This week’s newsletter describes two suggested improvements to LN static backups and links to a proposal for a new version of PSBTs. Also included are our regular sections with summaries of changes to services and client software, popular questions and answers from the Bitcoin StackExchange, new releases and release candidates, and notable changes to popular Bitcoin infrastructure projects.

Action items

None this week.


  • Proposed improvements to static LN backups: anyone receiving money in a payment channel, including those currently used by LN, needs to keep track of the channel’s latest state—all the details that make it possible to close the channel and receive the correct share of the channel’s funds onchain. Unfortunately, computers have a bad habit of losing data and regular periodic backups don’t help much when a channel’s state could have changed just milliseconds before a disk drive fails.

    LN has always provided some robustness against this type of problem: if your node is offline, your channel counterparty will eventually close the channel so that they can start spending their funds again. This will send your funds to the onchain part of your LN wallet, which you hopefully backed up using a normal BIP32 seed. This should be reasonably safe: LN’s regular penalty mechanism encourages your counterparty to close the channel in its latest state—if they use an old state, they could lose all their money from the channel.

    The downside of the above approach is that you have to wait for your counterparty to decide that you’re not coming back. This wait can be eliminated if you back up some static information about your channel (e.g. the ID of your peer), reconnect to the peer after you lose data, and request that the peer immediately close the channel. This does seem to indicate that you’ve lost data and so your peer could close the channel in an old state—but, if they try that and you still have your old data, you can penalize them.

    This week, Lloyd Fournier started two threads on the Lightning-Dev mailing list about possible improvements to the above mechanisms:

    • Fast recovery without backups: the static per-channel backups that allow fast recovery of funds require you to create a new backup each time you open a new channel. If you fail to make a backup, your only option is to wait until your channel counterparty decides to close the channel on their own. Fournier instead proposed a deterministic key derivation method that would allow a node to search through the list of public LN nodes, combine information about its private keys derived from its HD wallet with information about each node’s main public key, and determine whether or not it had a channel with that node. This backup strategy would only work for channels opened with public nodes, which are expected to be the most common type of channel for typical users.

    • Covert request for mutual close: the existing mechanism for closing a channel requires that your counterparty broadcast their unilateral commitment transaction. It would be better to use a mutual close transaction—this uses less space onchain, requires paying less fees, is not identifiable onchain as having belonged to an LN channel, and allows both parties to spend their funds immediately. However, mutual close transactions don’t contain any penalty mechanism, so if you request a channel be closed and your counterparty gives you an inaccurate mutual close transaction, there’s no way for you to penalize them. In the normal protocol, this isn’t an issue—you’d just broadcast the latest state, but if you’ve lost your state, then you have no remedy.

      Fournier proposed a solution using a cryptographic primitive called oblivious transfer that allows your counterparty to send you the mutual close transaction encrypted in a way that allows you to either use it (closing the channel) or prove that you can’t decrypt it (allowing them to safely continue accepting payments in the channel). If you use this procedure every time you reconnect, you don’t reveal to them that you lost any data until after they’ve provided you all the information you need to recover.

  • New PSBT version proposed: Andrew Chow, author of the BIP174 specification of Partially Signed Bitcoin Transactions (PSBTs), proposed a new version of PSBTs that will contain several backwards incompatible features, although it’ll be largely the same as the current version 0 PSBT.

Changes to services and client software

In this monthly feature, we highlight interesting updates to Bitcoin wallets and services.

  • Bitcoin Wallet Tracker adds descriptor support: Bitcoin Wallet Tracker’s 0.2.0 release adds support for tracking output script descriptors and introduces libbwt, a library for allowing Electrum-backed wallets to easily support a Bitcoin Core full node.

  • JoinMarket now defaults to native segwit addresses: While JoinMarket supported segwit since 0.5.1, version 0.8.0 now uses bech32 native segwit addresses by default for coinjoins.

  • Bisq adds segwit for trade transactions: Building on previous bech32 support for deposits and withdrawals, Bisq v1.5.0 adds segwit support within trade transactions as well as implementing fee optimizations.

