This week’s newsletter requests help testing release candidates for Bitcoin Core and LND, tracks continued discussion about the proposed noinput and anyprevout sighash flags, and describes several notable changes to popular Bitcoin infrastructure projects.
● Help test Bitcoin Core 0.19.0rc1: production users of Bitcoin Core are especially encouraged to test this latest release candidate to ensure that it fulfills all of your organization’s needs. Experienced users who plan to test are also asked to take a few moments to test the GUI and look for problems that might disproportionately affect less-experienced users who don’t normally participate in RC testing.
● Help test LND 0.8.0-beta-rc2: experienced users of LND are encouraged to help test the next release. For the first time ever, this testing can include creating a reproducible build of LND and verifying that it has the same hash as the binaries distributed by the LND developers.
● Continued discussion about noinput/anyprevout: this proposed sighash flag that would allow LN implementations to use eltoo was discussed again on the Bitcoin-dev and Lightning-dev mailing lists. After summarizing previous discussions, Christian Decker asked several questions: is the idea behind the proposal useful? (Respondents seemed to agree that it was.) Do people want mandatory chaperon signatures? (Respondents seemed moderately opposed.) Do people want mandatory output tagging? (Respondents seemed opposed, some strongly.)
In response to the question about output tagging, C-Lightning contributor ZmnSCPxj proposed an alternative tagging mechanism that would put the tag inside the taproot commitment, making it invisible unless a script-path spend was used. This could allow a spender who was worried about noinput to ensure they didn’t pay noinput-compatible scripts—the original goal behind output tagging—but without the decrease in privacy and fungibility created by output tagging. Several people seemed to express interest in this idea, although it wasn’t clear whether they wanted to see it as part of a proposal or they just preferred it to external output tagging (which, as noted above, was generally opposed by respondents).
The entire thread is more than 20 messages at present and started a spin-off discussion about
OP_CAT. Hopefully the discussion will be able to settle the major unresolved issues related to noinput and help get this proposal on track for inclusion in a subsequent soft fork.
Notable code and documentation changes
● Bitcoin Core #13716 adds
bitcoin-clithat allow it to read either an RPC or wallet passphrase from the standard input buffer rather than as a CLI parameter that would be stored in shell history. Echoing is also disabled on stdin during reading so that the passphrase isn’t visible to anyone watching your screen.
● Bitcoin Core #16884 switches the default address type for users of the RPC interface (including via
bitcoin-cli) from P2SH-wrapped P2WPKH to native segwit (bech32) P2WPKH. This change is on the master development code branch and is not expected to be released until Bitcoin Core 0.20.0 sometime in mid-2020. However, a previous change expected to be released as part of 0.19.0 in the next month or so will switch the default address type for GUI users to also use bech32 P2WPKH.
● Bitcoin Core #16507 fixes a rounding issue where a node would accept transactions into its mempool if they had a feerate greater than the node’s dynamic minimum feerate but wouldn’t relay those transactions to peers if the transactions’ feerates were less than minimum rounded up to next 0.00001000 BTC.
● LND #3545 adds code and documentation that allows users to create reproducible builds of LND. This should allow anyone with moderate technical skills to build identical binaries to those released by Lightning Labs, ensuring that users are running the peer-reviewed code from the LND repository and its dependencies.
● LND #3365 adds support for using
option_static_remotekeycommitment outputs as described later in this section. This new commitment protocol is particularly useful when something has gone wrong and you’ve lost data. If that happens, you need only wait for your channel counterparty to close the channel by paying a key directly derived from your HD wallet. Because the key was generated without any additional data (“tweaking”), your wallet doesn’t need any extra data in order to find and spend your funds. This is a simplified alternative to the data loss protection protocol that LND previously used and continues to understand.
● C-Lightning #2803 adds a new python package named
pylnthat includes a partial implementation of the LN specification. As described in its documentation, “This package implements some of the Lightning Network protocol in pure python. It is intended for protocol testing and some minor tooling only. It is not deemed secure enough to handle any amount of real funds (you have been warned!).”
● BOLTs #676 amends BOLT2 to specify that a node should not send the
funding_lockedmessage until it has validated the funding transaction. This warns future implementers about the problem that lead to the vulnerabilities described in last week’s newsletter.
● BOLTs #642 allows two peers opening a channel to negotiate an
option_static_remotekeyflag. If both peers set this flag, any commitment transactions they create which they’re able to spend unilaterally (e.g. to force close the channel) must pay their peer’s funds to a static address negotiated during the initial channel open. For example, if Alice has the address
bc1ally, Bob has the address
bc1bob, and they both request
option_static_remotekey, any commitment transactions that Alice can publish onchain must pay
bc1boband any commitment transactions that Bob can publish onchain must pay
bc1ally. If at least one of them doesn’t set this flag, they’ll fall back to the older protocol of using a different payout address for each commitment transaction, with the addresses created by combining the remote peer’s pubkey with a commitment identifier.
Always paying the same address allows that address to be a normal derivable address in the client’s HD wallet, making it possible for the user to recover their funds even if they’ve lost all of their state besides their HD seed. This is believed to be superior to the data loss protection protocol which depends on storing enough state to be able to at least contact the remote peer and identify the channel. With
option_static_remotekey, it can be assumed that the remote peer will eventually get tired of waiting for a missing peer to show up and will unilaterally close the channel, putting the funds onchain in an address where your HD wallet will find them.