"Recursive covenant" with CTV and CSFS

[Anthony Towns]

Hello world,

Some people on twitter are apparently proposing the near-term activation of
CTV and CSFS (BIP 119 and BIP 348 respectively), eg:

https://x.com/JeremyRubin/status/1895676912401252588
https://x.com/lopp/status/1895837290209161358
https://x.com/stevenroose3/status/1895881757288996914
https://x.com/reardencode/status/1871343039123452340
https://x.com/sethforprivacy/status/1895814836535378055

Since BIP 119's motivation is entirely concerned with its concept of
covenants and avoiding what it calls recursive covenants, I think it
is interesting to note that the combination of CSFS and CTV trivially
enables the construction of a "recursive covenant" as BIP 119 uses those
terms. One approach is as follows:

* Make a throwaway BIP340 private key X with 32-bit public key P.
* Calculate the tapscript "OP_OVER <P> OP_CSFS OP_VERIFY OP_CTV", and
its corresponding scriptPubKey K when combined with P as the internal public
key.
* Calculate the CTV hash corresponding to a payment of some specific value V
to K; call this hash H
* Calculate the BIP 340 signature for message H with private key X, call it S.
* Discard the private key X
* Funds sent to K can only be spent by the witness data "<H> <S>" that forwards
an amount V straight back to K.

Here's a demonstration on mutinynet:

https://mutinynet.com/address/tb1p0p5027shf4gm79c4qx8pmafcsg2lf5jd33tznmyjejrmqqx525gsk5nr58

I'd encourage people to try implementing that themselves with their
preferred tooling; personally, I found it pretty inconvenient, which I
don't think is a good indication of ecosystem readiness wrt deployment.
(For me, the two components that are annoying is doing complicated
taproot script path spends, and calculating CTV hashes)

I don't believe the existence of a construction like this poses any
problems in practice, however if there is going to be a push to activate
BIP 119 in parallel with features that directly undermine its claimed
motivation, then it would presumably be sensible to at least update
the BIP text to describe a motivation that would actually be achieved by
deployment.

Personally, I think BIP 119's motivation remains very misguided:

- the things it describes are, in general, not "covenants" [0]
- the thing it avoids is not "recursion" but unbounded recursion
- avoiding unbounded recursion is not very interesting when arbitrarily
large recursion is still possible [1]
- despite claiming that "covenants have historically been widely
considering to be unfit for Bitcoin", no evidence for this claim has
been able to be provided [2,3]
- the opposition to unbounded recursion seems to me to either mostly
or entirely be misplaced fear of things that are already possible in
bitcoin and easily avoided by people who want to avoid it, eg [4]

so, at least personally, I think almost any update to BIP 119's motivation
section would be an improvement...

[0] https://gnusha.org/pi/bitcoindev/20220719044...@erisian.com.au/
[1] https://gnusha.org/pi/bitcoindev/87k0dwr...@rustcorp.com.au/
[2] https://gnusha.org/pi/bitcoindev/0100017ee6472e02-037d355d-4c16...@email.amazonses.com/
[3] https://x.com/Ethan_Heilman/status/1194624166093369345
[4] https://gnusha.org/pi/bitcoindev/2022021715...@erisian.com.au/

Beyond being a toy example of a conflict with BIP 119's motivation
section, I think the above script could be useful in the context of the
"blind-merged-mining" component of spacechains [5]. For example, if
the commitment was to two outputs, one the recursive step and the other
being a 0sat ephemeral anchor, then the spend of the ephemeral anchor
would allow for both providing fees conveniently and for encoding the
spacechain block's commitment; competing spacechain miners would then
just be rbf'ing that spend with the parent spacechain update remaining
unchanged. The "nLockTime" and "sequences_hash" commitment in CTV would
need to be used to ensure the "one spacechain update per bitcoin block"
rule. (I believe mutinynet doesn't support ephemeral anchors however, so
I don't think there's anywhere this can be tested)

[5] https://gist.github.com/RubenSomsen/5e4be6d18e5fa526b17d8b34906b16a5#file-bmm-svg

(For a spacechain, miners would want to be confident that the private key
has been discarded. This confidence could be enhanced by creating X as a
musig2 key by a federation, in which case provided any of the private keys
used in generating X were correctly discarded, then everything is fine,
but that's still a trust assumption. Simple introspection opcodes would
work far better for this use case, both removing the trust assumption
and reducing the onchain data required)

If you're providing CTV and CSFS anyway, I don't see why you wouldn't
provide the same or similar functionality via a SIGHASH flag so that you
can avoid specifying the hash directly when you're signing it anyway,
giving an ANYPREVOUT/NOINPUT-like feature directly.

(Likewise, I don't see why you'd want to activate CAT on mainnet without
also at least re-enabling SUBSTR, and potentially also the redundant
LEFT and RIGHT operations)

For comparison, bllsh [6,7] takes the approach of providing
"bip340_verify" (directly equivalent to CSFS), "ecdsa_verify" (same but
for ECDSA rather than schnorr), "bip342_txmsg" and "tx" opcodes. A CTV
equivalent would then either involve simplying writing:

(= (bip342_txmsg 3) 0x.....)

meaining "calculate the message hash of the current tx for SIGHASH_SINGLE,
then evaluate whether the result is exactly equal to this constant"
providing one of the standard sighashes worked for your use case, or
replacing the bip342_txmsg opcode with a custom calculation of the tx
hash, along the lines of the example reimplementation of bip342_txmsg
for SIGHASH_ALL [8] in the probably more likely case that it didn't. If
someone wants to write up the BIP 119 hashing formula in bllsh, I'd
be happy to include that as an example; I think it should be a pretty
straightforward conversion from the test-tx example.

