status: draft

Maintainer(s): Cayle Sharrock


The 3-Clause BSD Licence.

Copyright 2020 The Tari Development Community

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

  1. Redistributions of this document must retain the above copyright notice, this list of conditions and the following disclaimer.
  2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
  3. Neither the name of the copyright holder nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.



The keywords "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY" and "OPTIONAL" in this document are to be interpreted as described in BCP 14 (covering RFC2119 and RFC8174) when, and only when, they appear in all capitals, as shown here.


This document and its content are intended for information purposes only and may be subject to change or update without notice.

This document may include preliminary concepts that may or may not be in the process of being developed by the Tari community. The release of this document is intended solely for review and discussion by the community of the technological merits of the potential system outlined herein.


This Request for Comment (RFC) presents a proposal for introducing TariScript into the Tari base layer protocol. Tari Script aims to provide a general mechanism for enabling further extensions such as side-chains, the DAN, one-sided payments and atomic swaps.

$$ \newcommand{\script}{\alpha} % utxo script \newcommand{\input}{ \theta } \newcommand{\cat}{\Vert} \newcommand{\so}{\gamma} % script offset \newcommand{\hash}[1]{\mathrm{H}\bigl({#1}\bigr)} $$


It is hopefully clear to anyone reading these RFCs that the ambitions of the Tari project extend beyond a Mimblewimble-clone-coin. It should also be fairly clear that vanilla Mimblewimble does not have the feature set to provide functionality such as:

  • One-sided payments
  • Multiparty side-chain peg outs and peg-ins
  • Generalised smart contracts

Extensions to Mimblewimble have been proposed for most of these features, for example, David Burkett's one-sided payment proposal for LiteCoin (LIP-004), this project's HTLC RFC and the pegging proposals for the Clacks side-chain.

Some smart contract features are possible, or partly possible in vanilla [Mimblewimble] using Scriptless script, such as

  • Atomic swaps
  • Hash time-locked contracts

This RFC makes the case that if Tari were to implement a scripting language similar to Bitcoin script, then all of these use cases will collapse and can be achieved under a single set of (relatively minor) modifications and additions to the current Tari and Mimblewimble protocol.

Scripting on Mimblewimble

To the author's knowledge, none of existing [Mimblewimble] projects have employed a scripting language, nor are there ambitions to do so.

Grin styles itself as a "Minimal implementation of the Mimblewimble protocol", so one might infer that this status is unlikely to change soon.

Beam recently announced the inclusion of a smart contract protocol, which allows users to execute arbitrary code (shaders) in a sandboxed Beam VM and have the results of that code interact with transactions.

Mimblewimble coin is a fork of Grin and "considers the protocol ossified".

Litecoin is in the process of adding Mimblewimble as a side-chain. As of this writing, there appear to be no plans to include general scripting into the protocol.

Scriptless scripts

Scriptless script is a wonderfully elegant technology and inclusion of TariScript does not preclude the use of Scriptless script in Tari. However, scriptless scripts have some disadvantages:

  • They are often difficult to reason about, with the result that the development of features based on scriptless scripts is essentially in the hands of a very select group of cryptographers and developers.
  • The use case set is impressive considering that the "scripts" are essentially signature wrangling, but is still somewhat limited.
  • Every feature must be written and implemented separately using the specific and specialised protocol designed for that feature. That is, it cannot be used as a dynamic scripting framework on a running blockchain.

TariScript - a brief motivation

The essential idea of TariScript is as follows:

Given a standard Tari UTXO, we add additional restrictions on whether that UTXO can be included as a valid input in a transaction.

As long as those conditions are suitably committed to, are not malleable throughout the existence of the UTXO, and one can prove that the script came from the UTXO owner, then these conditions are not that different to the requirement of having range proofs attached to UTXOs, which require that the value of Tari commitments is non-negative.

This argument is independent of the nature of the additional restrictions. Specifically, if these restrictions are manifested as a script that provides additional constraints over whether a UTXO may be spent, the same arguments apply.

This means that in a very hand-wavy sort of way, there ought to be no reason that TariScript is not workable.

Note that range proofs can be discarded after a UTXO is spent. This entails that the global security guarantees of Mimblewimble are not that every transaction in history was valid from an inflation perspective, but that the net effect of all transactions lead to zero spurious inflation. This sounds worse than it is, since locally, every individual transaction is checked for validity at the time of inclusion in the blockchain.

If it somehow happened that two illegal transactions made it into the blockchain (perhaps due to a bug), and the two cancelled each other out such that the global coin supply was still correct, one would never know this when doing a chain synchronisation in pruned mode.

But if there was a steady inflation bug due to invalid range proofs making it into the blockchain, a pruned mode sync would still detect that something was awry, because the global coin supply balance acts as another check.

With TariScript, once the script has been pruned away, and then there is a re-org to an earlier point on the chain, then there's no way to ensure that the script was honoured unless you run an archival node.

