Bitcoin Optech Newsletter #32
This week’s newsletter includes an announcement of the 2019 Chaincode Residency program, summarizes a few talks from the Stanford Blockchain Conference, and provides the usual list of notable code changes in popular Bitcoin infrastructure projects.
- Apply to the Chaincode Residency: Bitcoin Optech encourages any engineer who is interested in spending summer contributing to open source Bitcoin and Lightning projects to apply to the Chaincode Residency. Full details of the residency are in the News section below.
2019 Chaincode Residency: Chaincode Labs opened applications for its fourth residency program to be held in New York over summer 2019. The program combines a 3 week seminar and discussion series covering Bitcoin and Lightning protocol development with a two month period for working on an open source Bitcoin or Lightning project under the guidance of an established protocol developer. The list of confirmed speakers and mentors includes some of the most prolific contributors to Bitcoin and Lightning. The previous Chaincode Residency (focused on Lightning Network applications) was covered in Newsletter #21.
Chaincode is inviting developers who want to contribute to open source Bitcoin and Lightning protocol projects to apply to the residency. Applicants from all backgrounds are welcomed, and Chaincode will cover travel and accommodation costs and provide a stipend to support living expenses.
Notable talks from the Stanford Blockchain Conference
The third edition of this annual conference (formerly named BPASE) included more than two dozen presentations across three days. We found the following topics especially interesting from a Bitcoin perspective, although we encourage anyone who wants to learn more to look at the transcripts provided by Bryan Bishop or the videos provided by the organizers (day 1, day 2, day 3).
Accumulators for blockchains presented by Benedikt Bünz (transcript, video). Bitcoin full nodes maintain a ledger (called the UTXO set) that stores the ownership information for every currently-spendable grouping of bitcoins. Currently, that ledger contains over 50 million entries and uses about three gigabytes of disk space. This allows nodes who receive a transaction to ensure the bitcoins being spent exist in the UTXO set, retrieve the ownership information for those bitcoins, and verify that information against the signature and other witness data provided in the transaction.
But what if we also asked the spender to provide a copy of the ownership information in their transaction along with a cryptographic proof that the the information is actually in the UTXO set? In that case, we wouldn’t need to store the whole set, we’d only need to store a commitment to a set we knew was accurate and that could be referenced by accurate proofs. RSA accumulators are one idea (among several others) for how to create that commitment and its related proofs. Using an accumulator, the size of the UTXO commitment that nodes would store on disk would be tiny compared to the full state. Transactions would increase in size due to needing to provide ownership data and a proof that they were part of the current UTXO set, but the size increase would be modest (assuming current typical transactions, about 70 bytes of ownership information per input plus a proof that could be aggregated down to about 325 bytes per block).
Other considerations affect the suitability of accumulators to the task, including being relatively new to cryptography, requiring either a well-studied trusted setup or a more novel untrusted setup, and potentially making blocks take longer to verify than the current system given that accumulator verification is about 100x slower than alternative systems using merkle trees. In a new development compared to a previous version of this talk given at Scaling Bitcoin 2018, the speaker describes a potential major speedup to practical verification.
In summary, RSA accumulators remain an interesting area of investigation into how to reduce full node requirements for storing and quickly accessing the UTXO set. This may not be particularly important now, when the UTXO set is relatively small and fast, but it could make it easier to support initiatives that would change how people use the UTXO set in the future. For example:
If accumulator-based proofs can indeed be verified quickly, they could allow the UTXO set size to grow considerably (perhaps because of a block size increase) while still ensuring that miners can verify transaction inputs quickly enough to minimize the use of spy mining or the production of stale blocks.
Software that uses trusted UTXO sets with newly-started nodes to avoid the cost and delay of downloading and verifying the full block chain (an option some software already provides) could reduce those costs and delays even further by replacing the multi-gigabyte UTXO set with an accumulator a million times smaller.
It should be possible for eager experimenters to explore the use of accumulators in Bitcoin without changing any consensus rules, such as Tadge Dryja has been doing with his similar Utreexo system based on merkle trees.
Miniscript presented by Pieter Wuille (transcript, video, slides). Imagine you had a Bitcoin script that gave control over your bitcoins to your lawyer a year after you last moved them, in case he needed to distribute them to your heirs. That’s an easy script to write. But what if someone then asked you to join a 3-of-4 multisig contract where you’d be one of parties holding some funds. How hard would it be for you to insert your existing policy into their generic multisig contract and be sure you haven’t broken anything? That’s the question asked by this speaker, and his answer is the composable policy language miniscript.
Miniscript is a subset of the full Bitcoin Script language that focuses on just a few features such signatures, times, and hashes plus operators for combining them, such as and, or, or threshold. It’s compact, easy to read, and will compile down to the most efficient script implementing a given policy. From an existing script compatible with its operations, it will also reverse it back into a policy for easy review. By design, it uses a similar vocabulary to the output script descriptors Wuille has been implementing in Bitcoin Core and it can help the updater or finalizer in a BIP174 PSBT workflow figure out who needs to sign next or whether all data for finalizing the script has been received.
Looking at the problem posed in the introduction paragraph, we can define the example personal spending policy as either you providing a signature for your compressed pubkey,
pk(C), or your lawyer waiting for a year,
time(<seconds>), and then providing a signature for his own compressed pubkey. We combine these branches with an asymmetric “or,”
aor, to reduce the witness data required when following the first branch.
