[bitcoin-dev] An alternative: OP_CAT & OP_CHECKSIGFROMSTACK

[Russell O'Connor]

Hello ZmnSCPxj,

I agree that adding OP_CHECKSIGFROMSTACK doesn't preclude adding shortcuts
such as `SIGHASH_ANYPREVOUT` and `OP_CHECKOUTPUTSHASHVERIFY`, and I agree
we ought to support such operations directly, especially if we see
widespread use of these constructions in practice.

I think it is desirable to add OP_CHECKSIGFROMSTACK for its direct purposes
of enabling oracle verification and discreet log contracts.  Moreover, it
would be better decide if we do or do not want to do this first, because
whether or not we chose to implement a general OP_CHECKSIGFROMSTACK will
influence the design of these other proposals.

For example, if we choose to deploy OP_CHECKSIGFROMSTACK, then the design
of OP_CHECKOUTPUTSHASHVERIFY ought to be simplified to OP_PUSHOUTPUTHASH
and OP_PUSHNUMINPUTS (etc.) because the proposal would no longer be
extending the expressiveness of Bitcoin Script.  And while
OP_CHECKSIGFROMSTACK doesn't directly address whether SIGHASH_ANYPREVOUT
should be with or without a chaperone (as the simulated version with
OP_CHECKSIGFROMSTACK is necessarily chaperoned), we might get an
opportunity to learn if users are willing to take advantage of the
chaperone, or whether they rather bypass it by using a short well-known
pubkey: (e.g.
0x0200000000000000000000003b78ce563f89a0ed9414f5aa28ad0d96d6795f9c63)
and/or similar short signatures if we deploy OP_CHECKSIGFROMSTACK first.

Since most of the "scary" recursive convents are not available with
OP_CHECKSIGFROMSTACK within taproot (without further extensions), the
OP_CHECKSIGFROMSTACK proposal now has quite different consequences than
before.

On Thu, May 23, 2019 at 12:59 PM ZmnSCPxj <ZmnSCPxj at protonmail.com> wrote:

> Good morning Russell,
>
> While I am sympathetic to this argument from first principles, as I
> understand it, it requires that provided witness inputs will become larger,
> compared to "shortcuts" like `SIGHASH_ANYPREVOUT` and
> `OP_CHECKOUTPUTSHASHVERIFY`.
>
> For instance, to simulate `SIGHASH_ANYPREVOUT` with `OP_CAT` and
> `OP_CHECKSIGFROMSTACK`, I would effectively split the unsigned transaction
> into its "inputs" and "outputs" part, concat them and use
> `OP_CHECKSIGFROMSTACK` on the chaperone signature, and also use a normal
> `OP_CHECKSIGVERIFY` on that same chaperone signature, then dup the
> "outputs" part and use `OP_CHECKSIGFROMSTACK` on the "any
> prevout"/"noinput" signature.
> I would effectively give the transaction to itself as part of the witness,
> and further, I would also have to very carefully write the script
> (admittedly the writing of the template could be done once, but it would
> require far more review than simple usages of the "limited" operations like
> `SIGHASH_ANYPREVOUT`).
> So my witness stack would contain two signatures, and a duplicate of the
> transaction itself, plus a very much complicated script, whereas use of
> `SIGHASH_ANYPREVOUT` just requires two signatures and a script not much
> longer than pre-Schnorr multisig scripts.
>
>
> It seems to me desirable, to try to reduce bandwidth consumption on the
> Bitcoin blockchain, including witness data.
> Indeed, I had thought the whole exercise of putting `OP_CHECKSIGFROMSTACK`
> in a federated sidechain (Elements/Liquid) was to try to identify common
> patterns of usage for that opcode, and *then* to propose those common
> patterns as specific "optimized" opcodes as a sort of "jet" for Bitcoin
> itself (but not `OP_CHECKSIGFROMSTACK` itself).
>
> Regards,
> ZmnSCPxj
>
>
> Sent with ProtonMail Secure Email.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxfoundation.org/pipermail/bitcoin-dev/attachments/20190523/1d4853e6/attachment-0001.html>