[spencer-tb]

Status

Documentation/reference prototype for EIP-3298. This draft is not proposed for merge in its current form.

The protocol change is implemented in place in Amsterdam and activated for fixtures through the testing-side Bogota pseudo-fork, targeting Hegota. This work continues @marioevz's original execution-specs implementation, and the final commit preserves him as co-author. No substrate exception or dedicated Bogota spec module is used.

Description

Implements EIP-3298: removes the storage-clearing refund and EIP-3529 quotient cap while retaining the EIP-8038 net-metered STORAGE_WRITE reversal and applying it in full.

• Spec: removes REFUND_STORAGE_CLEAR, its grant and restoration reversal, and the one-fifth settlement cap; retains refund-counter journaling, child-frame accumulation, the write-reversal rule, and pre-refund EIP-7778 block accounting.
• Framework: registers EIP3298, models the zero clear-refund schedule and uncapped settlement through fork APIs, and activates the change through the undeployed Bogota testing pseudo-fork resolved against Amsterdam's spec module.
• Tests: covers removal of clear refunds, surviving and accumulated write reversals, uncapped settlement, calldata-floor ordering, pre-refund block accounting, and reverted-frame journaling. A complete fork-transition test is retained for use once a real post-Amsterdam spec fork exists.
• Benchmarks: keeps the SSTORE dirty-transition benchmarks active and documents their pre-EIP capped clear-refund and post-EIP uncapped write-reversal behavior.
• Existing tests: applies one fork-correct EIP-7778 filter update and parks eight refund-cap or clear-refund-dependent functions across seven files with valid_before("EIP3298"); no tests or vectors are deleted.

Required EIP follow-up

None identified. The pinned EIP revision explicitly specifies the self-bounded write-reversal invariant, refund-counter journaling across reverts, post-refund calldata-floor ordering, and pre-refund EIP-7778 block accounting used by this prototype.

Known limitations

• Because the Bogota pseudo-fork shares Amsterdam's spec module, it cannot provide a genuine pre-/post-fork execution boundary. The complete transition test is therefore skipped until a real post-Amsterdam spec fork exists.
• The eight parked inherited functions require post-EIP refund-boundary scenarios to be re-derived around the surviving write reversal before production fixture coverage is complete.
• The branch is frozen to the shared assessment baseline 43e3cd1; later upstream Amsterdam changes are intentionally not included in this isolated review.
• This is an isolated EIP-3298 assessment prototype, not a combined Hegota implementation branch.

References

• EIP revision used by these tests (548ebfb1): https://github.com/ethereum/EIPs/blob/548ebfb1e987ddb8dc260324e02ee9ec643525a1/EIPS/eip-3298.md
• Tracker: Execution Layer Testing Complexity Assessments For Hegota SFI'd & CFI'd & PFI'd EIPs ethspecs/pm#72
• Complexity assessment: https://github.com/ethspecs/pm/blob/6672c60213099cd1726af93a556963767229ff5a/complexity_assessments/EIPs/EIP-3298.md
• EIP discussion: https://ethereum-magicians.org/t/eip-3298-removal-of-refunds/5430