BIP for OP_INTERNALKEY

Hello list,

I'm currently failing to find the original reference discussion for adding OP_INTERNALKEY to tapscript. I believe it was in the context of the SIGHASH_ANYPREVOUT proposal which opted instead to access the internalkey as a special key with value `0x01`. Regardless, here[0] is a BIP for adding OP_INTERNALKEY to tapscript to allow access to the taproot internal key. As noted below, this helps certain classes of script come closer to matching segwitv0 in byte efficiency, which can be particularly useful for protocols such as Lightning where the same signers may need to sign a script path in some cases, and a key path in others.

------------ ## Abstract

This BIP describes a new tapscript opcode (`OP_INTERNALKEY`) which pushes the taproot internal key to the stack.

## Specification

When verifying taproot script spends having leaf version `0xc0` (as defined in [BIP 342]), `OP_INTERNALKEY` replaces `OP_SUCCESS203` (0xcb). `OP_INTERNALKEY` pushes the taproot internal key, as defined in [BIP 341], to the stack.

## Motivation

### Key spend with additional conditions

When building taproot outputs, especially those secured by an aggregate key representing more than one signer, the parties may wish to collaborate on signing with the taproot internal key, but only with additional script restrictions. In this case, `OP_INTERNALKEY` saves 8 vBytes.

### Mitigated control block overhead for scripts using hash locks

In cases where script path spending is not desired, the internal key may be set to a NUMS point whose bytes would otherwise be required in a tapscript. This could be used with any hash locked transaction, for example, to save 8 vBytes.

Note: The internal key must be the X coordinate of a point on the SECP256K1 curve, so any such hash must be checked and modified until it is such an X coordinate. This will typically take approximately 2 attempts.

## Reference Implementation

A reference implementation is provided here:

https://github.com/bitcoin/bitcoin/pull/29269

## Backward Compatibility

By constraining the behavior of an OP_SUCCESS opcode, deployment of the BIP can be done in a backwards compatible, soft-fork manner. If anyone were to rely on the OP_SUCCESS behavior of `OP_SUCCESS203`, `OP_INTERNALKEY` would invalidate their spend.

## Deployment

TBD

## Copyright

This document is licensed under the 3-clause BSD license.

[BIP 341]: https://github.com/bitcoin/bips/blob/master/bip-0341.mediawiki

[BIP 342]: https://github.com/bitcoin/bips/blob/master/bip-0342.mediawiki

------------

[0]: https://github.com/bitcoin/bips/pull/1534

I wonder, what is the reason to pick 0xcb as OP_INTERNALKEY, when we had pseudo-words OP_PUBKEY (assigned to 0xfe) and OP_PUBKEYHASH (assigned to 0xfd). Of course, those opcodes are invalid in the non-Taproot scripts, but they were intended to be used in cases like "OP_PUBKEY OP_CHECKSIG", and "OP_DUP OP_HASH160 OP_PUBKEYHASH OP_EQUALVERIFY OP_CHECKSIG". So, I guess we could reuse those opcodes, and make it a general rule, that OP_PUBKEY picks the next key from the stack (which in this specific case will give us the internal key, but I can imagine leaving some room for extending it in the future, with scripts like "OP_PUBKEY OP_CHECKSIGVERIFY OP_PUBKEY OP_CHECKSIG", where each of those OP_PUBKEYs will give us a different key, and act somewhat like OP_FROMALTSTACK, but working on a separate stack of public keys instead, where the first key of that stack would be the internal Taproot key).

czwartek, 25 kwietnia 2024 o 12:41:04 UTC+2 Brandon Black napisał(a):

