This week’s newsletter includes a special section about the recent Taproot proposal, news about a small potential change to the BIP174 PSBT format, and our regular sections about bech32 sending support and notable changes in popular infrastructure projects.
None this week.
- Higher feerates: at the time of writing, estimated feerates for confirmation in the next 2 blocks are ten times higher than estimates for confirmation within 30 blocks (six hours) and a hundred times higher than estimates for waiting 144 blocks (1 day) or more. This may mean there’s still an opportunity to cheaply consolidate a modest number of inputs in case feerates rise even further, but only if you can tolerate a potentially long wait for the consolidation transaction to confirm.
Soft fork discussion: several replies were posted to the Bitcoin-Dev mailing list in response to bip-taproot and bip-tapscript. Additionally, Anthony Towns posted a proposal for an additional soft fork change that uses bip-taproot features to enable functionality similar in purpose to BIP118 SIGHASH_NOINPUT. As this week’s newsletter already includes a special section about Taproot, we’re deferring summaries of the feedback and extension to next week’s newsletter so readers can have some time to digest the essentials of Taproot first.
Addition of derivation paths to BIP174 PSBTs: Stepan Snigirev posted a suggestion to the Bitcoin-Dev mailing list to allow PSBTs to include the BIP32 extended public key (xpub) and derivation path for the public keys used to generate the change output’s address. This can help multisig wallets determine whether or not the transaction’s change output pays back to the correct set of signers. The author of BIP174, Andrew Chow, was receptive to the idea, as was the developer for a hardware wallet.
Overview of the Taproot & Tapscript proposed BIPs
Last week, Pieter Wuille emailed the Bitcoin-Dev mailing list with links to two proposed BIPs. The first, bip-taproot, permits spending using either a Schnorr-style signature or merklized script. The second, bip-tapscript, defines a slight variation on Bitcoin’s existing Script language to be used in bip-taproot merkle spends.
For readers already familiar with Bitcoin scripting and ideas like MAST, it should be possible to understand the BIPs directly. For readers with less background, we’ve prepared the following summary of some key features from the proposals by looking at them from the point of view of an existing wallet that wants to upgrade to use Taproot and Tapscript. This only skims the surface of what these proposals make available, as one reason developers have been waiting for Schnorr signatures and MAST-based encumbrances is that they provide the building blocks for new features that were previously difficult or impossible to build on Bitcoin. We’ll leave the description of those advances for another time and focus here on how the two proposed BIPs can make existing uses of Bitcoin work even better than they do today.
Please note, all Taproot features will be opt-in for wallets, so no existing wallet will need to change how it works. Only wallets that want to take advantage of the benefits of Taproot and Tapscript will need to upgrade.
What’s not in the proposals
Before we look at what features the proposals may add to Bitcoin, let’s take a moment to look at some things that aren’t part of the proposals:
No cross-input signature aggregation: just like current Bitcoin transactions, each input will need to contain all its required signatures. This means transactions such as consolidations and coinjoins won’t receive any special discount. If the Taproot and Tapscript proposals become part of Bitcoin, we think developers will continue to look for ways to bring cross-input signature aggregation to Bitcoin in future soft forks, but they’ll need to find ways to address complications that were discovered during their research into the advanced techniques described by bip-taproot. (See Newsletter #3 for related discussion in the context of bip-schnorr.)
No new sighash types: although some existing sighash types are tweaked a bit, the proposals do not offer anything similar to BIP118 SIGHASH_NOINPUT. However, bip-tapscript does provide a forward-compatibility mechanism (tagged public keys) that will make it easy for future soft forks to extend the signature-checking opcodes with new sighash types or other changes.
No activation mechanism specified: if users decide they want to begin enforcing the soft fork’s new rules, safety requires that enough of them begin enforcing the new rules at the same block so that miners are deterred from creating blocks that violate the new rules. Various mechanisms have been used to accomplish this in the past and other options have been described for potential future use. However, bip-taproot doesn’t mention any of these techniques. Optech agrees with the primary author of the BIP that activation discussion is premature. We first need to ensure that there’s widespread agreement that the proposals are safe and desirable before we start a debate about the best way to activate them.
Single-sig spending using Taproot
To look at what’s in the proposals, we’ll primarily examine how existing use cases could be transferred to Taproot. The best place to start is by looking at the way most wealth is transferred via the Bitcoin protocol right now: single-sig P2PKH and P2WPKH spends.