This is broadly in keeping with the Mimblewimble security guarantees that, in pruned-mode synchronisation, individual transactions are not necessarily verified during chain synchronisation.

However, the guarantee that no additional coins are created or destroyed remains intact.

Put another way, the blockchain relies on the network at the time to enforce the TariScript spending rules. This means that the scheme may be susceptible to certain horizon attacks.

Incidentally, a single honest archival node would be able to detect any fraud on the same chain and provide a simple proof that a transaction did not honour the redeem script.

Additional requirements

The assumptions that broadly equate scripting with range proofs in the above argument are:

  • The script must be committed to the blockchain.
  • The script must not be malleable in any way without invalidating the transaction. This restriction extends to all participants, including the UTXO owner.
  • We must be able to prove that the UTXO originator provides the script and no-one else.
  • The scripts and their redeeming inputs must be stored on the block chain. In particular, the input data must not be malleable.

The next section discusses the specific proposals for achieving these requirements.

Protocol modifications

Please refer to Notation, which provides important pre-knowledge for the remainder of the report.

At a high level, TariScript works as follows:

  • The spending script \((\script)\) is recorded in the transaction UTXO.
  • UTXOs also define a new, sender offset public key \((K_{O})\).
  • After the script \((\script)\) is executed, the execution stack must contain exactly one value that will be interpreted as the script public key \((K_{S})\). One can prove ownership of a UTXO by demonstrating knowledge of both the commitment blinding factor \((k\)), and the script private key \((k_{S})\).
  • The script private key \((k_{S})\), commitment blinding factor \((k)\) and commitment value \((v)\) signs the script input data \((\input)\).
  • The sender offset private keys \((k_{O})\) and script private keys \((k_{S})\) are used in conjunction to create a script offset \((\so)\), which are used in the consensus balance to prevent a number of attacks.

UTXO data commitments

The script, as well as other UTXO metadata, such as the output features are signed for with the sender offset private key to prevent malleability. As we will describe later, the notion of a script offset is introduced to prevent cut-through and forces the preservation of these commitments until they are recorded into the blockchain.

There are two changes to the protocol data structures that must be made to allow this scheme to work.

The first is a relatively minor adjustment to the transaction output definition. The second is the inclusion of script input data and an additional public key in the transaction input field.

Transaction output changes

The current definition of a Tari UTXO is:

pub struct TransactionOutput {
    /// Options for an output's structure or use
    features: OutputFeatures,
    /// The homomorphic commitment representing the output amount
    commitment: Commitment,
    /// A proof that the commitment is in the right range
    proof: RangeProof,

Note: Currently, the output features are actually malleable. TariScript fixes this.

Under TariScript, this definition changes to accommodate the script and the sender offset public keys:

pub struct TransactionOutput {
    /// Options for an output's structure or use
    features: OutputFeatures,
    /// The homomorphic commitment representing the output amount
    commitment: Commitment,
    /// A proof that the commitment is in the right range
    proof: RangeProof,
    /// The serialised script
    script: Vec<u8>,
    /// The sender offset pubkey, K_O
    sender_offset_public_key: PublicKey
    /// UTXO signature signing the transaction output data and the homomorphic commitment with a combination 
    /// of the homomorphic commitment private values (amount and blinding factor) and the sender offset private key.
    metadata_signature: CommitmentSignature,

The commitment definition is unchanged:

$$ \begin{aligned} C_i = v_i \cdot H + k_i \cdot G \end{aligned} \tag{1} $$

The metadata signature is an aggregated Commitment Signature signed with a combination of the homomorphic commitment private values \( (v_i \, , \, k_i )\), with the spending key only known by the receiver, and sender offset private key \(k_{Oi}\), only known by the sender. The signature challenge consists of all the transaction output metadata, effectively forming a contract between the sender and receiver, making all those values non-malleable and ensuring only the sender and receiver can enter into this contract. (See Signature on Commitment values by F. Zhang et. al. and Commitment Signature by G. Yu. for details about this signature.)

Note that the Commitment Signature is an aggregated signature between the sender and receiver, which is constructed as follows.

The sender portion of the public nonce is:

$$ \begin{aligned} R_{MSi} &= r_{MSi_a} \cdot H + r_{MSi_b} \cdot G \end{aligned} \tag{2} $$

The sender sends \(K_{Oi}, R_{MSi}\) to the receiver, who now has all the required information to calculate the final challenge. The receiver portion of the public nonce is:

$$ \begin{aligned} R_{MRi} &= r_{MRi_a} \cdot H + r_{MRi_b} \cdot G \end{aligned} \tag{3} $$

The final challenge is:

$$ \begin{aligned} e &= \hash{ (R_{MSi} + R_{MRi}) \cat \script_i \cat F_i \cat K_{Oi} \cat C_i} \\ \end{aligned} \tag{4} $$