  • PSBT Toolkit v0.1.2 released: PSBT Toolkit, software that “aims to give you a nice gui that gives you functionality for PSBT interactions”, released various improvements in its 0.1.2 version.

  • Sparrow adds Replace-By-Fee: Sparrow 0.9.8 adds Replace-By-Fee (RBF) functionality and support for HWI 1.2.1.

  • Ledger Live adds Bitcoin Core full node support: Ledger Live, using the open source Ledger SatStack application, can now connect to a Bitcoin full node for sending transactions and providing balances in a more private way, without using Ledger’s explorers.

Selected Q&A from Bitcoin StackExchange

Bitcoin StackExchange is one of the first places Optech contributors look for answers to their questions—or when we have a few spare moments to help curious or confused users. In this monthly feature, we highlight some of the top-voted questions and answers posted since our last update.

Releases and release candidates

New releases and release candidates for popular Bitcoin infrastructure projects. Please consider upgrading to new releases or helping to test release candidates.

  • Bitcoin Core 0.21.0rc3 is a release candidate for the next major version of this full node implementation and its associated wallet and other software. Note, the macOS version of the signed binary will not run due to a problem with the code signing tool. The unsigned version (which can still be verified with PGP) should run if opened using the right-click (or control-click) context menu. Developers are working on fixing this problem for future release candidates and the final relase.

  • LND 0.12.0-beta.rc1 is the first release candidate for the next major version of this LN node. It makes anchor outputs the default for commitment transactions and adds support for them in its watchtower implementation, reducing costs and increasing safety, and adds generic support for creating and signing PSBTs. Also included are several bug fixes.

  • Bitcoin Core 0.20.2rc1 and 0.19.2rc1 are expected to be available sometime after the publication of this newsletter. They contain several bug fixes, such as an improvement described in Newsletter #110 that will prevent them from redownloading future taproot transactions that they don’t understand.

Notable code and documentation changes

Notable changes this week in Bitcoin Core, C-Lightning, Eclair, LND, Rust-Lightning, libsecp256k1, Hardware Wallet Interface (HWI), Bitcoin Improvement Proposals (BIPs), and Lightning BOLTs.

  • Bitcoin Core #20564 makes two changes to the way that Bitcoin Core signals support for addrv2 messages (BIP155):

    • Protocol version: Bitcoin Core will only negotiate support for addrv2 messages with peers that signal they are using P2P version 70016 or higher. This restriction isn’t required by BIP155, but release testing has revealed that some other implementations will disconnect from Bitcoin Core if they receive any unknown message, including sendaddrv2. This change may be reverted in future versions of Bitcoin Core, so other implementations are advised to tolerate unknown P2P messages at any time in the connection.

    • Updated BIP: The sendaddrv2 message will now be sent between the version and verack message, as required by the latest version of BIP155. See BIPs #1043 below for more information about that change to the BIP.

      This PR was backported to the latest V0.21 release candidate in Bitcoin Core #20612.

  • Bitcoin Core #19776 updates the getpeerinfo RPC with two new fields. bip152_hb_to indicates that we asked the peer to relay new blocks to us by sending a BIP152 compact block without waiting to ask if we need that specific block, called High-Bandwidth (HB) mode. bip152_hb_from indicates that the peer asked us to be one of their high-bandwidth peers. By default, each node selects up to 3 high-bandwidth compact block peers. (Despite the name, high-bandwidth mode doesn’t use much bandwidth compared to legacy block relay; its optimizations for fast relay of new blocks just uses more bandwidth than BIP152’s low-bandwidth mode.)

  • Bitcoin Core #19858 adds a new method for finding high-quality peers with the goal of making eclipse attacks more difficult. It opens an outbound connection every five minutes on average to a new peer and syncs headers with it. If the peer tells the node about new blocks, the node disconnects an existing block-relay-only peer and gives that connection slot to the new peer; otherwise, the new peer is dropped. Provided the node knows the IP address of at least one other honest node, this peer rotation should raise the cost of sustaining a partitioning attack against it, as the node should always eventually find the valid chain with the most work. The increased rotation and security will slightly reduce the number of open listening sockets on the network, but it is expected to help bridge the network as a whole via more frequent connections, add edges to the network graph, and provide more security against partitioning attacks in general.