Single-sig P2WPKH wallets today currently generate a private key, derive a public key (pubkey) from it, and hash that pubkey to create the witness program for a bech32 address. (P2PKH is functionally identical but uses a different script and a different address encoding.)
|Private key||read 32 bytes from CSPRNG, or using BIP32 HD derivation||
|Public key||point(0x807d[…]0101), or using BIP32 HD public derivation||
|Address||bech32.encode(‘bc’, 0, 0x006e[…]05d6)||
Under Taproot, the hashing step will be skipped and so the bech32
address will contain the pubkey directly, with one small change.
Currently, 33-byte Bitcoin-style pubkeys are encoded to start with either
a 0x02 or 0x03 to allow validators to reconstruct the key’s Y-coordinate
on the secp256k1 elliptic curve; in bip-taproot, the value of this byte
is reduced by two so that 0x02 becomes 0x00 and 0x03 becomes 0x01. The
meaning stays the same but using the low bit for the values frees up the
remaining bits for future soft forks.
Also the witness version is changed from the
for P2WPKH/P2WSH to a
|Private key||(Same as above)||
|Public key||(Same as above)||
|Alter key prefix||(key - 2),key[1:33])||
|Address||bech32.encode(‘bc’, 1, 0x00e5[…]3c23)||
Here’s an example of existing addresses compared to an example taproot address.
Spending from P2PKH or P2WPKH requires the spender include their public key in their input. For Taproot, the public key was provided in the UTXO being spent, so several vbytes are saved by not including it again. The spend must also include a signature; this is a Schnorr-style signature as defined by bip-schnorr with an optional sighash byte appended. If the default sighash is used, the signature is 64 bytes (16 vbytes); if a non-default is used, it is 65 bytes (16.25 vbytes1). Overall, this makes the cost to create and spend a Taproot single-sig output about 5% more expensive than P2WPKH. This is probably not significant: the cost to create a Taproot output is almost the same as to create a P2WSH output—which people pay all the time without issue—and the cost to spend a single-key Taproot is 40% cheaper than P2WPKH.
Besides the change from ECDSA to Schnorr for the signature algorithm, there are a few important (but easy to implement) changes to the transaction digest—the hash that a signature commits to in order to prove the transaction is an authorized spend of a UTXO.
Most notably, the hash used goes from the legacy protocol’s
double-SHA256 (SHA256d) to a single SHA256 operation. The extra hashing
previously used isn’t believed to have provided any meaningful security.
Additionally, the data hashed is prefixed with a value that’s specific
to this use of the hash function; this helps prevent problems like
CVE-2012-2459 where a hash from one context can be used in another
context. The prefix tag is
SHA256(tag) || SHA256(tag) where the tag
in this case is the UTF-8 string “TapSighash” and
|| stands for
concatenation. Software that needs to create or verify a large number
of signatures (such as active LN nodes or Bitcoin full nodes) can use a
version of their SHA256 function that’s been pre-initialized with the
prefix tag so it doesn’t need to repeat those operations for every
signature verification. Implementations that don’t require maximal
performance (such as ordinary wallets) can just implement the algorithm
as described using their default SHA256 library function.
There are also some changes to what data is included in the hash and how it is serialized. This includes improvements that can help make wallets without access to verified UTXO entries (e.g. hardware wallets and HSMs) more efficient because they don’t need to retrieve as much data in order to ensure they’re signing for correct UTXOs and amounts.
Although that may sound like a significant number of changes, we suspect it’s only an afternoon worth of work making minor serialization changes for any wallet that’s already segwit compatible and which can gain access to a library like libsecp256k1 for generating bip-schnorr-compatible signatures.
Simple multisig spending using Taproot
After single-sig UTXOs, the most common are simple multisig UTXOs. These are outputs that depend on a certain number of signatures from particular pubkeys but don’t have any other conditions. These are used both by individual users (e.g. requiring multiple devices to work together in a spend) and by multiple parties to a single transaction (e.g. escrows).
There are two ways to perform simple multisig spending using Taproot, the most efficient of which is key and signature aggregation as described in this subsection. We’ll examine the other mechanism in the next subsection.
For aggregation, two or more private keys are created and their pubkeys are derived. The pubkeys are then combined into a single pubkey that’s indistinguishable from any other Bitcoin pubkey. This is used as the segwit witness program as described in the previous subsection. Later, the owner or owners of some or all of the private keys create signatures that are combined into a single signature that’s also indistinguishable from any other bip-schnorr signature. This must commit to the same transaction digest as described in the previous subsection, but the result is a multisig spend that’s completely indistinguishable from a single-sig spend.
