becoming operational: which software would nodes run when the advertised deadlines arrived?^[13][3]

Three coordination plans crowded into the same calendar

BIP 148, published in March 2017 under the name Shaolinfry, set a flag day. From August 1, participating nodes would reject blocks that failed to signal for the existing SegWit deployment. This was not SegWit activation itself; it was a rule intended to make signalling mandatory and push the existing BIP 9 deployment toward lock-in. If miners and economic nodes chose incompatible paths, a chain split was possible.^[3]

“While this BIP is active, all blocks must set the nVersion header top 3 bits to 001 together with bit field (1<<1). Blocks that do not signal as required will be rejected.”^[3]

— BIP 148 specification

On May 23, Digital Currency Group published the New York Agreement. The page’s figures, stated as of May 25, described 58 companies in 22 countries and 83.28% of hashing power. The plan paired an 80% threshold for SegWit with a 2 MB hard fork within six months. Those figures were the coalition’s own account, not a binding census of nodes or users; the listed supporters included Gavin Andresen.^[5]

“We agree to immediately support the following parallel upgrades to the bitcoin protocol”^[5]

— New York Agreement

The agreement then listed SegWit and the later 2 MB hard fork. Combining both approaches made a compromise possible for its signatories. The political limit remained: companies could commit their infrastructure, but their signatures could not compel every independent node to accept incompatible blocks.^[5]

BIP 91, BIP 148 and BCH did not do the same job

James Hilliard’s BIP 91 lowered the miner-coordination threshold to 80% within its own shorter window. BIP 91 locked in on July 21, 2017 and began enforcement on July 23: participating miners rejected blocks without the bit 1 signal for SegWit. That was not the lock-in of BIP 141 itself. The original SegWit deployment reached lock-in at block 479,808 on August 9, after BIP 148’s August 1 deadline, and activated at block 481,824 on August 24. BIP 91 coordinated signalling before the deadline; SegWit’s own two milestones followed later.^[4][3][12]

Bitcoin Cash followed an incompatible route. Bitmain had publicly described a UAHF as a contingency against BIP 148 and said it would contribute hash power under specified conditions. The final technical specification set 12:20 UTC on August 1 as the default activation time, required the first fork block to exceed 1 MB and added replay-protected signatures. Those were BCH rules, distinct from BIP 148’s demand for SegWit signalling on Bitcoin.^[6][7]

The dates are close enough to invite a false sentence: “UASF activated SegWit and triggered BCH on August 1.” The ledger says otherwise. BCH’s first exclusive block was 478,559, with a header time of 17:59:21 UTC on August 1. Bitcoin’s SegWit rules became active at block 481,824 on August 24. BIP 148’s date, the BCH fork and SegWit activation belong in the same story, but they are three events.^[7][8][11]

Nor did one event prove a permanent hierarchy among users, developers and miners. BIP 148 nodes could reject blocks locally; miners could coordinate through BIP 91; UAHF nodes could follow a different rule set regardless of relative hash rate; exchanges and wallets could decide how to label and support the resulting assets. The outcome emerged from these choices meeting in public software and markets.^[3][4][7]

November ended the agreement, not the argument

After SegWit activated, the New York coalition still intended to ship its 2 MB hard fork. Opposition persisted among developers, node operators and some businesses, while services had to prepare for replay, naming and custody risks if two ledgers survived. Bitcoin.org’s October warning was explicitly partisan—it called the proposed chain an alternative currency—but its operational warning was concrete: a service choosing one chain could expose customers to transactions or labels they had not chosen.^[10]

On November 8, coordinator Mike Belshe and other signatories suspended the fork. Their message did not concede that one social class had constitutional supremacy. It said the proposed upgrade lacked enough agreement to be clean, and that proceeding risked dividing the community.^[9]

“Unfortunately, it is clear that we have not built sufficient consensus for a clean blocksize upgrade at this time.”^[9]

— Mike Belshe and SegWit2x co-signers

The durable lesson is less triumphant and more useful. Consensus software lets each participant refuse rules, but refusal alone does not decide which asset markets, hash power or businesses will support. In 2017, BIP 91 helped Bitcoin activate SegWit, BCH preserved a larger-block path on a separate ledger, and SegWit2x failed to launch. The block-size dispute did not crown a final sovereign. It produced two live experiments whose users continue to judge different trade-offs.^[4][7][9]