Reward for P2SH IsStandard() patch.

As you all know keeping the size of the UTXO set small is critical, and more recently we've also had problems with distasteful data being added to the UTXO set. (http://garzikrants.blogspot.se/2013_04_01_archive.html) Gregory Maxwell has an excellent solution to the distasteful data problem in the form of P2SH^2 (http://comments.gmane.org/gmane.comp.bitcoin.devel/1996) and Peter Todd pointed out how we can implement it with the existing P2SH form. We're also going to be implementing some kind of OP_RETURN <data> soon which handles the timestamping and similar use-cases, again without UTXO impact.

Right now the only scriptPubKey form with any significant use is the checksighash. Bare pubkey gets used by the odd miner, and by Deepbit due to their ancient codebase. The former isn't an issue as the miner mines the txout themselves, and the latter shouldn't find updating to be a big deal. OP_CHECKMULTISIG is used by Peter Todd's timestamper, but that can be changed to OP_RETURN without difficulty. However all that will (hopefully!) soon change as hardware wallets and the payment protocol make hardware wallets worthwhile, and we should make sure these protocols take the extra step of using P2SH before we get locked into a bunch OP_CHECKMULTISIG implementations.

We also have the problem that the IsStandard() code accepts up to 120 bytes of junk data as a pubkey, allowing injection of 240 bytes of *spendable* data into the UTXO set with bare OP_CHECKMULTISIG. This capability has to be stopped.

Thus I'm offering a reward of 1BTC for whomever creates a patch to change IsStandard() to accept only P2SH and pubkeyhash in a raw scriptSig, allowing other forms only when used with P2SH. I'm offering a further 1BTC to whomever gets such a patch accepted into mainline. It's a pretty easy patch, so I'm asking that all core-developers (that includes you Peter) hold off for one week to give less experienced developers a crack at it. If for some reason you want to remain anonymous that is ok by me as well provided you assign copyright to me. I do expect unittests. Should be about half a day to a days work.

Long-term we should be using P2SH with an inner OP_CHECKSIG for most addresses as it's a 1 byte savings. Change addresses can have this done first, although bitcoinj support will help so that satoshidice and similar sites can pay to P2SH change. As for multisig's P2SH overhead for a 1-of-2 and 2-of-2 and 3-of-3, is 10%, 8.6% and 6.2% respectively, all pretty minor, especially if you assume the blocksize limit will be raised.

On Sun, Jul 14, 2013 at 07:05:26PM +0000, John Dillon wrote:

> Long-term we should be using P2SH with an inner OP_CHECKSIG for most addresses
> as it's a 1 byte savings. Change addresses can have this done first, although
> bitcoinj support will help so that satoshidice and similar sites can pay to
> P2SH change. As for multisig's P2SH overhead for a 1-of-2 and 2-of-2 and
> 3-of-3, is 10%, 8.6% and 6.2% respectively, all pretty minor, especially if you
> assume the blocksize limit will be raised.

Small comment: the current implementation in the reference client uses a custom script encoder for the UTXO database, which stores every (valid) send-to-pubkey as 33 bytes and every send-to-pubkeyhash or send-to-scripthash as 21 bytes. So for "standard" address payment, there is no storage impact of using P2SH instead.

On Sun, Jul 14, 2013 at 7:28 PM, Pieter Wuille <pieter.wuille@gmail•com> wrote:

> On Sun, Jul 14, 2013 at 07:05:26PM +0000, John Dillon wrote:
>> Long-term we should be using P2SH with an inner OP_CHECKSIG for most addresses
>> as it's a 1 byte savings. Change addresses can have this done first, although
>> bitcoinj support will help so that satoshidice and similar sites can pay to
>> P2SH change. As for multisig's P2SH overhead for a 1-of-2 and 2-of-2 and
>> 3-of-3, is 10%, 8.6% and 6.2% respectively, all pretty minor, especially if you
>> assume the blocksize limit will be raised.
>
> Small comment: the current implementation in the reference client uses a custom
> script encoder for the UTXO database, which stores every (valid) send-to-pubkey
> as 33 bytes and every send-to-pubkeyhash or send-to-scripthash as 21 bytes.
> So for "standard" address payment, there is no storage impact of using P2SH
> instead.

By "impact" I am referring to the impact on transaction size and thus blockchain space and fees, not UTXO size as stored by nodes themselves. Specifically take the size of the txout and txin and compare the version using P2SH to the equivalent version not using it to get my numbers.

Anyway, given how much uncompressed keys are still used obviously fee pressure isn't even close to getting people to create efficient transactions.