You may have noticed that the preceding paragraph is vague about the exact mechanism used to aggregate the keys and signatures. The reason for that is because there are multiple known methods and any one of them can be used by the participants. It’s even possible for researchers to find new methods and implement them in Bitcoin wallets without any consensus changes. This is because the signature algorithm is only looking for a single pubkey and a single signature that are valid under rules described in the single-sig subsection above. That said, of the methods known, the MuSig protocol is probably the best studied aggregation protocol in the context of Bitcoin.
The number of bytes used for aggregated keys and signatures is exactly the same no matter how many signers are involved. This can be compared to P2WSH multisig where each additional pubkey adds 8.50 vbytes and each additional signature adds about 18.25 vbytes.
Complex spending with Taproot
Bitcoin’s Script language allows users to specify what conditions must be fulfilled in order for their bitcoins to be spent, conditions that sometimes require more than just signatures. Taproot not only supports this feature but enhances it in several ways.
We can consider this by looking at a basic Hash TimeLock Contract (HTLC) similar to those used in LN and cross-chain atomic swaps. This contract can terminate three ways:
- Alice receives the contracted money (a payment) by publishing a pre-image that releases the contract’s hashlock.
- Bob receives the contracted money (a refund) after a timelock has expired.
- Alice and Bob mutually agree about how to distribute the money, usually because one of them could use one of the preceding conditions to force a payment or refund.
To take the LN case of HTLCs as an example, we can produce two independent scripts, each of which handles one of the first two items above:
(1) OP_HASH256 <hash> OP_EQUALVERIFY <Alice pubkey> OP_CHECKSIG (2) <time> OP_CHECKLOCKTIMEVERIFY OP_DROP <Bob pubkey> OP_CHECKSIG
These separate scripts are then hashed so that they can be used as the leaves of a merkle tree. As described earlier, the data to be hashed is prefixed with a tag (itself hashed) indicating its purpose. “TapLeaf” is the tag in this case. The leaves must also include the size of the script and a version (only version 0xc0 is defined in this proposal; other versions are available for future soft fork upgrades).
After the two scripts are hashed, their digests are put in lexicographic order. This ordering will allow later verification of a merkle inclusion proof without knowing whether or not each leaf or branch appeared in the tree to the left or right of its pair sibling, thus reducing the amount of data that needs to be communicated as well as potentially improving privacy. After ordering, the two hashes will be hashed together with the prefix tag “TapBranch”. As this merkle tree only has two leaves, the resultant hash is the merkle root.
This merkle tree only covers two of the HTLC’s possible end states. For the third case where Alice and Bob mutually agree on a spend, they can use signature aggregation using something like MuSig. As in the multisig section above, they work together to create a single pubkey, called the Taproot internal key.
The merkle root and the internal key are then hashed together (prefix tag, “TapTweak”) and the resultant digest is used as a private key from which a pubkey is derived, this value being known as the tweak. The tweak pubkey is added to the internal key in order to derive the taproot output key—the key that is used in any bech32 addresses and scriptPubKeys that pay these three conditions.
When it comes time to spend this money, either Alice or Bob can provide the script they want to use, the data needed by it (a signature and, in Alice’s case, a hash preimage), the Taproot internal key, and the hash of the script they didn’t use. Alternatively, Alice and Bob can work together to use signature aggregation (after accounting for the tweak) to provide a signature in combination with the same single-key, single-signature spending pattern described in the previous subsections. As long as the data they provide in either case is correct, the spend will be accepted.
|scriptPubKey vbytes||witness vbytes||Total vbytes|
|P2WSH (1) Alice spends||34||55.25||89.25|
|P2WSH (2) Bob spends||34||47.25||81.25|
|P2WSH (3) Mutual spend||N/A||N/A||N/A|
|Taproot (1) Alice spends||35||58.50||93.50|
|Taproot (2) Bob spends||35||43.50||78.50|
|Taproot (3) Mutual spend||35||16.25||51.25|
Because we chose to use a small tree and a simple example script, the cost of using Taproot script-path spending is similar to the cost of the data that can be omitted from the unspent branch, leading to slightly higher costs for Taproot in the case where Alice spends. However, the case where Bob spends is slightly cheaper and the case of the mutual spend is significantly less expensive than any of the other options (and using it would completely hide that this was an HTLC).
