Here’s a half thought alternative to EIP 8198 that has quite some interesting advantages with minor tradeofs. Let’s get straight to the point.

  • Slots remain 12 seconds.
  • Proposer commits to a builder and a blockhash as in Gloas.
  • The builder reveals 3 blocks, one each 4 seconds.
  • Each block commits to the next block hash so blocks are still delayed 4 seconds as today, but transaction inclusion time reduces to a max of 4 seconds.
  • PTC members vote with a bitlist for the timeliness of each of the payloads.
  • The next proposer can reorg all payloads starting from the first one that there was a quorum that the payload was late/unavailable.

The bad things:

  • Builders can equivocate, make a mess of split views in forkchoice. Workable, if PTC does not reach quorum the next proposer will just reorg all payload equivocations. These equivocations can be made to not affect finality, they all occurr withing the already anchored block by the consensus layer.
  • Builders have many consecutive blocks (3 in the scheme above) and could potentially do multiple slot MEV, but they are forced to reveal the payloads in time and are constrainged by the next hash commitment.
  • The biggest drawback is that transaction inclusion does not get a beacon committee vote every block, not even a PTC vote every block, votes are only coming per slot. This forces latency bound traders to track for payload equivocations for example. So this is a weak preconfirmation but it’s visibly signed on chain. Traders that need the guarantee of a full beacon vote would need to wait a slot.

The good things:

  • Trivially implementable. Does not require any new machinery, we have everything in place and it’s just adding a couple of bits to the PTC attestations.
  • The forkchoice guarantees are what we are moving anyway in new consensus with small committees for the fast vote, notice that anyway we retain the full beacon committee every slot to anchor the beacon chain.
  • We get to immediately pump up transaction inclusion latency without affecting hardware, builders are already supposed to be beefier than attesters. We get to have 6 – 4 seconds blocks in a realistic frame.
  • The most important one: We keep the useless part of the slot exactly as today. Today the consensus block propagation and attestation propagation (because builders will wait to reveal) is dead time in the slot. If we reduce the slot time this dead time per block becomes a higher fraction of the useful time in the slot. With this proposal the dead time remains the same, about 1/3 of the slot in Gloas time, while the chain throughput doubles or triples. We don’t need to pay with extra dead time and thus we can keep the same Gas limit per second that we can achieve in Gloas.