Errata: EIP-2200 should require EIP-1884

[sorpaas]

This PR add clause requires: 1884 for EIP-2200, otherwise the variable name SLOAD_GAS in EIP-2200 can be a point of confusion.

[soc1c]

wouldn't accepting the modification of this final proposal force us to revert the clients?

• core/vm: use dedicated SLOAD gas constant for EIP-2200 go-ethereum#20646
• Add EIP-1884 SLOAD_GAS change to EIP-2200 openethereum/parity-ethereum#11474

See also https://github.com/hyperledger/besu/blob/master/ethereum/core/src/main/java/org/hyperledger/besu/ethereum/mainnet/IstanbulGasCalculator.java

[sorpaas]

Changed this to be a much simpler way -- add clause requires: 1884 so no point of confusion is possible.

[GregTheGreek]

@soc1c To answer your question, no it wouldn't. The way client teams implemented it was that they will take the SLOAD_GAS from EIP-1884, and use that number as the subtraction in the formula when calculating the SSTORE_DIRTY_GAS. Its a case where some clients implemented it as a bundle, and just didn't create a separate variable, since the two were identical in numerical value.

My thoughts:

• If you go with #86437e you haven't cleared up anything, its still confusing and misleading to have SLOAD_GAS a parameter when it is in fact not. I would suggest the original changes you made in 44feb9 that rename SLOAD_GAS to SSTORE_DIRTY_GAS ontop of this parameter.

[sorpaas]

I wouldn't mind either 86437ec or 86437ec + 44feb98 but I disagree for the point that the former doesn't clear up things. The only confusion point is if EIP-2200 is applied without EIP-1884. Applying 2200 without 1884 was never intended and is in fact buggy (SLOAD/SSTORE gas costs are quite messed up in that situation). Users who wish to do that should use EIP-1283 + EIP-1706 instead.

[meowsbits]

Applying 2200 without 1884 was never intended and is in fact buggy

But... (and I hate to be nitpicky here ☹️ ) if the EIP was Finalized without a specification for EIP1884 dependence (and has no mention of it in the document), then... maybe the EIP2200 is just buggy? And the intention would need a new EIP to establish a complete and un-buggy specification?

Or...? I'm not very familiar with the EIP process specs... can one simply amend a Final spec? Is that allowed? 😕

[sorpaas]

@meowsbits See #2388 which is an errata for EIP-1052. Your route works as well, but given there're precedents of errata, I do not plan to create new specifications at this moment.

And just FYI, EIP-1884 is indeed mentioned in the summary section of 2200:

This is an EIP that implements net gas metering. It's a combined version of EIP-1283 and EIP-1706, with a structured definition so as to make it interoperable with other gas changes such as EIP-1884.

I think it's over-exaggerate to call it "buggy". No changes mentioned here will require implementations to change anything or to cause consensus bugs in Ethereum. Rather, it's a "potential of misuse". Not specify 1884 can indeed lead to people from other chains trying to apply this specifications without it, and result in problematic chains.

[meowsbits]

@sorpaas Thanks for the clarification! (And sorry I missed the 1884 reference - my search went down, but not up all the way... oops).

I have one more question - does this specification (EIP2200) actually specify SLOAD gas cost to be 800, or was that an "assumption" (potentially soon-to-be-made-explicit) taken from the context of EIP1884? (I'm still confused with the SLOAD_GAS meaning 😕 ).

... And if SLOAD_GAS (== SSTORE_DIRTY_GAS?) was presumed to be "inherited" from the 1884 context, then maybe the spec can be revised from

- `SLOAD_GAS`: changed from `200` to `800`.

to

- `SLOAD_GAS`: unchanged from `800` (stays at `800`).

[sorpaas]

Thinking again I don't think this is needed at all. The so-called "bug" only affects ETC and has nothing to do with Ethereum or EIPs anyway. I don't think we should spend too much time dealing with unintended use cases like this.

I'm closing this PR for now. We can re-open it if it's really needed in the future.

[shanejonas]

why would you close this??