The maximum depth of a tree is 32 rows, allowing a tree to contain a maximum of around four billion scripts. However, only one of those scripts would appear in any spending transaction and only 32 hashes directly related to the merkle tree would need to be generated to prove the script was part of the tree—this means that more complex scripts than our simple HTLC can gain much larger efficiency improvements than seen above (see an article focused on MAST for more information).
Slight changes to scripting with Tapscript
When using Bitcoin Script with Taproot, some of the rules have been changed, most notably:
Unused and invalid opcodes redefined: the bytes of opcodes that were never assigned, or which were made invalid by Satoshi Nakamoto, or which were created for upgrades and have not been used yet have all been changed to have the semantics of an
OP_SUCCESSopcode that unconditionally renders any script containing that code valid. Because future soft forks can only add new rules by making previously-valid things invalid, this maximal validity today will allow maximal flexibility in future soft forks. Of course, receivers get to choose what scripts they use to receive payments, so no one should include an
OP_SUCCESSopcode in their scripts until a soft fork has given it a new consensus-enforced meaning.
Schnorr signatures required: ECDSA signatures will not be accepted in Tapscript.
New script-based multisig semantics: the current
OP_CHECKMULTISIGVERIFYopcodes are not available in Tapscript. There are two alternatives. First, the existing single-sig
OP_CHECKSIGVERIFYmay be used in series. For example (2-of-2 multisig):
<A pubkey> OP_CHECKSIGVERIFY <B pubkey> OP_CHECKSIG
Second, a new
OP_CSADD) opcode may be used to increment a counter if a signature matches a specified public key. For example (2-of-3):
<A pubkey> OP_CHECKSIG <B pubkey> OP_CSADD <C pubkey> OP_CSADD OP_2 OP_EQUAL
This change is made to allow for batch verification of multiple signatures, which can significantly speed up verification compared to checking each signature independently.
Redefined sigops limit: because verifying signatures is the most CPU expensive operation in Bitcoin Script, an early version of Bitcoin added a limit on the maximum number of signature-checking operations (sigops) a block could contain, and versions of this limit were applied to later P2SH and segwit v0 scripts. However, one problem with this limit is that miners must consider two factors when trying to select what valid transactions to include in a block: fee density (fee/vbyte) and sigop density (sigops/vbyte). This is much more difficult than optimizing block composition based on just fee density.
Taproot resolves this down to one parameter by requiring valid transactions using Taproot spends include a certain amount of data for each sigop that succeeds. The rule is one free sigop and then the witness must contain 50 bytes of data for each additional sigop. Since Schnorr signatures are at least 64 bytes, this should provide more than enough space to cover all expected uses, and it means that miners can simply include the most profitable valid Taproot transactions in their blocks without worrying about sigops.
Taproot and Tapscript summary and next steps
Together, these proposals bring Bitcoin two features that developers have been wanting for years. The first of these features, Schnorr signatures, will make available increased privacy and reduced fees for the growing number of Bitcoin users taking advantage of multisig security (including LN users), and research into advanced uses of Schnorr such as scriptless scripts and discrete log contracts may enable many significant improvements in efficiency, privacy, and financial contracting without further necessary consensus changes.
The second feature, Taproot-based MAST, allows developers to write scripts that are much more complex than are possible today but still minimize their onchain impact by allowing spenders to only put the minimum amount of data onchain—lowering the feerates for advanced users while also improving their privacy.
Finally, because there’s a strong chance that many single-sig, multisig, and MAST-based spends can all be resolved using nothing but a single public key and a single signature, it may become much harder to track different users who are using different Bitcoin features—an advantage that benefits all Bitcoin users by making bitcoin ownership harder to track and thus bitcoins more fungible and harder to efficiently censor in piecemeal fashion.
However, the future of these proposals is not certain. At this stage, they need careful review by experts in cryptography and the Bitcoin protocol, as well as from application developers who plan to use these features. Following that, they will need to be implemented for full nodes, which will require more review and also extensive testing (an example implementation is available, but it’s currently meant to help proposal reviewers). Finally, it will be up to the people who use their own full nodes for validating their incoming payments to decide whether or not they want to enforce this proposal.
Optech doesn’t know when—or if—any of those goals will be accomplished, but we’ll do our best to keep you appraised of any progress in future editions of this newsletter.
Bech32 sending support
Week 9 of 24. Until the second anniversary of the segwit soft fork lock-in on 24 August 2019, the Optech Newsletter will contain this weekly section that provides information to help developers and organizations implement bech32 sending support—the ability to pay native segwit addresses. This doesn’t require implementing segwit yourself, but it does allow the people you pay to access all of segwit’s multiple benefits.