We can define the generic 3-of-4 multisig policy similarly using compressed pubkeys (
A functionally equivalent policy that allows more flexibility would use the threshold operation:
This allows you to just replace one of the public keys with your personal policy:
When compiled, the result is the following script:
[pk] CHECKSIG SWAP [pk] CHECKSIG ADD SWAP [pk] CHECKSIG ADD TOALTSTACK IF [pk] CHECKSIGVERIFY 0x8033e101 CHECKSEQUENCEVERIFY 0NOTEQUAL ELSE [pk] CHECKSIG ENDIF FROMALTSTACK ADD 3 EQUAL
A final benefit of miniscript is that it should allow a policy written today to be automatically translated into a structure that makes optimal use of MAST, taproot, or other possible Bitcoin protocol upgrades. That means, as the Bitcoin protocol advances, the user or developer who invested time into crafting a policy won’t have to redo their work in order to continue using it with newer technology.
Probabilistic Bitcoin soft forks (sporks) presented by Jeremy Rubin (transcript, video). By March 2017, almost all Bitcoin nodes were ready to begin enforcing the segwit soft fork but miners seemed unwilling to send the activation signal. This created confusion: do miners get to veto protocol upgrades? If they do, is segwit dead? If they don’t, what do users do to override them? Ultimately the situation was resolved, but it was a mess that many would prefer not to repeat.
The speaker identifies the root cause of the problem as miners being able to delay activation at no cost to themselves. He proposes a solution: use the randomness remaining in a block header hash to determine whether or not a block signals for activation. For example, we’d choose a target that only 1 header hash out of 26,000 would match. A block matching that target would be produced once every six months on average, although nobody would know exactly when (about 10% of the time, it’d be within 3 weeks; but another 10% of the time, it’d take more than a year).
Miners would have no control over whether or not their block signaled for activation, but they would have control over whether they published that block. A miner who refused to publish his own block if it signaled for activation would lose the income from that block but would successfully prevent the fork from activating at least until the next signaling block was produced (which could be tomorrow or could be two years later). This gives miners a real chance to hold back a change but only by sacrificing real income. The end of the talk suggests some variations on the method with different tradeoffs.
The idea needs to be analyzed for possible problems, but it does provide an interesting alternative to BIP9-style miner-activated soft forks (MASFs) and BIP8-style user-activated soft forks (UASFs).
At the conclusion of his talk, this speaker also announced that the next Scaling Bitcoin conference and EdgeDev++ training sessions will be later in 2019 in Tel Aviv, Israel.
Notable code changes
Notable code changes this week in Bitcoin Core, LND, C-Lightning, Eclair, and libsecp256k1.
Bitcoin Core #14929 allows peers that your node automatically disconnected for misbehavior (e.g. sending invalid data) to reconnect to your node if you have unused incoming connection slots. If your slots fill up, the misbehaving nodes are disconnected to make room for nodes without a history of problems (unless the misbehaving node helps your node in some other way, such as by connecting to a part of the Internet from which you don’t have many other peers). Previously, Bitcoin Core banned the IP addresses of misbehaving peers for a period of time (default of 1 day); this was easily circumvented by attackers with multiple IP addresses. This solution gives those peers a chance to be useful but provides priority to potentially more helpful peers. If you manually ban a peer, such as by using the
setbanRPC, connections from that peer will still be rejected.
Bitcoin Core #13926 adds a new
bitcoin-wallettool to the executables Bitcoin Core provides. Without using RPCs or any network access, this tool can currently create a new wallet file or display some basic information about an existing wallet, such as whether the wallet is encrypted, whether it uses an HD seed, how many transactions it contains, and how many address book entries it has. This helps people who have a wallet file that hasn’t been synced to the most recent chain tip; they can do a quick inspection on the wallet to see if it’s interesting before they perform the lengthy rescan necessary to import it. Developers plan to add more features to the tool in the future.
Bitcoin Core #15159 changes the
getrawtransactionRPC so that it will now only return transactions in the mempool by default. If you have enabled the optional transaction index (txindex), it will also return confirmed transactions as well. Prior to this change, even if you didn’t have the txindex enabled, it would return confirmed transactions if they hadn’t yet had all their outputs spent. This previous behavior confused users: the call would work on some confirmed transactions but not others, and sometimes transactions that previously worked would suddenly stop working. This change makes the RPC act more consistently (although, of course, mempools vary between nodes and change over time).
LND #2538 increases the default time between sending updates about what public nodes exist on the network from 30 seconds to 90 seconds. This slows down propagation on the network, which has grown hugely in size, in order to conserve bandwidth. Separately from this PR, LN protocol devs are preparing changes to the gossip protocol to be more bandwidth efficient, although a lower update frequency will still save bandwidth there as well. (See also C-Lightning #2297 for the fix this week to a bug some nodes were encountering because the volume of gossip they received was so large.)
LND #2554 deprecates the
IncorrectHtlcAmountonion error in favor of the
UnknownPaymentHasherror that now includes the amount of the failed payment. LND will still handle the old error code but it will no longer generate it.
LND #2500 disconnects any peers that don’t support the Data Loss Protection (DLP) protocol. This ensures that LND’s new backup format will be usable. See the notable commits section and footnote from last week’s newsletter for information about the new backup protocol and the existing DLP protocol.