The receiver can now calculate their portion of the aggregated Commitment Signature as:

$$ \begin{aligned} R_{MRi} &= r_{MRi_a} \cdot H + r_{MRi_b} \cdot G \\ a_{MRi} &= r_{MRi_a} + e(v_{i}) \\ b_{MRi} &= r_{MRi_b} + e(k_i) \end{aligned} \tag{5} $$

The receiver sends \( s_{MRi} = (a_{MRi}, b_{MRi}, R_{MRi} ) \) along with the other partial transaction information to the sender. The sender starts by calculating the final challenge (16) and then completes their part of the aggregated Commitment Signature.

$$ \begin{aligned} a_{MSi} &= r_{MSi_a} \\ b_{MSi} &= r_{MSi_b} + e(k_{Oi}) \end{aligned} \tag{6} $$

Note that (6) is a slight deviation to the Commitment Signature due to the fact that we do not have a private value for \( a_{MSi} \). This is equivalent to having a commitment to the value of zero (i.e. \( C_i = 0 \cdot H + k_{Oi} \cdot G \)).

The final aggregated Commitment Signature is combined a follows:

$$ \begin{aligned} s_{Mi} &= (a_{Mi}, b_{Mi}, R_{Mi} ) \\ &= ((a_{MSi} + a_{MRi}), (b_{MSi} + b_{MRi}), (R_{MSi} + R_{MRi}) ) \end{aligned} \tag{7} $$

This is verified by the following:

$$ \begin{aligned} a_{Mi} \cdot H + b_{Mi} \cdot G = R_{Mi} + (C_i + K_{Oi})e \end{aligned} \tag{8} $$

However, when evaluating (8) it is evident that the receiver can calculate \( r_{MSi_a} \) as follows:

$$ \begin{aligned} a_{MSi} &= a_{Mi} - a_{MRi} \\ r_{MSi_a} &= a_{MSi} \end{aligned} \tag{9} $$

To not leak any private nonces from the sender to the receiver, \( r_{MSi_a} \) can be set to zero, effectively turning the sender portion of the aggregated signature into a normal Schnorr signature. Therefore, (2), (6) and (7) can be rewritten as follows:

$$ \begin{aligned} R_{MSi} &= r_{MSi_b} \cdot G \end{aligned} \tag{10} $$

$$ \begin{aligned} a_{MSi} &= 0 \\ b_{MSi} &= r_{MSi_b} + e(k_{Oi}) \end{aligned} \tag{11} $$

$$ \begin{aligned} s_{Mi} &= (a_{Mi}, b_{Mi}, R_{Mi} ) \\ &= (a_{MRi}, (b_{MSi} + b_{MRi}), (R_{MSi} + R_{MRi}) ) \end{aligned} \tag{12} $$

Note that:

  • The UTXO has a positive value v like any normal UTXO.
  • The script and the output features can no longer be changed by the miner or any other party. This includes the sender and receiver; they would need to cooperate to enter into a new contract to change any metadata, otherwise the metadata signature will be invalidated.
  • We provide the complete script on the output.

Transaction input changes

The current definition of an input is

pub struct TransactionInput {
    /// The features of the output being spent. We will check maturity for all outputs.
    pub features: OutputFeatures,
    /// The commitment referencing the output being spent.
    pub commitment: Commitment,

In standard Mimblewimble, an input is the same as an output sans range proof. The range proof doesn't need to be checked again when spending inputs, so it is dropped.

The updated input definition is:

pub struct TransactionInput {
    /// Options for an output's structure or use
    features: OutputFeatures,
    /// The homomorphic Pedersen commitment representing the output amount
    commitment: Commitment,
    /// The serialised script
    script: Vec<u8>,
    /// The script input data, if any
    input_data: Vec<u8>,
    /// Signature signing the script, input data, [script public key] and the homomorphic commitment with a combination 
    /// of the homomorphic commitment private values (amount and blinding factor) and the [script private key].
    script_signature: CommitmentSignature,
    /// The sender offset pubkey, K_O
    sender_offset_public_key: PublicKey

The script_signature is an aggregated Commitment Signature signed with a combination of the homomorphic commitment private values \( (v_i \, , \, k_i )\) and script private key \(k_{Si}\) to prove ownership thereof. It signs the script, the script input, script public key and the commitment:

$$ \begin{aligned} s_{Si} = (a_{Si}, b_{Si}, R_{Si} ) \end{aligned} \tag{13} $$


$$ \begin{aligned} R_{Si} &= r_{Si_a} \cdot H + r_{Si_b} \cdot G \\ a_{Si} &= r_{Si_a} + e(v_{i}) \\ b_{Si} &= r_{Si_b} + e(k_{Si}+k_i) \\ e &= \hash{ R_{Si} \cat \alpha_i \cat \input_i \cat K_{Si} \cat C_i} \\ \end{aligned} \tag{14} $$

This is verified by the following:

$$ \begin{aligned} a_{Si} \cdot H + b_{Si} \cdot G = R_{Si} + (C_i+K_{Si})e \end{aligned} \tag{15} $$

The script public key \(K_{Si}\) needed for the script signature verification is not stored with the TransactionInput, but obtained by executing the script with the provided input data. Because this signature is signed with the script private key \(k_{Si}\), it ensures that only the owner can provide the input data \(\input_i\) to the TransactionInput.

