Re: [bitcoin-dev] [BIP] OP_CHECKPRIVPUBPAIR

[Anthony Towns <aj@erisian.com.au>]

On Fri, Nov 27, 2015 at 09:02:37AM +0100, Mats Jerratsch via bitcoin-dev wrote:
> <private key> <public key> OP_CHECKPRIVPUBPAIR
> As there are requests for all sort of general crypto operations in
> script, we can also introduce a new general OP_CRYPTO and prepend one
> byte for the operation, so
> 0x01 OP_CRYPTO = OP_CHECKPRIVPUBPAIR
> 0x02-0xff OP_CRYPTO = OP_NOP
> to allow for extension at some later point.

This wouldn't be a softfork -- a single prefixed 0x01 byte would just
push "OP_CRYPTO" onto the stack... If you had OP_CRYPTO look at the top
item on the stack to determine what to do, you could have:

  OP_[0-16] OP_CRYPTO
  OP_PUSHDATA1 [0x11-0xFF] OP_CRYPTO
  OP_PUSHDATA2 [0x00-0xFF] [0x01-0xFF] OP_CRYPTO
  ...

to get 17 different crypto ops in two bytes, and the next 238 in three,
and be arbitrarily expandable from there with multibyte pushes.

> Alternatives:
> In the attached discussion there are some constructions that would
> allow breaking the signature scheme, but they are either very large in
> script language or expensive to calculate.

I think that's good enough to try them out in prototypes though -- and
presumably if they're demonstrably useful in prototypes that's a good
argument for adding a dedicated op code to script?

Are there any other crypto ops that might be worth adding into a BIP for
a check-verify crypto toolkit op like this? The only ones that come to mind
as having practical uses in the near term are:

 Base-point multiply on secp256k1 (ie, CHECKPUBPRIVPAIR)
 Schnorr-signature of transaction with secp256k1 curve (smaller,
   faster, more-anonymous N-of-N multisig)

But perhaps there's also uses for some of:

 General point addition on secp256k1
 General point multiply on secp256k1
 SHA3-256 / SHA2-512 / SHA3-512
 ECDSA/Schnorr signature of value from stack
 ...?

Then again, I gather that if the segregated witness soft-fork turns out
to be plausible, re-enabling/changing/adding *any* sort of opcode could
be done as a soft-fork, not just turning a NOP into CHECK_foo_VERIFY...
So it might be better to wait and see how that goes before putting too
much time into drafting a BIP or similar?

Cheers,
aj