Summary: Covenants Support - Bitcoin Wiki

[/dev /fd0]

Hi Bitcoin Developers,

I had shared covenants support wiki page link here on [mailing list][1] in the last week of November 2024. Multiple changes were made based on the feedback:

- Removed 'community support' from 'No'. Rephrased definitions for 'Prefer' and 'Evaluating'.
- Added LNHANCE category for a combination of opcodes.
- Added links for BIP drafts and a column for 'rationale'.
- Created a separate table for evaluations without a rationale.

Murch and Gloria shared their feedback in the bitcoin optech [podcast 333][2]. I have started working on a [page][3] that lists use cases, prototype links and primitives used. We can still add more use cases in it. This list does not include use cases enabled by [OP_CHECKSIGFROMSTACK][4] alone or in combination with other opcodes like [LN SYMMETRY][5].

I had verified each entry to avoid spam and fake evaluations. Rearden was assigned moderator permissions on 8 December 2024 by Theymos to help me with the moderations. Some edits have been approved by other moderators.

Some insights from the table that could help developers working on different covenant proposals:

1. Multiple ways to achieve LN symmetry were discovered. SIGHASH_APO lacks interest among developers, contrary to the belief prior to this exercise.
2. OP_CHECKSIGFROMSTACK has unanimous support and is a part of multiple covenant proposals.
3. OP_PAIRCOMMIT, OP_INTERNALKEY and OP_CHECKCONTRACTVERIFY are not reviewed by enough developers. OP_PARCOMMIT seems to be controversial at this moment.

Objections:

```
======================
SIGHASH_APO
======================

LN SYMMETRY is possible with combination of a few opcodes which is more efficient. Its not the best option for covenants and cannot be used to improve Ark. Some developers prefer opcodes and not sighash flags.

Seems to be the result of an attempt to fix signatures to make them work for a specific use-case, but the end-result is hard-to-reason (for me) and not flexible. In general, SIGHASH flags are an encoding of specific predicates on the transaction, and I think the Script is better suited to carry the predicate. There is no interesting SIGHASH flag that couldn't be functionally simulated by introspection + CHECKSIGFROMSTACK (or other Script-based approaches), and that seems to me a much cleaner and ergonomic way to achieve the same goals.

======================
OP_TXHASH
======================

More expressive, many flag configurations, untested and undesirable use cases. Unaddressed comments in the BIP and the delay doesn't make sense because OP_CHECKTEMPLATEVERIFY can be upgraded later to achieve the same thing. Makes hash caching complex, potentially opening up DoS vectors or quadratic sighash.

Most templates you'd obtain with various combinations of parameters are meaningless. It implements state-carrying UTXOs in a very dirty way: adding additional inputs/outputs with no other meaning than "storing some state". This is ugly, inefficient, and bloats the UTXO set - and it definitely will happen if TXHASH is enabled without also enabling a clean way to carry state.

Follow up with an upgrade to OP_CHECKTEMPLATEVERIFY can fine tune it to what people are actually using covenants for, instead of prematurely optimizing everything with no data.

======================
OP_VAULT
======================

No demand for vaults. Customized for a specific use case.

======================
OP_CAT
======================

Can be used for various complex scripts including undesirable use cases (DEX, AMM and Hashrate Escrow). Enables granular transaction introspection through abuse of schnorr signatures and OP_CHECKSIG. Can be used for interesting use cases but alone does it poorly and inefficiently.

People can and will litter the chain with inefficient/ugly Scripts if activated alone. Since it happens to enable generic introspection by accident, and therefore an ugly version of state-carrying UTXOs, it shouldn't be enabled without more ergonomic opcodes for those use cases.

======================
OP_INTERNALKEY
======================

There are 3 'No' in the table, I couldn't find anything relevant in the rationales.

======================
OP_PAIRCOMMIT
======================

Adds unnecessary complexity, redundant if OP_CAT is activated in future and added for specific use case. LN SYMMETRY is possible without this opcode. It does not compose with anything that involves transaction introspection due to its specified tagged hash. Some developers prefer OP_CAT.

Not convinced it is impossible that there is no way to simulate CAT with PAIRCOMMIT, nor confident how much less powerful PAIRCOMMIT is than CAT.

======================
OP_CHECKTEMPLATEVERIFY
======================

Limited in scope and not recursive.
```

I have tried my best to remain unbiased in writing this summary and approving edits. There are a few things that I want to share and it could be a result of the aggressive marketing:

- A spammer had edited the table to remove all evaluations except in favor of OP_CAT and it was rejected.
- [Rationale][6] added by Aaron (sCrypt) does not mention anything about other opcodes and SIGHASH_APO. It is only focused on OP_CAT however evaluations exist in the table.
- I [requested][7] Udev (CatSwap) to add details about evaluation of other opcodes and SIGHASH_APO.
- Last [edit][8] by Roujiamo (bitdollar) has a rationale with incorrect signet stats and seems to be rephrased version of another rationale. Evaluation had 'weak' for OP_CTV before adding the rationale.
- An edit with duplicate rationale (in support of OP_CAT) was rejected because sharing the link for a rationale submitted by other developer adds no value in the table.

