Unresolved TODOs from PR #4828

[kevaundray]

We merged in #4828 and decided to work on the unresolved items iteratively.

Here are the review threads that I think we did not reach a conclusion on:

Prover Registration and Whitelist Mechanism

See review thread here

There was a conversation around what the whitelist should be and the mechanism for registering and updating provers.

• Suggestion to use a BPO-like (Prover Parameter Only) config loaded as node configuration
• Main concern for this mechanism was around DoS if execution_proof topic is spammed with invalid proofs
• Reference doc: https://hackmd.io/@kevaundray/SJjGnqz8bl

Stateless validation mode

See review thread here

Discussion around whether the specs should support running without an EL

• Original purpose of optional proofs was to be able to run without an EL. However, we may want to allow both ExecutionEngine and ProofEngine

Reuse ExecutionPayloadHeader

@jihoonsong notes here that ExecutionPayloadHeader is constructed twice with the same fields and could be reused from the verify_new_payload_request_header call

proof_gen_id vs proof_type Naming

See review thread here

• Clarification needed on what "proof instance" means

Async Proof delivery and failure handling

See review thread here

• Suggestion to add a note about fallback behavior

Fork Choice and Proof Verification Timing

See review thread here

Discussion about when verify_new_payload_request_header runs.

K-out-of-N Policy Documentation

@jihoonsong noted that the k-out-of-n proof policy should be described somewhere in the spec.

[kevaundray]

cc @frisitano for visibility

[kevaundray]

On the prover registration mechanism, the strategy discussed on discord (that is implemented in the PR) is to use the validator set initially. There is a discussion on discord between @MarcoPolo and @frisitano on adding a reputation system in order to mitigate a potential DoS attack, that was summarised here: https://discord.com/channels/595666850260713488/1464245495294722089/1465329442636169421