BIP173 forbids bech32 addresses from using mixed case. The preferred way to write a bech32 address is in all lowercase, but there’s one case where all uppercase makes sense: QR codes. Take a look at the following two QR codes for the same address with the only difference being lowercase versus uppercase:
This is a deliberate design feature of bech32. QR codes can be created in several modes that support different character sets. The binary mode character set is used for legacy addresses because they require mixed case. However, Bech32 addresses for Bitcoin2 can be represented using only numbers and capital letters, so they can use the smaller uppercase alphanumeric character set. Because this set is smaller, it uses fewer bits to encode each character in a QR code, allowing the resultant code to be less complex.
Bitcoin addresses are often used in BIP21 URIs. BIP21 technically
requires base58check formatting, but at least some wallets that support
native segwit addresses (such as Bitcoin Core) allow them to be used
with bech32. Although the preferred form of BIP21’s scheme identifier
bitcoin: is lowercase, it can also be uppercased as allowed by both
BIP21 and RFC3986. This also produces a less complex image
(although in this case, our QR encoding library gave both images the
& needed for passing additional parameters
in a BIP21 URI are not part of the QR code uppercase character set, so
only binary mode can be used for those characters. Additionally, BIP21 specifies that query
parameter names such as
label are case sensitive, so
uppercase versions of them aren’t expected to work anyway.
However, QR codes can support mixed character sets and doing so will always be at least slightly more efficient when used with a string that either begins or ends with an all-caps substring containing a bech32 address. This is because the minimum allowed size of a bech32 address (14 characters) combined with the efficiency gain from using uppercase mode (31.25%) exceeds the worst-case overhead of switching modes (20 extra bits). At least two QR code encoders we’re aware of, libqrencode (C) and node-qrcode (JS), automatically mix character sets by default as necessary to produce the least-complex QR code possible:
In summary, when using bech32 addresses in QR codes, consider uppercasing them and any other adjacent characters that can be uppercased in order to produce smaller and less complex QR codes. (However, for all other purposes, bech32 addresses should use all lowercase characters.)
Correction: an earlier version of this section claimed that QR codes which included BIP21 query parameters needed to use binary mode. Nadav Ivgi kindly informed us that it was possible to mix character modes, and we’ve updated the final two paragraphs of this section accordingly.
Notable code and documentation changes
Bitcoin Core #15730 extends the
getwalletinfoRPC with a
scanningfield that tells the user how far along the program is in rescanning the block chain for transactions affecting their wallet (if the user requested a rescan or it was automatically triggered). Otherwise, it simply returns
Bitcoin Core #15930 deprecates the
getunconfirmedbalanceRPC and the three different balance fields in the
getwalletinfoRPC. In their place, a
getbalancesRPC is added that provides provides two sets of fields. For balances the wallet can spend (“IsMine”), a
trustedfield for outputs either created by the wallet itself or which have at least one confirmation; an
untrusted_pendingfield for outputs in the mempool that pay the wallet; and an
immaturefield for outputs from a miner generation transaction that haven’t yet received 100 confirmations (the earliest they can be spent). For balances the wallet is only watching (“watchonly”), the same three fields are provided under a different object.
C-Lightning #2524 records details about failed attempts to forward payments so that it can later display those details in the
Bitcoin Core #15939 removes the build target for 32-bit Windows, meaning there will be no Win32 binaries for future versions of Bitcoin Core. Evidence suggests that very few (if any) Bitcoin Core users are using 32-bit Windows.
We thank Anthony Towns and Pieter Wuille for their insightful reviews of a draft of this newsletter’s Taproot overview. Any remaining errors are the fault of the newsletter author.
Technically, vbytes are an integer unit to make them backwards-compatible with the bytes unit in legacy Bitcoin block weighting. We use them in this document as a floating-point unit for additional precision. ↩
Bech32 addresses have three parts, a Human Readable Prefix (HRP) such as
bc, a separator (always a
1), and a data part. The separator and the data part are guaranteed to be part of QR code’s uppercase alphanumeric set, but the range of characters allowed in the HRP according to BIP173 includes punctuation that isn’t part of that uppercase alphanumeric set. Specifically, the following characters are allowed in bech32 HRPs but are not part of the QR code uppercase alphanumeric set:
None of the bech32 HRPs used in Bitcoin (bc, tb, bcrt) use any of these characters, and neither does any other application as far as we know. However, you may not want to make any assumptions in your code about uppercase always providing smaller QR codes for non-Bitcoin bech32 addresses. ↩