Evaluations without a rationale have some 'No' in different cells. Although none of them are backed by a rationale so cannot be considered for consensus discussion. The table is still updated regularly so you may see some of them with a rationale in 2025. Any suggestions to help achieve technical consensus are most welcome.

What's next?

- More rationales in the table
- Discuss objections on mailing list (if any)
- Workshops
- Add a table for economic nodes and their opinion
- Build activation client and discuss parameters

Finally, I would thank all the developers who added their evaluations in the table and everyone who shared updates on twitter. It was a coordinated effort to reach some technical consensus. You can read all the rationales in detail to understand different perspectives and reasons to support a combination of opcodes over others.

[1]: https://groups.google.com/g/bitcoindev/c/fdxkE1Al4TI/m/CeEuls2IAQAJ
[2]: https://bitcoinops.org/en/podcast/2024/12/17/
[3]: https://en.bitcoin.it/wiki/Covenants_Uses
[4]: https://github.com/bitcoin/bips/blob/master/bip-0348.md
[5]: https://gist.github.com/Ademan/4a14614fa850511d63a5b2a9b5104cb7
[6]: https://gist.github.com/gitzhou/dc92c41db1987db16fe665c26bc56dd9
[7]: https://gist.github.com/udevswap/b768d20d62549922b9e72428ef9eb608?permalink_comment_id=5359072#gistcomment-5359072
[8]: https://en.bitcoin.it/w/index.php?title=Covenants_support&diff=prev&oldid=70520

/dev/fd0
floppy disk guy

[moonsettler]

Hi All,

One thing I would like to make clear before people get the wrong idea and think this is some form of voting, OP_INTERNALKEY and OP_PARCOMMIT is part of LNhance and will be part of the activation client we release soon. The only way to change that is to demonstrate actual harm. You liking something else more, is your problem. What you can do about it, is write your activation client and try to gain consensus on that. There are plenty of version bits available. Replacing PAIRCOMMIT with CAT would be really easy, but while CAT is indeed very popular and has a wide support base it is also strongly opposed by many who did not choose to participate. I'm not convinced that this table represents actual developer, let alone ecosystem consensus. If you decide you want to run an alternative activation effort with CAT instead of PAIRCOMMIT feel free to fork our repo!

======================
OP_PARCOMMIT
======================

> OP_PARCOMMIT seems to be controversial at this moment.

I strongly disagree. In my book that's not what controversial means. Literally nobody managed to come up with a single use case anyone worth noting objects to for PAIRCOMMIT. Also inclined to ignore the "No" from those that prefer CAT as plain trolling. This BIP is young, there is a clear correlation between the age of the proposals and their support with the sole exception of APO.

> Adds unnecessary complexity

That's a subjective value judgement it enables something that was no possible before which is interacting with Merkle trees and multi-element commitments in script. PAIRCOMMIT is not significantly more complicated than CAT, and in a lot of actual use cases CAT was desired for it's more complex and resource intensive to safely use CAT than PAIRCOMMIT due to witness malleability.

> Not convinced it is impossible that there is no way to simulate CAT with PAIRCOMMIT, nor confident how much less powerful PAIRCOMMIT is than CAT.

This is sufficiently addressed in the BIP.

======================
OP_VAULT
======================

> No demand for vaults.

It's safe to completely ignore that "argument".

BR,
moonsettler

> --
> 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/38a6f252-afe9-4155-a341-11a42a9a9007n%40googlegroups.com.

[/dev /fd0]

Hi monnsettler,

I have never mentioned this as voting in any of my posts. You are free to build activation clients that nobody will run. 

/dev/fd0

floppy disk guy

[Ethan Heilman]

One of the CAT authors here

> > [PAIR_COMMIT] Adds unnecessary complexity

> That's a subjective value judgement it enables something that was no possible before which is interacting with Merkle trees and multi-element commitments in script. PAIRCOMMIT is not significantly more complicated than CAT, and in a lot of actual use cases CAT was desired for it's more complex and resource intensive to safely use CAT than PAIRCOMMIT due to witness malleability.

PAIR_COMMIT (BIP-442) for all intents and purposes is as simple in
implementation at CAT (BIP-347). I have no technical objection to
PAIRCOMMIT and it provides much needed functionality.

