The consensus and peer-to-peer changes behind segwit touch 1,486 lines of code, with 3,338 more lines of tests. Testing ran from Pieter Wuille's Elements sidechain version in 2015 through segnet to Bitcoin's regular testnet in May 2016. Schnorr signatures and MAST are named as later upgrades segwit would make easier.
Miners have until block 419328 to act: the CSV soft fork locked in at block 417312 after 1946 of 2016 blocks (96.53%) signaled for BIP68, BIP112 and BIP113. All mining and backup nodes must run Bitcoin Core or Knots 0.12.1, and anyone hardcoding the block version has to unset bit 0.
BIP9 lets miners signal readiness for up to 29 concurrent soft forks through bits in the block header. A proposal locks in when 95% of blocks in a 2016-block window signal, then activates one retarget later, or fails at its timeout. Unlike IsSuperMajority, it won't orphan a non-signaling block that obeys the new rules.
Peers using compact block relay (BIP152) exchange a sketch of each new block, about 9KB for an average 1MB block, plus overhead for transactions the receiver lacks. In live high-bandwidth experiments with up to 6 prefilled transactions, well over 90% of blocks propagated with no follow-up request.
Signaling for the CSV soft fork starts at midnight on May 1st, 2016 with Bitcoin Core 0.12.1, the first deployment through BIP9 version bits. It covers BIP68, BIP112 and BIP113; OP_CHECKSEQUENCEVERIFY lets coins be frozen for a set time after confirmation, which Lightning Network will depend on. The p2p alert system now defaults to off.
Marco Falke is Bitcoin Core's new Testing & QA maintainer, Wladimir van der Laan announced. He takes charge of merging test framework pull requests, a natural step after a year spent on unit tests, coverage and reviewing other contributors' work.
Security issues and new releases now have their own Bitcoin Core announcement list, opt-in and announcement-only. Users had asked for a way to receive only the critical notices, and the project expects very low volume.
A 16x16 Sudoku solution sold for 0.10 BTC in the first successful Zero-Knowledge Contingent Payment on Bitcoin, demonstrated live at Financial Cryptography 2016 in Barbados. Sean Bowe of the Zcash team was the seller and did almost all the engineering. Under the protocol, the data changes hands only if payment is made, with no third party to arbitrate.
OpenSSL is out of signature checks in Bitcoin Core 0.12.0, replaced by libsecp256k1, which cuts reindexing and new block validation to under half their old time. Mempools get a 300 MB default cap, closing a way attackers could crash low-RAM nodes. Also new: opt-in Replace-by-Fee (BIP 125) and -maxuploadtarget for limiting upload traffic.
Bitcoin Core says the site that first hosted the Bitcoin paper is not affiliated with the project, which has its own official website. Development runs on IRC, GitHub and the bitcoin-dev mailing list. The project takes no responsibility for other forums and asks community members not to brigade or launch denial-of-service attacks.
Third-party malleability ends for transactions whose inputs are all segwit spends, because signatures move to a witness that no longer feeds the txid. Wallet authors and the Lightning Network both gain from that. Signature hashing also turns linear, with each byte hashed at most twice; a single block has taken 25 seconds to validate under the old scheme.
Developers can now test Segregated Witness on its own network by compiling the segwit branch and adding -segnet to bitcoind and bitcoin-cli. Ledger, Trezor, Electrum and Bitgo have said they will support segwit, while mSIGNA and Green Address offer early previews.
A video maps 2015 commit activity in the Bitcoin Core repository, a year with 2 major versions, 5 maintenance releases and two activated soft forks: BIP66 and BIP65, the latter adding CHECKLOCKTIMEVERIFY. Reviewers, testers and translators don't show up in commit data, the project points out.
Consensus rules are not for the Bitcoin Core team to decide, its contributors say; users decide by choosing which software to run, and that is why it ships without auto-update. They favour soft forks, reserving hard forks for rare cases with near-universal agreement, and see additional layers on top of the protocol as the route to extremely high scale.
Blocks under segwit's new cost limit can reach just under 4 MB, but Anthony Towns's calculations put a block of single-signature P2PKH transactions near 1.6 MB and one of 2-of-2 multisig near 2.0 MB. The FAQ's tentative schedule has segwit deploying in 0.13.1 in October 2016, with weak blocks, IBLT or Lightning following in 2017.
Signing on to the capacity roadmap, 57 people including Pieter Wuille, Gregory Maxwell and Wladimir J. van der Laan call it the best continuation of their scalability work within Bitcoin Core. The list stays open to further signatures.
Gregory Maxwell's roadmap, sent to the bitcoin-dev mailing list as Scaling Bitcoin Hong Kong wrapped up, builds on libsecp256k1, BIP9 versionbits and Pieter Wuille's Segregated Witness. He reports ConnectTip running 7x faster in 0.12, and describes segwit as a soft fork that takes signatures outside the current blocksize limit, amounting to a 4MB increase at worst.
Technical contributors including Wladimir J. van der Laan and Pieter Wuille signed a letter asking the community not to prejudge scaling proposals. They point instead to the consensus process and two open Scaling Bitcoin workshops: Montreal on September 12-13 and Hong Kong on December 6-7.