If bllsh were live today (eg on either signet or mainnet), and it were
desired to softfork in a more optimised implementation of either CTV or
ANYPREVOUT to replace people coding their own implementation in bllsh
directly, both would simply involve replacing calls to "bip342_txmsg"
with calls to a new hash calculation opcode. For CTV behaviour, usage
would look like "(= (bipXXX_txmsg) 0x...)" as above; for APO behaviour,
usage would look like "(bip340_verify KEY (bipXXX_txmsg) SIG)". That
is, the underlying "I want to hash a message in such-and-such a way"
looks the same, and how it's used is the wallet author's decision,
not a matter of how the consensus code is written.

I believe simplicity/simfony can be thought of in much the same way;
with "jet::bip_0340_verify" taking a tx hash for SIGHASH-like behaviour
[9], or "jet::eq_256" comparing a tx hash and a constant for CTV-like
behaviour [10].

[6] https://github.com/ajtowns/bllsh/
[7] https://delvingbitcoin.org/t/debuggable-lisp-scripts/1224
[8] https://github.com/ajtowns/bllsh/blob/master/examples/test-tx
[9] https://github.com/BlockstreamResearch/simfony/blob/master/examples/p2pk.simf
[10] https://github.com/BlockstreamResearch/simfony/blob/master/examples/ctv.simf

For me, the bllsh/simplicity approach makes more sense as a design
approach for the long term, and the ongoing lack of examples of killer
apps demonstrating big wins from limited slices of new functionality
leaves me completely unexcited about rushing something in the short term.
Having a flood of use cases that don't work out when looked into isn't
a good replacement for a single compelling use case that does.

Cheers,
aj

[Olaoluwa Osuntokun]

Hi AJ,

First a standard disclaimer: the contents of this email shouldn't be
interpreted as an endorsement of one covenants proposal over another.

> Since BIP 119's motivation is entirely concerned with its concept of
> covenants and avoiding what it calls recursive covenants

If we look at the motivation section of BIP 119, we find this sentence:

> This BIP introduces a simple covenant called a *template* which enables a
> limited set of highly valuable use cases without significant risk. BIP-119
> templates allow for non-recursive fully-enumerated covenants with no
> dynamic state.

You appear to have latched onto the "non-recursive" aspect, ignoring the
subsequent qualifiers of "fully-enumerated" and "with no dynamic state".

The example that you've come up with to "directly undermine" the claimed
motivations of BIP 119 is still fully enumerated (the sole state is declared
up front), and doesn't contain dynamic state (I can't spend the contract on
chain and do something like swap in another hash H, or signature S).

> I found it pretty inconvenient, which I don't think is a good indication
> of ecosystem readiness wrt deployment. (For me, the two components that
> are annoying is doing complicated taproot script path spends, and
> calculating CTV hashes)

What language/libraries did you use to produce the spend? In my own
development tooling of choice, producing complicated taproot script path
spends is pretty straight forward, so perhaps the inconvenience you ran into
says more about your dev tooling than the ecosystem readiness.

It's also worth pointing out that your example relies on private key
deletion, which if deemed acceptable, can be used to emulate CTV as is today
(though you can't create a self-referential loop that way afaict).

> For me, the bllsh/simplicity approach makes more sense as a design
> approach for the long term

Simplicity certainly has some brilliant devs working on it, but after all
these years it still seems to be struggling to exit research mode with some
"killer apps" on Liquid.

bllsh on the other hand is a very new (and cool!) project that has no
development uptake beyond its creator. Given its nascent state, it seems
rather premature to promote it as a long term solution.

Both of them are effectively a complete rewrite of Script, so compared to
some of the existing covenant proposals on the table (many of which have a
small core code footprint in the interpreter), they represent a radically
expanded scope (ecosystem changes, wallets, consensus code) and therefore
additional risks. The current Overton Window appears to be focused on a
small (LoC wise) package to enable a greater degree of permissionless
innovation on Bitcoin, while leaving the research landscape open for more
dramatic overhauls (bllsh/Simplicity) in the future.

-- Laolu

--
You received this message because you are subscribed to the Google Groups "Bitcoin Development Mailing List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bitcoindev+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/Z8eUQCfCWjdivIzn%40erisian.com.au.

[Antoine Poinsot]

Hi,

Just picking on one thing Laolu said:

The current Overton Window appears to be focused on a small (LoC wise) package to enable a greater degree of permissionless innovation on Bitcoin

For what it's worth i'm not sure this is the correct focus. Bitcoin Script is so notoriously unpredictable and hard to reason about that most of what matters is outside of the lines of code changed. Of course it depends on the specifics, but rewriting a clean interpreter that we can actually reason about does not strike me as a necessarily riskier approach than "just changing a few lines of code" in an interpreter that hardly anyone knows how it really behaves in all cases.

Antoine

To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/CAO3Pvs-1H2s5Dso0z5CjKcHcPvQjG6PMMXvgkzLwXgCHWxNV_Q%40mail.gmail.com.

[moonsettler]

Hi AJ,

I don't even think about this "recursion" as an issue in itself anymore. The way CSFS enables "recursion" with deleted key covenants basically is useful for some things not so much for others. Useful for vaults, possibly somewhat useful for spacechains, pretty useless for tokens.

It's not even a really a meaningful distinction as you noted in general.

What's more interesting is "do these set of changes allow for 'native' fungible tokens you can _identify_ and interact with in script in a way that is enforced by consensus"? Can you build AMMs for them? For a lot of proposals currently discussed we actually know how to do that. Anything fully generic will trivially unlock this capability.

The two primitives involved are state carrying (amounts) and quining (aka recursion). These are the truly significant thresholds for changes that can possibly alter the nature of bitcoin and how people use it. Only one of them is not enough. Beyond these there remains the issue of novel trust minimized two way pegs to other chains like Ethereum which would also be in high demand, in fact probably prioritized in funding over all other things we are discussing in relation to covenants.

After all these years I'm confident that for LNhance (CTV+CSFS+IKEY+PC) the answer is NO.

