Consensus-layer Call 107

[djrtwo]

Consensus-layer Call 107 Agenda

prev: call 106

Meeting Date/Time: Thursday 2023/4/20 at 14:00 UTC
Meeting Duration: 1.5 hours
Livestream

1. Capella
   • fork retro
2. Deneb
   • Potential blockers for 4844-devnet-5
   • PR to merge getPayloadV3 & getBlobsBundleV1
   • Blob re-orgs & newPayload(context)
   • under consideration:
     • add EIP-4788 feature consensus-specs#3319
     • EIP6914 - Reuse indexes with full sweep consensus-specs#3307
     • historic summaries backfill
     • Not allow slashed validator to become a proposer consensus-specs#3175
   • Engine api
     • State that payloadId should be unique for each PayloadAttributes instance execution-apis#401
3. Research, spec, etc
   • Consideration for RPC to validate beacon header
   • Attnet revamp: Subnet backbone structure based on beacon nodes consensus-specs#3312
   • Verge number field in json in engine api
4. Open Discussion/Closing Remark

[timbeiko]

Two things that came up on the 4844 call today which we should discuss on ACDC:

1. Potential blockers for 4844-devnet-5
2. PR to merge getPayloadV3 & getBlobsBundleV1
3. Blob re-orgs & newPayload(context)

[mkalinin]

I will likely not make this call, couple of Engine API things:

• State that payloadId should be unique for each PayloadAttributes instance execution-apis#401
  • if given PayloadAttributes does not match attributes of the existing build process a client MUST start a new one; every new build process MUST be uniquely identified by returned payloadId
  • client SHOULD NOT restart existing build process if given PayloadAttributes matches
• Merge getPayloadV3 and getBlobsBundleV1 execution-apis#402

Aiming to merge both PRs on Friday, last call for objections.

[mkalinin]

Another thing to consider for Deneb:

• Not allow slashed validator to become a proposer consensus-specs#3175 -- it started as a small change to compute_proposer_index function, but it appears that the complexity of this fix is a bit bigger as with the change to compute_proposer_index we can't use get_beacon_proposer_index inside of process_operations anymore as any slashing would change the output of this function. So, slash_validator, process_attestation and process_sync_aggregate undergoes small changes as well.

The probability of missing proposal because of a single slashing is trivial. While it increases significantly in the case of correlated slashing, for instance, 10% of slashed validators results in ~1 missed proposal per epoch on average.

[gballet]

If there's time, I would like to have a <10 minute conversation about the Verge, namely for having a number field in the JSON of the execution payload.

Context: In the spec, the SuffixDiff object has a Suffix byte. It seems that the agreement for the API so far, is to always use hex strings for any number. Given how often this field is going to be repeated, however, it would be wasteful to use a hex string here.

Discussion point: We would like to use a regular JSON number for this field, instead of a hex string.

[mkalinin]

Discussion point: We would like to use a regular JSON number for this field, instead of a hex string.

We can use JSON integer type for this field, implementation complexity should be trivial.