EIP-7044: Lock voluntary exit domain on capella

[dapplion]

Motivation

For staker UX it's useful that a signed exit is valid in perpetuity. Currently signed voluntary exits are valid for only two forks, due the state only holding current and previous fork.

Description

Compute voluntary exit domain as min(voluntary_exit.epoch, CAPELLA_FORK_EPOCH). This change is backwards compatible if merged before deneb.

I have committed this change in the phase0 spec since "technically" it does not change spec logic, but I can see that leaking CAPELLA_FORK_EPOCH into phase0 directory is not ideal.

[nisdas]

It is too late at this stage to be introducing a consensus change unless it is a security issue.

[dapplion]

It is too late at this stage to be introducing a consensus change unless it is a security issue.

This PR would qualify as a deneb fork change, not capella.

[dapplion]

Moved the change to /deneb directory. See 1c35eb1 for previous iteration committing the change to phase0

[mcdee]

I'm not sure this PR does what it wants to do. If we imagine that we are a few hard forks down the line, let's call it G, and someone submits a voluntary exit signed during Capella, then get_domain() is passed the epoch CAPELLA_FORK_EPOCH and the line:

fork_version = state.fork.previous_version if epoch < state.fork.epoch else state.fork.current_version

will return the fork version for F, so the signature will not verify.

Given that this is going to be a change to Deneb, could we achieve the desired result by replacing

domain = get_domain(state, DOMAIN_VOLUNTARY_EXIT, min(voluntary_exit.epoch, CAPELLA_FORK_EPOCH))

with

domain = compute_domain(DOMAIN_VOLUNTARY_EXIT, CAPELLA_FORK_VERSION, state.genesis_validators_root)

[ralexstokes]

I'd rather move to EL-initiated exits rather than mess w/ the exit via CL operation infrastructure

main reason: we can get much more secure exit authorization strategies via e.g. 4337 smart wallets

[mcdee]

I would suggest that this change is orthogonal to any sort of EL-initiated exit. It's a simple UX boost with only a minimal change to the consensus layer and no alteration to the execution layer. If EL-initiated exits ultimately replace this then fine, but that seems to me to be a fair few hard forks away compared to this change, which could easily slip in to deneb.

[g11tech]

lgtm!

[zah]

This change is related to an earlier proposal for reducing the required trust between staking pool operators and their stakers. From conversations with multiple pool operators, I know this is a real problem they face.

#2666

Admittedly, smart contract initiated exits will be able to solve the problem as well.

[hwwhww]

@djrtwo @ralexstokes

I git merge #3406 and maked EIP7044 change in this branch:deneb-4844-clean...pr3288-comment

I will push it to this PR once #3406 is merged.

[hwwhww]

LGTM

[djrtwo]

looks good! just a couple of minor comments

[djrtwo]

Looks good to me. I added a comment about how it breaks valid/invalid convention