My primary concern is not PAIRCOMMIT itself, but the rationale for PAIRCOMMIT.

The rationale for PAIRCOMMIT rests on the assumption that the Bitcoin
community does not want the expressiveness of CAT. If we assume this
is the case, then we should be very careful PAIRCOMMIT does not enable
this expressiveness as well. On the other hand, if the Bitcoin
community does want the expressiveness of CAT, then we should merge
CAT. PAIRCOMMIT is well designed to be less expressive than CAT and it
is likely that you can not simulate CAT with PAIRCOMMIT. That said, I
am not convinced it is impossible that there is no way to simulate CAT
with PAIRCOMMIT, nor I do feel confident that I know how much less

powerful PAIRCOMMIT is than CAT.

Playing devil's advocate for a second, if I was opposed to CAT on
grounds that we should limit expressiveness I would want to really
understand the limits of PAIRCOMMIT. For instance can you do arbitrary
computation by building STARKs with PAIRCOMMIT merkle trees? If not,
why not?

That said, I have not heard any argument against PAIRCOMMIT from those
against CAT, so perhaps they are comfortable with it.

Since I am in favor of CAT, I am also in favor of PAIRCOMMIT.

> To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/rp07_AsZrGYA3kFwZweIhzZVonmcuQktAz9r51MgKvrG101_T9NBTTMCFK_q3bMzIH0-QzfFtzC6uJGEKOIMi6Hl6qwbDtMWXXV2frBWXac%3D%40protonmail.com.

[/dev /fd0]

Hi Ethan,

OP_CAT is not proposed as an opcode to enable LN SYMMETRY. Whereas OP_PAIRCOMMIT is a part of LNHANCE.

In this context, OP_PAIRCOMMIT adds unnecessary complexity because LN SYMMETRY can be achieved with other opcodes.

Note: The objections shared in this thread are a summarised version of all the rationales and not my person opinion.

/dev/fd0

floppy disk guy

To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/CAEM%3Dy%2BV9Gu0n7pLv1d%2BK1HfaFsB3kXg-LbtppyZG0xjAa7DBaA%40mail.gmail.com.

[moonsettler]

Hi Floppy,

Neither INTERNALKEY nor PAIRCOMMIT enables LN-Symmetry, LNhance does. They make it more efficient, and they also help other contracts.
Among them: Resumeable LN channels, Multi-party LN channels, Vaults, etc.

Main benefit for the network: we can reduce the number of SigOps on-chain which benefits everyone that runs a validating node by making it more economic to use a single signature for multiple elements instead of using something like the ReKey technique.

Calling it "unnecessary complexity" is not a valid technical observation in any shape or form. It would provide optimization for many contracts and use cases even if we had CAT. I explained this to you in private first, yet you keep insisting on this completely invalid objection.

BR,
moonsettler

PS: I largely agree with everything Ethan said.

Sent with Proton Mail secure email.

[/dev /fd0]

Hi moonsettler,

>  Neither INTERNALKEY nor PAIRCOMMIT enables LN-Symmetry, LNhance does. They make it more efficient, and they also help other contracts.
Among them: Resumeable LN channels, Multi-party LN channels, Vaults, etc.

I am aware of this and have used the comparison table in my [rationale][1]. LNHANCE is not an opcode. CTV and CSFS enable LN SYMMETRY.

>  Calling it "unnecessary complexity" is not a valid technical observation in any shape or form. It would provide optimization for many contracts and use cases even if we had CAT. I explained this to you in private first, yet you keep insisting on this completely invalid objection.

It is a valid objection and I find it disappointing that you think people will change their opinion about an opcode if you build [activation client][2] or a different table. If you read all the rationales, its not just me who finds this opcode irrelevant. Please respect the developers who shared their evaluations in the table even if you disagree with them. If you cannot appreciate efforts to review proposals and reach technical consensus, at least avoid calling reviews/evaluations as "voting". 

"Unnecessary" because LN symmetry (efficient) is possible without it. "Complexity" is introduced because it will be used for everything possible with it in bitcoin script and not just the use cases described in your email.

/dev/fd0
floppy disk guy 

[1]: https://gitlab.com/-/snippets/4777553
[2]: https://groups.google.com/g/bitcoindev/c/QOnpM4Ixmms/m/eodYc71lDAAJ

[moonsettler]

Hi FLoppy,

Of course I appreciate thoughtful evaluations.

But my point stands, I encourage everyone to look at this table as means to engage in discussion not some indicator of "consensus" by any means.

And I emphasized this, because there was a weird perception emerging, and even your own summary had this feel to it.

BR,
moonsettler

Sent with Proton Mail secure email.

> To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/c411dee9-1937-42fd-bc66-d347ddbff506n%40googlegroups.com.