BR,
moonsettler

[Antoine Riard]

Hi AJ,

> I don't believe the existence of a construction like this poses any
> problems in practice, however if there is going to be a push to activate
> BIP 119 in parallel with features that directly undermine its claimed
> motivation, then it would presumably be sensible to at least update
> the BIP text to describe a motivation that would actually be achieved by
> deployment.

I do...

https://gnusha.org/pi/bitcoindev/f594c2f8-d712-48e4-a010-778dd4d0cadb@Spark/
https://blog.bitmex.com/txwithhold-smart-contracts/

Terminology in the article is the following, as a reminder:
- target transaction: a tx to be withheld
- target tx: a fee, which a victim pays for the target tx inclusion
- attacker: an actor willing to withhold the target tx
- victim: a spender of the target tx
- reward transaction: a tx paying out a reward to the miner which withheld a target tx

With OP_CHECKSIGFROMSTACK, which is iirc <signature> <pubkey> <message>
as an input stack, one can have the <message> being an already spent transaction.

From then, one can build a TxWithhold for a LN commitment transaction, where
the <message> is the latest valid commitment transaction to spend a chan funding
output (each counterparty re-build the remote commitment tx for counter-signature
of the 2-of-2 P2WSH).

After that, one can build a script: <proof-of-target-UTXO-mining=commitment_tx"
OP_CSFS> OR <<bounty_timelock> <OP_CHECKLOCKTIMEVERIFY> <recursive_bounty_sig |
SIGHASH_SINGLE> OP_CHECKSIG. Using SIGHASH_SINGLE the TxWithhold attacker can
make the funding UTXO amount available as a "anyone-can-spend" and force a re-
commitment to the same tx-withholding script.

There is no hash dependency, as the "proof-of-target-UTXO-mining" is a hash given
as part of the input stack, be it for legacy spends or segwit spends.

One can have multiple script branches for all the versions of the commitment
transaction, at least the latest 2 no-penalty one.

The TxWithhold attacker can publish the script in an unrelated inscription
in the chain itself, to make the "bribing" TxWithhold contract available to
any miner anonymously wishing to engage in a TxWithhold to maximize its income
for given hashrate capabilities.

Bonus point: if you the reader can come with a construction to do a TxWithold on
the spend of any coinbase output, at least more than the delay of COINBASE_MATURITY=
100 blocks to get the `blockReward`.

Bonus Bonus point: if you the reader can come with a detailed protocol for all
the miners in a N retarget period to get a reward % share of a TxWithhold instance
weighted by their respective hashrate capabilities, by only using on-chain
inscriptions.

Best,
Antoine
OTS hash: 28d42f8caeedb4760f3a4ddb39adf5443edaf63741560e3cee264237c0c126b5

[Greg Sanders]

> Of course it depends on the specifics, but rewriting a clean interpreter that we can actually reason about does not strike me as a necessarily riskier approach than "just changing a few lines of code" in an interpreter that hardly anyone knows how it really behaves in all cases.

It's certainly something to consider when weighing further off Bitcoin Script updates: From here is something like "Great Script Restoration" ever the right choice vs a from scratch overhaul? I am less persuaded that consensus risk is particularly high for very narrowly scoped changes, ignoring the "fixed" costs of changing consensus, maintenance burden, MEVil risks, etc. The risk-reward ratio may be suboptimal of course.

Greg

[moonsettler]

Hi All,

> I am less persuaded that consensus risk is particularly high for very narrowly scoped changes

Agreed.

Some people out there seem to conflate execution risks with crypto-economic incentives risks.
Better designed script systems obviously reduce execution risks and unintended consensus failure risks and make maintenance easier.
They also quiet easily blow the lid off other types of risks by nature of being better and more capable.

Paradoxically the more expressive bitcoin script becomes over time, the less likely that a script system overhaul comes with a nasty surprise.

BR,
moonsettler

> To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/17e7eb49-77b7-4f2f-be40-a6649e610ce5n%40googlegroups.com.

[Antoine Riard]

Hi all,

> After that, one can build a script: <proof-of-target-UTXO-mining=commitment_tx"
> OP_CSFS> OR <<bounty_timelock> <OP_CHECKLOCKTIMEVERIFY> <recursive_bounty_sig |
> SIGHASH_SINGLE> OP_CHECKSIG. Using SIGHASH_SINGLE the TxWithhold attacker can
> make the funding UTXO amount available as a "anyone-can-spend" and force a re-
> commitment to the same tx-withholding script.

Correcting myself, after more thinking I believe to get a non-forgeable "proof-of
-target-UTXO-mining" there should be a merkle branch proof of the txid back to
the mined block. Otherwise, any participant in the target UTXO / transaction (e.g
a LN channel) can confirm the target transaction to execute the contracting protocol
according to its semantics _and_ generate the "proof-of-target-UTXO-mining" to
claim the "anyone-can-spend" bribing output.

So to do a merkle branch proof, script would have to be of the rough form e.g:
< <leaf_node_ab> <OP_SHA256> <OP_CAT> <leaf_node_cd> <OP_SHA256> <OP_CAT>, or
any equivalent opcode primitive allowing to get concatenation in the script.

For the design of a TxWithhold, all that you wish is a proof-of-existence of a
block B + transaction T (in the mathematical sense). That way if a timelock is
reached after X blocks, and no proof-of-existence as been brought until then,
this logically implies that B + T do not exist. Or at least have not been published,
which for the purpose of the chain being the publication space and anti-double-spend
oracle of UTXO spends is the same.

I don't believe with OP_CHECKSIGFROMSTACK alone you can build such proof-of-target-UTXO
-mining, though I'm not sure of such statement. All cov primitives proposals are coming
without formal analysis of the limits of the expressivity extensions, there has been
progress since OP_EVAL, though it's always tooling on a subset of current Bitcoin
Script not extension.

At the very least, and in reference to what is described in naumenkogs's original
TxWithhold article, the introduction of OP_CSFS let you have m-of-n oracles doing
attestations in the script that a block has been confirmed, eventually with a tx.
This is already a powerful building block for TxWithhold. At first thought, I don't
see how you can have the equivalent with OP_CHECKTEMPLATEVERIFY, by design the scope
is reduced.

