>OP_RETURN <sidechain_id> <critical hash>

I think it is redundant here to have the <sidechain id>, we can implicitly assume what the sidechain_id is since we have a fixed set of drivechains. I.e. mining reward is index 0, mimblewimble sidechain is index 1, etc. CryptAxe has specific indexes defined already in his implementation:  https://github.com/drivechain-project/bitcoin/blob/mainchainBMM/src/sidechain.h#L26-L30.

I think it would be wise to include a version byte to allow us to upgrade this commitment structure in the future. Similar to how witness program's are now versioned.

><block height> <sidechain_id> <critical hash> OP_BRIBE_VERIFY

If <block height> is an argument that OP_BRIBE_VERIFY takes, doesn't that mean the mainchain miner has to validate this *is* the actual block height on the sidechain? Does that take the 'blindness' away from BMM?

Overall, I think we need to work on the commitment structure to the coinbase tx. If I understand the current implementation correctly we can have up to 256 OP_RETURNs embedded in the coinbase tx signifying new blocks mined on drivechains.. this seems less than ideal. It might be prudent to make these outputs ANYONECANSPEND, and then have miners spending these outputs to themselves for every block mined.. maybe this doesn't have a benefit over using OP_RETURNs though?

The structure could be something like:
<version> <critical hash>

and then in a subsequent block the miner spends that output to themselves. I will admit I'm not super familiar with how OP_RETURNs work with the UTXO set -- maybe this scheme doesn't have any benefit.

-Chris

On Wed, May 24, 2017 at 3:50 AM, Tier Nolan via bitcoin-dev <bitcoin-dev@lists.linuxfoundation.org> wrote:
On Tue, May 23, 2017 at 3:22 PM, Paul Sztorc <truthcoin@gmail.com> wrote:

If you haven't seen http://www.truthcoin.info/blog/drivechain/ , that is
probably the most human-readable description.

I guess I was looking for the detail you get in the code, but without having to read the code.

My quick reading gives that the sidechain codes (critical hashes) are added when a coinbase is processed.

Any coinbase output that has the form "OP_RETURN <32 byte push>" counts as a potential critical hash.

When the block is processed, the key value pair (hash, block_height) is added to a hash map.

The OP_BRIBE opcode checks that the given hash is in the hash map and replaces the top element on the stack with the pass/fail result.

It doesn't even check that the height matches the current block, though there is a comment that that is a TODO.

I agree with ZmnSCPxj, when updating a nop, you can't change the stack.  It has to fail the script or do nothing.

OP_BRIBE_VERIFY would cause the script to fail if the hash wasn't in the coinbase, or cause a script failure otherwise.

Another concern is that you could have multiple bribes for the same chain in a single coinbase.  That isn't fair and arguably what the sidechain miner is paying for is to get his hash exclusively into the block.

I would suggest that the output is

OP_RETURN <sidechain_id> <critical hash>

Then add the rule that only the first hash with a particular sidechain id actually counts.

This forces the miner to only accept the bribe from 1 miner for each sidechain for each block.  If he tries to accept 2, then only the first one counts.

OP_BRIBE_VERIFY could then operate as follows

<block height> <sidechain_id> <critical hash> OP_BRIBE_VERIFY

This causes the script to fail if
  <block height> does not match the block height, or
  <critical hash> is not the hash for the sidechain with <sidechain_id>, or
  there is no hash for that sidechain in the block's coinbase

If you want reduce the number of drops, you could serialize the info into a single push. 

This has the advantage that a sidechain miner only has to pay if his block is accepted in the next bitcoin block.  Since he is the only miner for that sidechain that gets into the main bitcoin block, he is pretty much guaranteed to form the longest chain.

Without that rule, sidechain miners could end up having to pay even though it doesn't make their chain the longest.

How are these transactions propagated over the network?  For relaying, you could have the rule that the opcode passes as long as <block height> is near the current block height.  Maybe require that they are in the future.  They should be removed from the memory pool once the block height has arrived, so losing miners can re-spend those outputs.

This opcode can be validated without needing to look at other blocks, which is good for validating historical blocks.

I am still looking at the deposit/withdrawal code.

_______________________________________________
bitcoin-dev mailing list
bitcoin-dev@lists.linuxfoundation.org
https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev