ECIP-1056 (Agharta): revise draft w/r/t EIP-1702 EVM Account Versioning

[meowsbits]

The previous iteration of this document downplayed the introduction
of making the EVM use versioned contract accounts.

This is a complex and significant change and demands adequate attention in the
document and discussion.

This change (including EIP1702 in this fork's "Meta-ECIP") was originally made here: ethereumclassic/ECIPs@4d227a2

[lialan]

Like the authors stated in the proposal, this change is fully backward compatible and is a necessity for future improvement.

My concern here is the storage it takes to support account versioning. A 256-bit scalar is used which will take quite a lot of disk spaces.

Consider using higher bits of the Nonce field which is most likely sitting there being all zeros. By doing this we will have almost minimal change to the client. Need inputs from original proposer @sorpaas

[sorpaas]

@lialan That's not true. RLP encoding is compact, meaning all zero bits are dropped. This means currently it only takes 1-bit for non-legacy versions, and it's simply impossible to use any nonce bits.

[sorpaas]

I have been suggesting to apply account versioning together with versionless EVM, which has the benefits of making version 1 forward-compatible and thus we will only need 2 versions (compared with one version per six month if only account versioning is applied). You can see my devcon talk for more details on this. If it's proven there're enough interests in ETC community on this, I'd be happy to write more about this.

• ECIP: ECIP-1065: Kunming EVM and Protocol Upgrades ethereumclassic/ECIPs#141
• Devcon slides: https://git.that.world/talks/20191011-compatibility.git/plain/presentation.pdf

Also I want to note that I won't be championing any of the specifications for ETC hard forks. It's purely up to ETC community. If you think we shouldn't use account versioning at all -- you're right.

[soc1c]

Let's discuss this in the call tomorrow.

[lialan]

@sorpaas oh right I did not consider the RLP encoding. In that case it looks good.

[soc1c]

This is obsolete now ethereumclassic/ECIPs#155