On the distinction between execution risks with crypto-economic incentives risks
brought by some on this thread, this is unclear what is meant exactly. Of course
with execution risks, it might more think affecting full-node security (e.g a DoS
vector, reason a lot of opcodes were disabled back in ~2010), though it could be
just a risk of misuage of the primitive by use-cases toolchain (e.g generating
a consensus-invalid tx due to misuse of the timelock flags) or an attack driven
by crypto-ecnomic incentives (e.g selfish-mining or block-withholding attack at
the miner-level). Though even for a simple DoS, one can evaluate the CPU / bandwidth
cost in quantitative terms to gauge if it's a serious risk, there is an area of
uncertainty, or it's not a risk.

When we we talk about security risks, one has to see things that it's more a continuum,
where an attacker vector can be used as a building block for more sophisticated
attacks. In the original bitcoin paper, crypto-incentives themselves are weighted
to evaluate the soundness of the system in section 6. and in section 11.

Best,
Antoine
OTS hash: 2c3e2e41ed67484f5d58138413cfca7f0aba5d7b4d448f2895a1b70b2886e9d8

[Anthony Towns]

On Tue, Mar 04, 2025 at 10:14:59PM -0800, Olaoluwa Osuntokun wrote:
> > Since BIP 119's motivation is entirely concerned with its concept of
> > covenants and avoiding what it calls recursive covenants
> If we look at the motivation section of BIP 119, we find this sentence:
> > This BIP introduces a simple covenant called a *template* which enables a
> > limited set of highly valuable use cases without significant risk. BIP-119
> > templates allow for non-recursive fully-enumerated covenants with no
> > dynamic state.
> You appear to have latched onto the "non-recursive" aspect, ignoring the
> subsequent qualifiers of "fully-enumerated" and "with no dynamic state".
>
> The example that you've come up with to "directly undermine" the claimed
> motivations of BIP 119 is still fully enumerated (the sole state is declared
> up front), and doesn't contain dynamic state (I can't spend the contract on
> chain and do something like swap in another hash H, or signature S).

The reason "fully-enumerated" provides any "safety" is that it occurs
when the scriptPubKey is chosen -- without the availability of CSFS,
you either include the CTV hash in the scriptPubKey or the use of CTV
provides no protection at all.

My example does not include the CTV hash in the scriptPubKey, which is
what allows the CTV hash to then commit to the scriptPubKey, which in turn
allows for the unbounded recursion.

If you instead did not delete the CSFS private key would allow you to
swap in another hash H or signature S in future. That would perhaps
allow designing an unbounded state machine where a master key can add
new states in future. It's not immediately obvious to me if there's
anything interesting that can be done with that.

In any event, if there is some weird subset of use cases that are somehow
both scary and still prevented even by the combination of CTV and CSFS
the BIP should be updated to document that.

> > For me, the bllsh/simplicity approach makes more sense as a design
> > approach for the long term
> Simplicity certainly has some brilliant devs working on it, but after all
> these years it still seems to be struggling to exit research mode with some
> "killer apps" on Liquid.

https://github.com/ElementsProject/elements/pull/1427 suggests
Simplicity's potentially going live on Liquid any day now.

> The current Overton Window appears to be focused on a
> small (LoC wise) package to enable a greater degree of permissionless
> innovation on Bitcoin, while leaving the research landscape open for more
> dramatic overhauls (bllsh/Simplicity) in the future.

The concept of an "Overton window" is a political one, used for when
there has been successful political pressure to exclude some subjects
from discussion for reasons other than their underlying merits. That's
not a good idea if you want to maintain high quality, and it's probably
not compatible at all with a project that aims to be decentralised in
any meaningful way.

Certainly a small change (though LoC is a bad measure of that -- how
many LoC does it take to drop the 21M limit, or to drop the subsidy from
3.125 BTC to 0 BTC?) is better than a large change all else being equal;
but all else isn't equal: different changes enable different feature
sets. The question you should be asking is whether we're getting useful
feature sets from the small changes being proposed.

If you want "permissionless innovation", then many small incremental
consensus changes are not a good way of doing it -- as that involves
asking the global network of Bitcoin users for permission for each
individual change.

Cheers,
aj

[Anthony Towns]

On Wed, Mar 05, 2025 at 02:46:08PM -0800, Antoine Riard wrote:
> > I don't believe the existence of a construction like this poses any
> > problems in practice, however if there is going to be a push to activate
> > BIP 119 in parallel with features that directly undermine its claimed
> > motivation, then it would presumably be sensible to at least update
> > the BIP text to describe a motivation that would actually be achieved by
> > deployment.
> I do...
> https://gnusha.org/pi/bitcoindev/f594c2f8-d712-48e4-a010-778dd4d0cadb@Spark/
> https://blog.bitmex.com/txwithhold-smart-contracts/

I don't believe being able to pay for censorship on-chain is any more
threatening than being able to pay for censorship off-chain.

The bitmex blog post there relies on having a trusted oracle to release
DLC payments if the target tx wasn't mined. If you have that level of
trust anyway, then just putting funds in escrow, having miners register
bolt12 invoices with the oracle, and having the oracle make the payments
when it's satisfied blocks are sufficiently confirmed has a pretty
similar risk profile.

> With OP_CHECKSIGFROMSTACK, which is iirc <signature> <pubkey> <message>

It's <signature> <message> <pubkey> with pubkey at the top of the stack.
https://github.com/bitcoin/bips/blob/master/bip-0348.md

The same is also true of both Elements' CSFS and Bitcoin-Cash's CHECKDATASIG.

Cheers,
aj

[Anthony Towns]

On Thu, Mar 06, 2025 at 09:17:41AM -0800, Greg Sanders wrote:
> > Of course it depends on the specifics, but rewriting a clean interpreter
> > that we can actually reason about does not strike me as a necessarily
> > riskier approach than "just changing a few lines of code" in an interpreter
> > that hardly anyone knows how it really behaves in all cases.
> It's certainly something to consider when weighing further off Bitcoin
> Script updates: From here is something like "Great Script Restoration" ever
> the right choice vs a from scratch overhaul? I am less persuaded that
> consensus risk is particularly high for very narrowly scoped changes,
> ignoring the "fixed" costs of changing consensus, maintenance burden, MEVil
> risks, etc. The risk-reward ratio may be suboptimal of course.

I think "narrowly scoped changes" is doing a lot of work there. Was
tapscript a narrowly scoped change? I would have said so. However,
one thing it did was remove the limit on sript size/opcodes, which
necessitated a change to how OP_IF/OP_ELSE were implemented to avoid
potential quadratic execution time in the length of the script.

https://github.com/bitcoin/bitcoin/pull/16902

On the other hand, treating the current interpreter code as an artifact
of wise ancients, whose knowledge of how to safely modify it has been
lost to time doesn't really seem like a great approach to me, either.

Cheers,
aj

[James O'Beirne]

On Friday, March 7, 2025 at 4:26:36 PM UTC-5 Anthony Towns wrote:

If you instead did not delete the CSFS private key would allow you to
swap in another hash H or signature S in future. That would perhaps
allow designing an unbounded state machine where a master key can add
new states in future.

I'm not sure what your point here is - that we can do stupid things with CTV + CSFS? That's universally true in bitcoin; I can have an "unbounded state machine" by sending myself the same UTXO back and forth indefinitely with CHECKSIG.

As with everything in bitcoin, the chain is insulated from stupidity like that by fees, the UTXO model, and block cadence, so what's the problem?

 

https://github.com/ElementsProject/elements/pull/1427 suggests
Simplicity's potentially going live on Liquid any day now.

Probably worth noting that CSFS and many advanced introspection opcodes (which allow synthesizing CTV) have been live for almost *four years now* on Liquid (https://github.com/ElementsProject/elements/blob/master/doc/tapscript_opcodes.md#new-opcodes-for-additional-functionality).

 

The concept of an "Overton window" is a political one, used for when
there has been successful political pressure to exclude some subjects
from discussion for reasons other than their underlying merits. That's
not a good idea if you want to maintain high quality, and it's probably
not compatible at all with a project that aims to be decentralised in
any meaningful way.

Sorry to tell you, but given that changing consensus requires soliciting buy-in from a wide variety of people across the Bitcoin ecosystem with varying levels of technical ability, it is inherently a political process.

Beyond that, "Overton window" is an appropriate device in the sense that roasbeef is using it because the more substantial a proposed change is, the more time is needed by the technical ecosystem to digest it, both in terms of tooling and safety analysis. Introducing an entirely new script interpreter on the basis of a Lisp (or combinator calculus, or whatever Simplicity's claim is) is indeed far outside of that "Overton" window - much farther than tacking on what is essentially just a new SIGHASH mode to the existing interpreter.

 

Certainly a small change (though LoC is a bad measure of that -- how
many LoC does it take to drop the 21M limit, or to drop the subsidy from
3.125 BTC to 0 BTC?) is better than a large change all else being equal;
but all else isn't equal: different changes enable different feature
sets. The question you should be asking is whether we're getting useful
feature sets from the small changes being proposed.

Let's not be petty here - it's clear he's talking about LoC within the script interpreter, which is a different context than the codebase as a whole. Within the context of the interpreter, LoC is indeed a decent heuristic for marginal risk. Of course, nobody's saying it's perfect.

James

[Nadav Ivgi]

For some historical context, it's interesting to note that CTV originally had the constexpr requirement which prevented your CSFS+CTV construct by requiring the CTV hash to come from a literal push preceding the opcode (i.e. not dynamically computed/pre-verified and not from the witness). This was eventually removed because it was deemed an unnecessary safeguard and to simplify the implementation.

Also, it seems that APO alone also enables the same kind of construct as what you're describing. For example, a 'recursive' APO signature spending back to the same address could similarly be used as the basis for a BMM Spacechain.

>  I'd encourage people to try implementing that themselves with their preferred tooling; personally, I found it pretty inconvenient

Here's my go at it using Minsc: https://min.sc/v0.3/#gist=3aa538b384d3aaf4282d1acaf943f608

Example transaction: https://mutinynet.com/tx/9c941dc9c0068eb817b2d18416b174468203c6ad090d8d5735cc9b3732959e39

Note however that what you described is creating 0-fee transactions, with no way to attach additional fees. So this would also need either an anchor output or a CTV template that allows the inclusion of an additional input, which I did not implement in the gist.

> For me, the two components that are annoying is doing complicated taproot script path spends, and calculating CTV hashes

I've been working on improving Minsc to make these kinds of things easier. It supports Taproot, CTV, PSBT, Descriptors, Miniscript, raw Script, BIP32, and more. Much of it is still undocumented (the min.sc website is very much outdated), but I have some examples demonstrating different uses that might be of interest:

- A simple one that involves calculating CTV hashes and a multi-script Taproot tree: https://min.sc/v0.3/#github=examples/ctv-simple.minsc

- More advanced example implementing a CTV vault: https://min.sc/v0.3/#github=examples/ctv-vault.minsc

- Another vault, with key delegation using CTV+CSFS+PAIRCOMMIT+INTERNALKEY: https://min.sc/v0.3/#gist=82d92568c2fcba62203157f8df11354e

- Decaying multisig using PSBT/Miniscript: https://min.sc/v0.3/#github=examples/3of3-into-2of3.minsc

More examples are available on the README: https://github.com/shesek/minsc

Cheers,

shesek

[Antoine Riard]

Hi AJ,

> I don't believe being able to pay for censorship on-chain is any more
> threatening than being able to pay for censorship off-chain.
>
> The bitmex blog post there relies on having a trusted oracle to release
> DLC payments if the target tx wasn't mined. If you have that level of
> trust anyway, then just putting funds in escrow, having miners register
> bolt12 invoices with the oracle, and having the oracle make the payments
> when it's satisfied blocks are sufficiently confirmed has a pretty
> similar risk profile.