Script Offset

For every transaction an accompanying script offset \( \so \) needs to be provided. This is there to prove that every
script public key \( K_{Sj} \) and every sender offset public key \( K_{Oi} \) supplied with the UTXOs are the correct ones. The sender will know and provide sender offset private keys \(k_{Oi} \) and script private keys \(k_{Si} \); these are combined to create the script offset \( \so \), which is calculated as follows:

$$ \begin{aligned} \so = \sum_j\mathrm{k_{Sj}} - \sum_i\mathrm{k_{Oi}} \; \text{for each input}, j,\, \text{and each output}, i \end{aligned} \tag{16} $$

Verification of (16) will entail:

$$ \begin{aligned} \so \cdot G = \sum_j\mathrm{K_{Sj}} - \sum_i\mathrm{K_{Oi}} \; \text{for each input}, j,\, \text{and each output}, i \end{aligned} \tag{17} $$

We modify the transactions to be:

pub struct Transaction {
    /// A scalar offset that links outputs and inputs to prevent cut-through, enforcing the correct application of
    /// the output script.
    pub script_offset: BlindingFactor,

All script offsets (\(\so\)) from (16) contained in a block is summed together to create a total script offset (18) so that algorithm (16) still holds for a block.

$$ \begin{aligned} \so_{total} = \sum_k\mathrm{\so_{k}}\; \text{for every transaction}, k \end{aligned} \tag{18} $$

Verification of (18) will entail:

$$ \begin{aligned} \so_{total} \cdot G = \sum_j\mathrm{K_{Sj}} - \sum_i\mathrm{K_{Oi}} \; \text{for each input}, j,\, \text{and each output}, i \end{aligned} \tag{19} $$

As can be seen all information required to verify (18) is contained in a block's inputs and outputs. One important distinction to make is that the Coinbase output in a coinbase transaction does not count towards the script offset. This is because the Coinbase UTXO already has special rules accompanying it and it has no input, thus we cannot generate a script offset \( \so \). The coinbase output can allow any script \(\script_i\) and sender offset public key \( K_{Oi} \) as long as it does not break any of the rules in RFC 120 and the script is honored at spend. If the coinbase is used as in input, it is treated exactly the same as any other input.

We modify Blockheaders to be:

pub struct BlockHeader {
    /// Sum of script offsets for all kernels in this block.
    pub total_script_offset: BlindingFactor,

This notion of the script offset \(\so\) means that the no third party can remove any input or output from a transaction or the block, as that will invalidate the script offset balance equation, either (17) or (19) depending on whether the scope is a transaction or block. It is important to know that this also stops cut‑through so that we can verify all spent UTXO scripts. Because the script private key and
sender offset private key is not publicly known, its impossible to create a new script offset.

Certain scripts may allow more than one valid set of input data. Users might be led to believe that this will allow a third party to change the script keypair \((k_{Si}\),\(K_{Si})\). If an attacker can change the \(K_{Si}\) keys of the input then he can take control of the \(K_{Oi}\) as well, allowing the attacker to change the metadata of the UTXO including the script. But as shown in Script offset security, this is not possible.

If equation (17) or (19) balances then we know that every included input and output in the transaction or block has its correct script public key and sender offset public key. Signatures (2) & (13) are checked independently from script offset verification (17) and (19), and looked at in isolation those could verify correctly but can still be signed by fake keys. When doing verification in (17) and (19) you know that the signatures and the message/metadata signed by the private keys can be trusted.

Consensus changes

The Mimblewimble balance for blocks and transactions stays the same.

In addition to the changes given above, there are consensus rule changes for transaction and block validation.

Verify that for every valid transaction or block:

  1. The metadata signature \( s_{Mi} \) is valid for every output.
  2. The script executes successfully using the given input script data.
  3. The result of the script is a valid script public key, \( K_S \).
  4. The script signature, \( s_{Si} \), is valid for every input.
  5. The script offset is valid for every transaction and block.


Let's cover a few examples to illustrate the new scheme and provide justification for the additional requirements and validation steps.

Standard MW transaction

For this use case we have Alice who sends Bob some Tari. Bob's wallet is online and is able to countersign the transaction.

Alice creates a new transaction spending \( C_a \) to a new output containing the commitment \( C_b \) (ignoring fees for now).

To spend \( C_a \), she provides:

  • An input that contains \( C_a \).
  • The script input, \( \input_a \).
  • A valid script signature, \( s_{Si} \) as per (13),(14) proving that she owns the commitment \( C_a \), knows the private key, \( k_{Sa} \), corresponding to \( K_{Sa} \), the public key left on the stack after executing \( \script_a \) with \( \input_a \).
  • A sender offset public key, \( K_{Ob} \).
  • The sender portion of the public nonce, \( R_{MSi} )\, as per (10).
  • The script offset, \( \so\) with: $$ \begin{aligned} \so = k_{Sa} - k_{Ob} \end{aligned} \tag{20} $$

Alice sends the usual first round data to Bob, but now because of TariScript also includes \( K_{Ob} \) and \( R_{MSi} \). Bob can then complete his side of the transaction as per the standard Mimblewimble protocol providing the commitment \(C_b\), its public blinding factor, its rangeproof and the partial transaction signature. In addition, Bob also needs to provide a partial metadata signature as per (5) where he commits to all the transaction output metadata with a commitment signature. Because Alice is creating the transaction, she can suggest the script \( \script_b \) to use for Bob's output, similar to a bitcoin transaction, but Bob can choose a different script \(\script_b\). However, in most cases the parties will agree on using something akin to a NOP script \(\script_b\). Bob has to return this consistent set of information back to Alice.

Alice verifies the information received back from Bob, check if she agrees with the script \( \script_b \) Bob signed, and calculates her portion of the metadata signature \( s_{Mb} \) with:

$$ \begin{aligned} s_{Mb} = r_{mb} + k_{Ob} \hash{ \script_b \cat F_b \cat R_{Mb} } \end{aligned} \tag{21} $$

Alice then constructs the final aggregated metadata signature \(s_{Mb}\) as per (12) and replaces Bob's partial metadata signature in Bob's TransactionOutput.

She completes the transaction as per standard Mimblewimble protocol and also adds the script offset \( \so \), after which she sends the final transaction to Bob and broadcasts it to the network.

Transaction validation

Base nodes validate the transaction as follows:

  • They check that the usual Mimblewimble balance holds by summing inputs and outputs and validating against the excess signature. This check does not change nor do the other validation rules, such as confirming that all inputs are in the UTXO set etc.
  • The metadata signature \(s_{Ma}\) on Bob's output,
  • The input script must execute successfully using the provided input data; and the script result must be a valid public key,
  • The script signature on Alice's input is valid by checking:

$$ \begin{aligned} a_{Sa} \cdot H + b_{Sa} \cdot G = R_{Sa} + (C_a + K_{Sa})* \hash{ R_{Sa} \cat \alpha_a \cat \input_a \cat K_{Sa} \cat C_a} \end{aligned} \tag{22} $$

  • The script offset is verified by checking that the balance holds:

$$ \begin{aligned} \so \cdot{G} = K_{Sa} - K_{Ob} \end{aligned} \tag{23} $$

Finally, when Bob spends this output, he will use \( K_{Sb} \) as his script input and sign it with his script private key \( k_{Sb} \). He will choose a new sender offset public key \( K_{Oc} \) to give to the recipient, and he will construct the script offset, \( \so_b \) as follows:

$$ \begin{aligned} \so_b = k_{Sb} - k_{Oc} \end{aligned} \tag{24} $$

One sided payment

In this example, Alice pays Bob, who is not available to countersign the transaction, so Alice initiates a one-sided payment,

$$ C_a \Rightarrow C_b $$

Once again, transaction fees are ignored to simplify the illustration.

Alice owns \( C_a \) and provides the required script to spend the UTXO as was described in the previous cases.

Alice needs a public key from Bob, \( K_{Sb} \) to complete the one-sided transaction. This key can be obtained out-of-band, and might typically be Bob's wallet public key on the Tari network.

Bob requires the value \( v_b \) and blinding factor \( k_b \) to claim his payment, but he needs to be able to claim it without asking Alice for them.

This information can be obtained by using Diffie-Hellman and Bulletproof rewinding. If the blinding factor \( k_b \) was calculated with Diffie-Hellman using the sender offset keypair, (\( k_{Ob} \),\( K_{Ob} \)) as the sender keypair and the script keypair, \( (k_{Sb} \),\( K_{Sb}) \) as the receiver keypair, the blinding factor \( k_b \) can be securely calculated without communication.

Alice uses Bob's public key to create a shared secret, \( k_b \) for the output commitment, \( C_b \), using Diffie-Hellman key exchange.

Alice calculates \( k_b \) as

$$ \begin{aligned} k_b = k_{Ob} * K_{Sb} \end{aligned} \tag{25} $$

Next Alice uses Bulletproof rewinding, see RFC 180, to encrypt the value \( v_b \) into the the Bulletproof for the commitment \( C_b \). For this she uses \( k_{rewind} = \hash{k_{b}} \) as the rewind_key and \( k_{blinding} = \hash{\hash{k_{b}}} \) as the blinding key.

Alice knows the script-redeeming private key \( k_{Sa}\) for the transaction input.

Alice will create the entire transaction, including generating a new sender offset keypair and calculating the script offset,

$$ \begin{aligned} \so = k_{Sa} - k_{Ob} \end{aligned} \tag{26} $$

She also provides a script that locks the output to Bob's public key, PushPubkey(K_Sb). This will only be spendable if the spender can provide a valid signature as input that demonstrates proof of knowledge of \( k_{Sb}\) as well as the value and blinding factor of the output \(C_b\). Although Alice knowns the value and blinding factor of the output \(C_b\) only Bob knows \( k_{Sb}\).

Any base node can now verify that the transaction is complete, verify the signature on the script, and verify the script offset.

For Bob to claim his commitment he will scan the blockchain for a known script because he knowns that the script will be PushPubkey(K_Sb). In this case, the script is analogous to an address in Bitcoin or Monero. Bob's wallet can scan the blockchain looking for scripts that he would know how to resolve.

When Bob's wallet spots a known script, he requires the blinding factor, \( k_b \) and the value \( v_b \). First he uses Diffie-Hellman to calculate \( k_b \).

Bob calculates \( k_b \) as

$$ \begin{aligned} k_b = K_{Ob} * k_{Sb} \end{aligned} \tag{27} $$

Next Bob's wallet calculates \( k_{rewind} \), using \( k_{rewind} = \hash{k_{b}}\) and (\( k_{blinding} = \hash{\hash{k_{b}}} \), using those to rewind the Bulletproof to get the value \( v_b \).

Because Bob's wallet already knowns the script private key \( k_{Sb} \), he now knows all the values required to spend the commitment \( C_b \)

For Bob's part, when he discovers one-sided payments to himself, he should spend them to new outputs using a traditional transaction to thwart any potential horizon attacks in the future.

To summarise, the information required for one-sided transactions are as follows:

Transaction inputSymbolsKnowledge
commitment\( C_a = k_a \cdot G + v \cdot H \)Alice knows the blinding factor and value.
features\( F_a \)Public
script\( \alpha_a \)Public
script input\( \input_a \)Public
script signature\( s_{Sa} \)Alice knows \( k_{Sa},\, r_{Sa} \) and \( k_{a},\, v_{a} \) of the commitment \(C_a\).
sender offset public key\( K_{Oa} \)Not used in this transaction.
Transaction outputSymbolsKnowledge
commitment\( C_b = k_b \cdot G + v \cdot H \)Alice and Bob know the blinding factor and value.
features\( F_b \)Public
script\( \script_b \)Script is public; only Bob knows the correct script input.
range proofAlice and Bob know opening parameters.
sender offset public key\( K_{Ob} \)Alice knows \( k_{Ob} \).
metadata signature\( s_{Mb} \)Alice knows \( k_{Ob} \), \( (k_{b},\, v) \) and the metadata.

HTLC-like script

In this use case we have a script that controls where it can be spent. The script is out of scope for this example, but has the following rules:

  • Alice can spend the UTXO unilaterally after block n, or
  • Alice and Bob can spend it together.

This would be typically what a lightning-type channel requires.

Alice owns the commitment \( C_a \). She and Bob work together to create \( C_s\). But we don't yet know who can spend the newly created \( C_s\) and under what conditions this will be.

$$ C_a \Rightarrow C_s \Rightarrow C_x $$

Alice owns \( C_a\), so she knows the blinding factor \( k_a\) and the correct input for the script's spending conditions. Alice also generates the sender offset keypair, \( (k_{Os}, K_{Os} )\).

Now Alice and Bob proceed with the standard transaction flow.

Alice ensures that the sender offset public key \( K_{Os}\) is part of the output metadata that contains commitment \( C_s\). Alice will fill in the script with her \( k_{Sa}\) to unlock the commitment \( C_a\). Because Alice owns \( C_a\) she needs to construct \( \so\) with:

$$ \begin{aligned} \so = k_{Sa} - k_{Os} \end{aligned} \tag{28} $$

The blinding factor, \( k_s\) can be generated using a Diffie-Hellman construction. The commitment \( C_s\) needs to be constructed with the script that Bob agrees on. Until it is mined, Alice could modify the script via double-spend and thus Bob must wait until the transaction is confirmed before accepting the conditions of the smart contract between Alice and himself.

Once the UTXO is mined, both Alice and Bob possess all the knowledge required to spend the \( C_s \) UTXO. It's only the conditions of the script that will discriminate between the two.

The spending case of either Alice or Bob claiming the commitment \( C_s\) follows the same flow described in the previous examples, with the sender proving knowledge of \( k_{Ss}\) and "unlocking" the spending script.

The case of Alice and Bob spending \( C_s \) together to a new multiparty commitment requires some elaboration.

Assume that Alice and Bob want to spend \( C_s \) co-operatively. This involves the script being executed in such a way that the resulting public key on the stack is the sum of Alice and Bob's individual script keys, \( k_{SsA} \) and \( k_{SaB} \).

The script input needs to be signed by this aggregate key, and so Alice and Bob must each supply a partial signature following the usual Schnorr aggregate mechanics, but one person needs to add in the signature of the blinding factor and value.