> Hello list,
>
> I'm currently failing to find the original reference discussion for
> adding OP_INTERNALKEY to tapscript. I believe it was in the context of
> the SIGHASH_ANYPREVOUT proposal which opted instead to access the
> internalkey as a special key with value `0x01`. Regardless, here[0] is a
> BIP for adding OP_INTERNALKEY to tapscript to allow access to the
> taproot internal key. As noted below, this helps certain classes of
> script come closer to matching segwitv0 in byte efficiency, which can be
> particularly useful for protocols such as Lightning where the same
> signers may need to sign a script path in some cases, and a key path in
> others.
>
> ------------
> ## Abstract
>
> This BIP describes a new tapscript opcode (`OP_INTERNALKEY`) which
> pushes the taproot internal key to the stack.
>
> ## Specification
>
> When verifying taproot script spends having leaf version `0xc0` (as
> defined in [BIP 342]), `OP_INTERNALKEY` replaces `OP_SUCCESS203` (0xcb).
> `OP_INTERNALKEY` pushes the taproot internal key, as defined in [BIP
> 341], to the stack.
>
> ## Motivation
>
> ### Key spend with additional conditions
>
> When building taproot outputs, especially those secured by an aggregate
> key representing more than one signer, the parties may wish to
> collaborate on signing with the taproot internal key, but only with
> additional script restrictions. In this case, `OP_INTERNALKEY` saves 8
> vBytes.
>
> ### Mitigated control block overhead for scripts using hash locks
>
> In cases where script path spending is not desired, the internal key may
> be set to a NUMS point whose bytes would otherwise be required in a
> tapscript. This could be used with any hash locked transaction, for
> example, to save 8 vBytes.
>
> Note: The internal key must be the X coordinate of a point on the
> SECP256K1 curve, so any such hash must be checked and modified until it
> is such an X coordinate. This will typically take approximately 2
> attempts.
>
> ## Reference Implementation
>
> A reference implementation is provided here:
>
> https://github.com/bitcoin/bitcoin/pull/29269
>
> ## Backward Compatibility
>
> By constraining the behavior of an OP_SUCCESS opcode, deployment of the
> BIP can be done in a backwards compatible, soft-fork manner. If anyone
> were to rely on the OP_SUCCESS behavior of `OP_SUCCESS203`,
> `OP_INTERNALKEY` would invalidate their spend.
>
> ## Deployment
>
> TBD
>
> ## Copyright
>
> This document is licensed under the 3-clause BSD license.
>
> [BIP 341]: https://github.com/bitcoin/bips/blob/master/bip-0341.mediawiki
>
> [BIP 342]: https://github.com/bitcoin/bips/blob/master/bip-0342.mediawiki
>
> ------------
>
> [0]: https://github.com/bitcoin/bips/pull/1534
>
> --
> --Brandon
>

Hi Garlo, thanks for your thoughts!

On 2024-04-26 (Fri) at 01:09:24 -0700, Garlo Nicon wrote:

> I wonder, what is the reason to pick 0xcb as OP_INTERNALKEY,

The choice of 0xcb was fairly arbitrary. An early draft had used another OP_SUCCESS but we moved to 0xcb to avoid conflict with BIP345.

> when we had pseudo-words OP_PUBKEY (assigned to 0xfe) and
> OP_PUBKEYHASH (assigned to 0xfd).

Given that OP_PUBKEY was never an actual opcode, reusing that pseudo-word for this purpose seems more confusing than beneficial.

> So, I guess we could reuse those opcodes, and make it a general rule,
> that OP_PUBKEY picks the next key from the stack (which in this
> specific case will give us the internal key, but I can imagine leaving
> some room for extending it in the future, with scripts like "OP_PUBKEY
> OP_CHECKSIGVERIFY OP_PUBKEY OP_CHECKSIG", where each of those
> OP_PUBKEYs will give us a different key, and act somewhat like
> OP_FROMALTSTACK, but working on a separate stack of public keys
> instead, where the first key of that stack would be the internal
> Taproot key).

In terms of creating a stack of pubkeys which can be accessed sequentially in script, the only two keys I can see that being used for are the taproot internal key and the taproot external key. Any other keys will already explicitly appear in the script being validated. An alternative design could be OP_TAPKEY which can be called up to two times in script, first pushing the internal key and then pushing the external key. My inclination here is that if a case can be made for accessing both of these keys, 2 separate opcodes that repeatedly push the same key is more ergonomic for script developers than 1 opcode that pushes internal, then external, (then fails?). I'd even opt for OP_TAPKEYS that pushes both keys and the script can then DROP or NIP away the one they don't need over a stack of keys.

Best,

--Brandon