I think your reasoning point out exactly what is the "unknown" part of such
tx-withholding risk in bitcoin, and what Gleb never formalized further in the
bitmex blog or his email on costless bribes to miner, as far as I know. Namely
"If you have that level of trust anyway", which I'm understanding that we're
putting here in the shoes of the attacker to evaluate a tx-withhold attacks _cost_.

Generally, when we evaluate a threat model (e.g for internet DDoS) we'll have
few notions like a goal (e.g DoS a Minecraft server), knowledge (e.g know-how
on do to the most efficient DDoS) and capabilities (e.g number of botnet available
to reach a DoS threshold), all 3 elements combined in an attack scenario.

Here, if we reason by analogy to mount a tx-withholding attack as an attacker,
we'll have a goal, i.e censor a LN commitment-tx, a knowledge, i.e txid of the
commitment transaction to be withheld and inner know-how of the LN protocol and
about capabilities i.e DLC oracles available and what can be paid as bribing
on-chain fees (to simplify, let's say on-chain fees == off-chain fees).

Now, in this tx-withholding attack, we can consider that the attacker capabilities
are constituted, among others, of the number of trusted oracles available to release
DLC payment signing the equivalent of a "proof-of-target-UTXO-mining" for a
tx-withholding "bribing" contract.

And that's where your remark on "If you have that level of trust anyway" is
pertinent, I think as this underscores the _accumulation cost_ for an attacker
to build _trust_ in the given DLC oracles that will be used for a tx-withhold
attack. Accumulation or acquisition cost of a Sybil node is a metric considered
in the litterature about Sybil attacks.

Now, and this where is the crux of my interrogation, by extending the expressivity
of bitcoin script and removing the necessity to use an off-chain twist, are we
slashing the _accumulation_ cost for a tx-withholding attack ? An attacker could
from now on just rely on the "blockchain-as-a-judge" paradigm, which is the one
explained in the LN paper iirc and attacks become far more practical.

If you have another methodology to evaluate such tx-censorship risk, I'm of
course curious...The bitmex blog post have also a recension of the literature
in Ethereum analyzing HTLC-attacks, among fews.

> It's <signature> <message> <pubkey> with pubkey at the top of the stack.
https://github.com/bitcoin/bips/blob/master/bip-0348.md

> The same is also true of both Elements' CSFS and Bitcoin-Cash's CHECKDATASIG.

Okay, so this is the latest CSFS draft. I still had in mind the O'Connor's FS'17
paper where it was <signature> <message> <pubkey> with sig-first as a stack order,
as example. Fwiw, what matters is that you can freely sign a <message>, which is
not iself the implicit signature digest, though exact code matters to analyze
opcode composability, of course.

Best,
Antoine
OTS hash: 2e731833f7408e21832605e904e5db0cb21d29a453fbb1cd232eb6c766441f2a

[Anthony Towns]

On Mon, Mar 10, 2025 at 07:14:30AM +0200, Nadav Ivgi wrote:
> For some historical context, it's interesting to note that CTV originally
> had the constexpr requirement which prevented your CSFS+CTV construct by
> requiring the CTV hash to come from a literal push preceding the opcode
> (i.e. not dynamically computed/pre-verified and not from the witness). This
> was eventually removed because it was deemed an unnecessary safeguard and
> to simplify the implementation.

I think the original COSHV implementation had the hash appear a push *after*
the CTV opcode.
https://github.com/JeremyRubin/bips/blob/op-checkoutputshashverify/bip-coshv.mediawiki

> Also, it seems that APO alone also enables the same kind of construct as
> what you're describing. For example, a 'recursive' APO signature spending
> back to the same address could similarly be used as the basis for a BMM
> Spacechain.

With either APO or CTV alone you can do an arbitrarily long chain of commitments
(with APO|SINGLE allowing you to add an input/output for fees and change when
using those commitments), adding CSFS and discarding the CSFS private key allows
you to have a single commitment that can be reused indefinitely.

> Here's my go at it using Minsc:
> https://min.sc/v0.3/#gist=3aa538b384d3aaf4282d1acaf943f608

Nice!

> Note however that what you described is creating 0-fee transactions, with
> no way to attach additional fees.

Yes; I don't think you can do a good job of that on mutinynet with just
CTV/CSFS, since I don't think 0-value anchor outputs are supported. (You
could have CTV commit to two inputs, with the second input's entire
value being burnt to fees, but that's fairly annoying)

Cheers,
aj

[Nadav Ivgi]

On Wed, Mar 12, 2025 at 5:48 AM Anthony Towns <a...@erisian.com.au> wrote:

I think the original COSHV implementation had the hash appear a push *after*
the CTV opcode.
https://github.com/JeremyRubin/bips/blob/op-checkoutputshashverify/bip-coshv.mediawiki

Yes you are of course correct. My memory failed me, should've looked up the BIP :)

With either APO or CTV alone you can do an arbitrarily long chain of commitments,

adding CSFS and discarding the CSFS private key allows
you to have a single commitment that can be reused indefinitely.

 

With APO alone, you can use one of two constructs:

1. Trustless, Finite - using a chain of pre-computed transactions where the signature for the next transaction is committed to in the scriptPubKey. This has similar properties to what you can get with CTV alone.

2. Trusted, Infinite - using a simple non-committed signature spending back to the same address. This has similar properties to your CTV+CSFS construct.

Does adding CSFS enable any additional designs?

I think it's impossible to get Trustless, Infinite short of having full introspection abilities (CAT/TXHASH/Elements-like), right?