In an analogous fashion, Alice and Bob also generate an aggregate sender offset private key \( k_{Ox}\), each using their own \( k_{OxA} \) and \( k_{OxB}\).

To be specific, Alice calculates her portion from

$$ \begin{aligned} \so_A = k_{SsA} - k_{OxA} \end{aligned} \tag{29} $$

Bob will construct his part of the \( \so\) with:

$$ \begin{aligned} \so_B = k_{SsB} - k_{OxB} \end{aligned} \tag{30} $$

And the aggregate \( \so\) is then:

$$ \begin{aligned} \so = \so_A + \so_B \end{aligned} \tag{31} $$

Notice that in this case, both \( K_{Ss} \) and \( K_{Ox}\) are aggregate keys.

Notice also that because the script resolves to an aggregate key \( K_s\) neither Alice nor Bob can claim the commitment \( C_s\) without the other party's key. If either party tries to cheat by editing the input, the script validation will fail.

If either party tries to cheat by creating a new output, the script offset will not validate correctly as it locks the output of the transaction.

A base node validating the transaction will also not be able to tell this is an aggregate transaction as all keys are aggregated Schnorr signatures. But it will be able to validate that the script input is correctly signed, thus the output public key is correct and that the \( \so\) is correctly calculated, meaning that the commitment \( C_x\) is the correct UTXO for the transaction.

To summarise, the information required for creating a multiparty UTXO is as follows:

Transaction inputSymbolsKnowledge
commitment\( C_a = k_a \cdot G + v \cdot H \)Alice knows the blinding factor and value.
features\( F_a \)Public
script\( \alpha_a \)Public
script input\( \input_a \)Public
script signature\( s_{Sa} \)Alice knows \( k_{Sa},\, r_{Sa} \) and \( k_{a},\, v_{a} \) of the commitment \(C_a\).
sender offset public key\( K_{Oa} \)Not used in this transaction.

Transaction outputSymbolsKnowledge
commitment\( C_s = k_s \cdot G + v \cdot H \)Alice and Bob know the blinding factor and value.
features\( F_s \)Public
script\( \script_s \)Script is public; Alice and Bob only knows their part of the correct script input.
range proofAlice and Bob know opening parameters.
sender offset public key\( K_{Os} = K_{OsA} + K_{OsB}\)Alice knows \( k_{OsA} \), Bob knows \( k_{OsB} \), neither party knows \( k_{Os} \).
metadata signature\( (a_{Ms} , b_{Ms} , R_{Ms}) \)Alice knows \( k_{OsA} \), Bob knows \( k_{OsB} \), both parties know \( (k_{s},\, v) \). Neither party knows \( k_{Os}\).

When spending the multi-party input:

Transaction inputSymbolsKnowledge
commitment\( C_s = k_s \cdot G + v_s \cdot H \)Alice and Bob know the blinding factor and value.
features\( F_s \)Public
script\( \alpha_s \)Public
script input\( \input_s \)Public
script signature\( (a_{Ss} ,b_{Ss} , R_{Ss}) \)Alice knows \( (k_{SsA},\, r_{SsA}) \), Bob knows \( (k_{SsB},\, r_{SsB}) \), both parties know \( (k_{s},\, v_{s}) \), neither party knows \( k_{Ss}\).
sender offset public key\( K_{Os} \)As above, Alice and Bob each know part of the sender offset key.


A major issue with many Mimblewimble extension schemes is that miners are able to cut-through UTXOs if an output is spent in the same block it was created. This makes it so that the intervening UTXO never existed; along with any checks and balances carried in that UTXO. It's also impossible to prove without additional information that cut-through even occurred (though one may suspect, since the "one" transaction would contribute two kernels to the block).

In particular, cut-through is devastating for an idea like TariScript which relies on conditions present in the UTXO being enforced.

This is a reason for the presence of the script offset in the TariScript proposal. It mathematically links all inputs and outputs of all the transactions in a block and that tallied up to create the script offset. Providing the script offset requires knowledge of keys that miners do not possess; thus they are unable to produce the necessary script offset when attempting to perform cut-through on a pair of transactions.

Lets show by example how the script offset stops cut-through, where Alice spends to Bob who spends to Carol. Ignoring fees, we have:

$$ C_a \Rightarrow C_b \Rightarrow C_c $$

For these two transactions, the total script offset is calculated as follows:

$$ \begin{aligned} \so_1 = k_{Sa} - k_{Ob}\\ \so_2 = k_{Sb} - k_{Oc}\\ \end{aligned} \tag{32} $$

$$ \begin{aligned} \so_t = \so_1 + \so_2 = (k_{Sa} + k_{Sb}) - (k_{Ob} + k_{Oc})\\ \end{aligned} \tag{33} $$

In standard Mimblewimble cut-through can be applied to get:

$$ C_a \Rightarrow C_c $$

After cut-through the total script offset becomes:

$$ \begin{aligned} \so'_t = k_{Sa} - k_{Oc}\\ \end{aligned} \tag{34} $$

As we can see:

$$ \begin{aligned} \so_t\ \neq \so'_t \\ \end{aligned} \tag{35} $$

A third party cannot generate a new script offset as only the original owner can provide the script private key \(k_{Sa}\) to create a new script offset.