  • Bitcoin Core #18766 disables the ability to get fee estimates when using the blocksonly configuration option. This bandwidth-saving option stops Bitcoin Core from requesting and relaying unconfirmed transactions. However, Bitcoin Core’s fee estimates are also dependent on tracking how long it takes transactions to become confirmed. Previously, when blocksonly was enabled, Bitcoin Core stopped updating its estimates but continued to return the estimates it had already generated, producing increasingly out of date estimates. After this change, it won’t return any estimates at all when blocksonly mode is enabled.

  • C-Lightning #4255 is the first of a series of pull requests for an initial version of offers—the ability to request and receive invoices over the LN. A common use of this feature would be that a merchant generates a QR code, the customer scans the QR code, the customer’s LN node sends some of the details from the QR code (such as an order ID number) to the merchant’s node over LN, the merchant’s node returns an invoice (also over LN), the invoice is displayed to the user (who agrees to pay), and the payment is sent. Although this use case is already addressed today using BOLT11 invoices, the ability for the spending and receiving nodes to communicate directly before attempting payment provides much more flexibility. It also enables features that aren’t possible with BOLT11’s one-time-use hashlocks, such as recurring payments for subscriptions and donations. See the BOLT12 draft for more information.

  • Eclair #1566 abstracts out all connection-handling logic into a front node, which can be distributed across multiple hosts to achieve high availability for production deployments. These front nodes also handle CPU/bandwidth-intensive BOLT7 messages, such as routing table related gossip and syncing requests, in a distributed manner, improving scalability for larger node deployments like ACINQ’s. For readers looking to deploy this change, an AWS Beanstalk bundle is available, and the author recommends using the AWS Secrets Manager to store the node’s private key, a topic previously covered in the SuredBits field report.

  • Eclair #1610 allows overriding the default relay fees when opening a new channel using the new feeBaseMsat and feeProportionalMillionths options.

  • LND #4779 allows the node to claim payments (HTLCs) that weren’t yet settled at the time a channel using anchor outputs was closed.

  • BIPs #1043 changes the way that support for BIP155 is negotiated between peers. Previously, the BIP specified that a node should send a sendaddrv2 message to signal support for BIP155 after receiving a verack message from its peer. The BIP now specifies that the node must send the sendaddrv2 message earlier in connection establishment, between sending its version and verack messages. This is consistent with how BIP339 negotiates wtxid relay support and also with a generic method for negotiating features proposed to the mailing list earlier this year.

    John Newbery posted a summary of all the changes to BIP155 since it was proposed in February 2019 to the Bitcoin-dev mailing list.

  • BOLTs #803 updates BOLT5 with recommendations for preventing a transaction pinning attack. The recent anchor outputs update to the LN specification allows multiple payments (HTLCs) that were pending at the time a channel was unilaterally closed to all be settled in the same transaction. A potential problem is that some of those outputs may pay your channel counterparty, giving them the ability to pin the transaction and prevent the other HTLCs in the batch from confirming until after their timelocks expire. The recommendation is to allow batching the HTLCs for maximum efficiency when there’s plenty of time left but to split each HTLC into a separate transaction when the timelock expiration is approaching so that pinning isn’t a problem.


Newsletter #87 incorrectly claimed that “previous versions of Bitcoin Core would terminate a new connection if certain messages didn’t appear in a particular order”. We were alluding to a belief that the version message needed to be immediately followed by the verack message introduced in Bitcoin 0.2.9 (May 2010). Subsequent code review and testing of old versions of Bitcoin Core by Optech contributors did not substantiate this statement and we’ve added a correction to the original text. We apologize for the error.

Holiday publication schedule

Happy holidays! This issue is our final regular newsletter for the year. Next week we’ll publish our annual special year-in-review issue. We’ll return to regular publication on Wednesday, January 6th.