With Elements it's pretty straightforward, something like `OP_PUSHCURRENTINPUTINDEX OP_INSPECTINPUTSCRIPTPUBKEY <0> OP_INSPECTOUTPUTSCRIPTPUBKEY OP_ROT OP_EQUALVERIFY OP_EQUAL`

You could have CTV commit to two inputs, with the second input's entire
value being burnt to fees, but that's fairly annoying

Yes, preparing an exact-sized utxo for fees is indeed annoying. However it's not much different from CPFP - an extra tx with the same overall number of inputs/outputs, only around 46vB less efficient[0] (assuming you need change[1]). So at least for some use-cases it's not terrible either.

One important difference is that for a BMM-like bidding system, losing bids would still pay tx fees (and take up chain space) for creating the fee utxo, unlike with CPFP where only the winning bid pays fees. But for other cases, the difference would be negligible.

Of course, the most WU-optimal construct is APO|SINGLE (implying ACP) that you mentioned, where no extra transaction is needed at all.

[0] Assuming you need to use N coins for fees and that change is needed, with an exact-sized fee utxo you have a N-in, 2-out tx to prepare the fee utxo, then a 2-in, 1-out for the main tx - for a total of N+2 inputs and 3 outputs. With CPFP, its 1-in, 2-out for the main tx, then a N+1-in, 1-out tx for the CPFP fee bump - also for a total of N+2 inputs and 3 outputs. However, with CPFP you could use Pay-to-Anchor (while the fee utxo has to be key-encumbered), saving 65WU for one input plus 30vB/120WU for one output, making it more efficient by a total of 185WU/46vB (assuming the fee utxo uses key-path P2TR).

[1] If you happen to have an existing utxo to pay fees with, the 2-input CTV template would be as efficient as it gets, on par with APO|SINGLE. This can be extended to combos of multiple utxos, by adding more CTV tapleaves for different numbers of inputs (at a slight cost of larger control blocks). With large wallets it could be plausible to find such combos.

Cheers,

shesek

[Nadav Ivgi]

> One important difference is that for a BMM-like bidding system [..]. But for other cases, the difference would be negligible.

Actually, thinking about this some more, this could be an issue even without multiple bidders - the tx preparing the fee utxo might pay sufficient fees to get itself confirmed, yet not provide enough fees for the main transaction to confirm too. In this case you'll basically also have 'losing bids' that waste fees/space and will have to make another tx to prepare a larger fee utxo.

So yes, pretty annoying. :< More than I initially thought.

shesek

[Anthony Towns]

On Wed, Mar 12, 2025 at 12:02:27PM +0200, Nadav Ivgi wrote:
> > adding CSFS and discarding the CSFS private key allows
> > you to have a single commitment that can be reused indefinitely.
> With APO alone, you can use one of two constructs:

> 2. Trusted, Infinite - using a simple non-committed signature spending back
> to the same address. This has similar properties to your CTV+CSFS construct.

Right, that's

<01 P> OP_CHECKSIG

as the scriptPubKey, "<sig ANYPREVOUT|ALL>" as the witness, committing to
that scriptPubKey (and an anchor output probably), and after generating
that signature, the private key is discarded.

> Does adding CSFS enable any additional designs?
> I think it's impossible to get Trustless, Infinite short of having full
> introspection abilities (CAT/TXHASH/Elements-like), right?

Direct introspection certainly seems like the easiest approach.
TLUV/OP_VAULT should also get you there I think, despite not providing
full introspection.

> > You could have CTV commit to two inputs, with the second input's entire
> > value being burnt to fees, but that's fairly annoying
> Yes, preparing an exact-sized utxo for fees is indeed annoying. However
> it's not much different from CPFP - an extra tx with the same overall
> number of inputs/outputs, only around 46vB less efficient[0] (assuming you
> need change[1]). So at least for some use-cases it's not terrible either.

CPFP is easier to RBF, so still superior, I think.

> Of course, the most WU-optimal construct is APO|SINGLE (implying ACP) that
> you mentioned, where no extra transaction is needed at all.

A big problem with that is that a griefer can potentially attach large
inputs or outputs to your tx and get their package relayed before yours,
making RBF attempts expensive in high-feerate scenarios, potentially
resulting in nothing being confirmed for an extended period. Could
potentially be solved by TRUC or similar rules (or pure replace-by-feerate
rules), though.

Also, for better or worse, an even more WU-optimal construction is simply
paying miners directly, out of band.

Cheers,
aj

[Antoine Riard]

Hi James,

> As with everything in bitcoin, the chain is insulated from stupidity like that
> by fees, the UTXO model, and block cadence, so what's the problem?

See my novel answer to AJ that by extending bitcoin script, I think you might *actually*
diluting the UTXO model as you can now on cross-inspect the status of an outpoint.
This might amplify tx-withhold style risk for bitcoin contracting protocols, e.g vaults.

> Probably worth noting that CSFS and many advanced introspection opcodes (which allow synthesizing
> CTV) have been live for almost *four years now* on Liquid (https://github.com/ElementsProject/elements/blob/master/doc/tapscript_opcodes.md#new-opcodes-for-additional-functionality).

Live != Deployed use-cases built on top of those advanced introspection opcodes.

I've not verified this by myself, though from what is available from public data
L-BTC is <= to LN today in terms of locked coins. In terms of decentralization,
a federated side-chain is not even like a network of payment channels, where
everyone can join by confirming a 2-of-2 in the chain. So have skilled dev really
try do "wrong things" with L-BTC and all those shiny introspection opcodes ? I'm
not sure, and please accept my skepticism here.

> Sorry to tell you, but given that changing consensus requires soliciting buy-in from a
> wide variety of people across the Bitcoin ecosystem with varying levels of technical
> ability, it is inherently a political process.

No -- Apologies to tell you that science / engineering != politics.

Somehow, if one has taste and familiarity with all postmodernism sociology,
there is well one domain where it completely failed to demonstrate that things
are following a "political process", this is indeed the domain of science. In
science, mathematical or physical truth always prevails, no matter what.