Script offset security

If all the inputs in a transaction or a block contain scripts such as just NOP or CompareHeight commands, then the hypothesis is that it is possible to recreate a false script offset. Lets show by example why this is not possible. In this Example we have Alice who pays Bob with no change output:

$$ C_a \Rightarrow C_b $$

Alice has an output \(C_{a}\) which contains a script that only has a NOP command in it. This means that the script \( \script_a \) will immediately exit on execution leaving the entire input data \( \input_a \)on the stack. She sends all the required information to Bob as per the standard mw transaction, who creates an output \(C_{b}\). Because of the NOP script \( \script_a \), Bob can change the script public key \( K_{Sa}\) contained in the input data. Bob can now use his own \(k'_{Sa}\) as the script private key. He replaces the sender offset public key with his own \(K'_{Ob}\) allowing him to change the script \( \script_b \) and generate a new signature as in (2). Bob can now generate a new script offset with \(\so' = k'_{Sa} - k'_{Ob} \). Up to this point, it all seems valid. No one can detect that Bob changed the script to \( \script_b \).

But what Bob also needs to do is generate the signature in (13). For this signature Bob needs to know \(k_{Sa}, k_a, v_a\). Because Bob created a fake script private key, and there is no change in this transaction, he does know the script private key and the value. But Bob does not know the blinding factor \(k_a\) of Alice's commitment and thus cannot complete the signature in (13). Only the rightful owner of the commitment, which in Mimblewimble terms is the person who knows \( k_a, v_a\), can generate the signature in (13).

Script lock key generation

At face value, it looks like the burden for wallets has tripled, since each UTXO owner has to remember three private keys, the spend key, \( k_i \), the sender offset key \( k_{O} \) and the script key \( k_{S} \). In practice, the script key will often be a static key associated with the user's node or wallet. Even if it is not, the script and sender offset keys can be deterministically derived from the spend key. For example, \( k_{S} \) could be \( \hash{ k_i \cat \alpha} \).

Blockchain bloat

The most obvious drawback to TariScript is the effect it will have on blockchain size. UTXOs are substantially larger, with the addition of the script, script signature, and a public key to every output.

These can eventually be pruned, but will increase storage and bandwidth requirements.

Input size of a block will now be much bigger as each input was previously just a commitment and output features. Each input now includes a script, input_data, the script signature and an extra public key. This could be compacted by just broadcasting input hashes along with the missing script input data and signature, instead of the full input in transaction messages, but this will still be larger than inputs are currently.

Every header will also be bigger as it includes an extra blinding factor that will not be pruned away.

Fodder for chain analysis

Another potential drawback of TariScript is the additional information that is handed to entities wishing to perform chain analysis. Having scripts attached to outputs will often clearly mark the purpose of that UTXO. Users may wish to re-spend outputs into vanilla, default UTXOs in a mixing transaction to disassociate Tari funds from a particular script.


Where possible, the "usual" notation is used to denote terms commonly found in cryptocurrency literature. Lower case characters are used as private keys, while uppercase characters are used as public keys. New terms introduced by TariScript are assigned greek lowercase letters in most cases. The capital letter subscripts, R and S refer to a UTXO receiver and script respectively.

\( \script_i \)An output script for output i, serialised to binary.
\( F_i \)Output features for UTXO i.
\( f_t \)Transaction fee for transaction t.
\( (k_{Oi}, K_{Oi}) \)The private - public keypair for the UTXO sender offset key.
\( (k_{Si}, K_{Si}) \)The private - public keypair for the script key. The script, \( \script_i \) resolves to \( K_S \) after completing execution.
\( \so_t \)The script offset for transaction t, see (16)
\( C_i \)A Pedersen commitment to a value \( v_i \), see (1)
\( \input_i \)The serialised input for script \( \script_i \)
\( s_{Si} \)A script signature for output \( i \), see (13 - 15)
\( s_{Mi} \)A metadata signature for output \( i \), see (2 - 12)



TariScript places restrictions on who can spend UTXOs. It will also be useful for Tari digital asset applications to restrict how or where UTXOs may be spent in some cases. The general term for these sorts of restrictions are termed covenants. The Handshake white paper has a fairly good description of how covenants work.

It is beyond the scope of this RFC, but it's anticipated that TariScript would play a key role in the introduction of generalised covenant support into Tari.

Lock-time malleability

The current Tari protocol has an issue with Transaction Output Maturity malleability. This output feature is enforced in the consensus rules, but it is actually possible for a miner to change the value without invalidating the transaction.

With TariScript, output features are properly committed to in the transaction and verified as part of the script offset validation.


Thanks to David Burkett for proposing a method to prevent cut-through and willingness to discuss ideas.