Don't get me wrong -- This doesn't mean I cannot share the opinion that current
consensus changes process is a complete conoundrum. Sadly it might have devolved
in the view of numerous segments of the Bitcoin ecosystem that what matters w.r.t
consensus build-up is who is sitting on a bunch of surreanous github repository
permissions, and not necessarily if the consensus change is benefitful in an
objective fashion for bitcoin, not even if such _consensus change_ might be
required for the long-term flourishing of bitcoin (at least fixing the technical
debt).

There is difference in saying what makes you socially popular to be vetted with
empty github permissions and let you vegetate on them by a weak social consensus
and there is a difference in saying unpopular truth to nurture real changes on
the long-term. I won't open the conversation more on that subject.

Now, back to the more technical analysis of covenant, personally my position
is still the same than it was discussed on Bitcoin Core #28550. The communication
tone was rude, for sure, though I still share the same belief we shouldn't be
careful in designing contracting primitives and be completely $YOLO if said
primitives works well for a second-layer.

On concerning CheckTemplateVerify itself, which is itself simpler than any
other proposed soft-forks due to the non-malleable templating, as I said 18
months ago I'm still open in my free time to review/test a CTV-based vault
(https://github.com/jamesob/simple-ctv-vault), the same way I did for Eltoo-based
LN channels in the past. This is one thing to come with a proof-of-concept
working on a single workstation, this is another thing to have tested it
under real operational guidelines, where dynamic fees and keys ceremonies
have to be deal with.

I don't think CTV is altering the UTXO model significantly, though pleasure
to be proven wrong here. On the other hand CTV is bringing this idea of
immutability in a chain of transactions, that we cannot do currently with
sigs-based covenant. That immutability in theory has consequential value
for anyone building cold wallet or vault, that is clear as once the funding
utxo is confirmed, the UTXO signing keys leaks are void.

Now, I might be the last guy in the ecosystem to think that CTV is not
the promised silver-bullet by its author all over the world to design
and deploy vaults. Fine.

> Beyond that, "Overton window" is an appropriate device in the sense that roasbeef
> is using it because the more substantial a proposed change is, the more time is
> needed by the technical ecosystem to digest it, both in terms of tooling and safety
> analysis. Introducing an entirely new script interpreter on the basis of a Lisp (or
> combinator calculus, or whatever Simplicity's claim is) is indeed far outside of
> that "Overton" window - much farther than tacking on what is essentially just a
> new SIGHASH mode to the existing interpreter.

I do not wish to be overly pessimitic, though I think whatever "just-drop-a-new
script-interpreter" approach is just the wrong design approach, if we do not go
to fix first all the dynamic fees issues grieving all the contracting protocols
and bitcoin second-layers.

Side-note: if you're used with playing with your compiler to target exotic
ISAs, one could build the equivalent for any Lisp interpreter where there is
an IR and in function of the target you compile to current supported bitcoin
opcodes, disabling interpreter syntax that are not (currently) supported or
that cannot be translated. I think this would be already a formal verification
gain for any smart-contract toolchain using it, even if it's a subset.

> Let's not be petty here - it's clear he's talking about LoC within the script
>interpreter, which is a different context than the codebase as a whole. Within
> the context of the interpreter, LoC is indeed a decent heuristic for marginal risk. Of course, nobody's saying it's perfect.

I agree that LoC is a decent heuristic for marginal risk.

Best,
Antoine
OTS hash: 7f00760799b4defd9fe551673a7926c01704274e522d44f3dc8e701b320243de

[moonsettler]

Hi All,

> You could have CTV commit to two inputs, with the second input's entire
> value being burnt to fees, but that's fairly annoying

Sometimes it's annoying, other times it enables new applications, because
it can force the submitter of the transaction to contribute to the output(s).

This could serve different uses like griefing deterrent (immortal statechains)
or collect yield from blind merged mining (surfchains).

BR,
moonsettler

On Wednesday, March 12th, 2025 at 4:48 AM, Anthony Towns <a...@erisian.com.au> wrote:

> On Mon, Mar 10, 2025 at 07:14:30AM +0200, Nadav Ivgi wrote:
>
> > For some historical context, it's interesting to note that CTV originally
> > had the constexpr requirement which prevented your CSFS+CTV construct by
> > requiring the CTV hash to come from a literal push preceding the opcode
> > (i.e. not dynamically computed/pre-verified and not from the witness). This
> > was eventually removed because it was deemed an unnecessary safeguard and
> > to simplify the implementation.
>
>

> I think the original COSHV implementation had the hash appear a push after

> the CTV opcode.
> https://github.com/JeremyRubin/bips/blob/op-checkoutputshashverify/bip-coshv.mediawiki
>
> > Also, it seems that APO alone also enables the same kind of construct as
> > what you're describing. For example, a 'recursive' APO signature spending
> > back to the same address could similarly be used as the basis for a BMM
> > Spacechain.
>
>
> With either APO or CTV alone you can do an arbitrarily long chain of commitments
> (with APO|SINGLE allowing you to add an input/output for fees and change when
> using those commitments), adding CSFS and discarding the CSFS private key allows
> you to have a single commitment that can be reused indefinitely.
>
> > Here's my go at it using Minsc:
> > https://min.sc/v0.3/#gist=3aa538b384d3aaf4282d1acaf943f608
>
>
> Nice!
>
> > Note however that what you described is creating 0-fee transactions, with
> > no way to attach additional fees.
>
>
> Yes; I don't think you can do a good job of that on mutinynet with just
> CTV/CSFS, since I don't think 0-value anchor outputs are supported. (You
> could have CTV commit to two inputs, with the second input's entire
> value being burnt to fees, but that's fairly annoying)
>
> Cheers,
> aj
>

> --
> You received this message because you are subscribed to the Google Groups "Bitcoin Development Mailing List" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to bitcoindev+...@googlegroups.com.

> To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/Z9ED_dez7_UHxjK0%40erisian.com.au.