I can’t believe it’s already time for the final development update! Time flew since the start of the program, and it’s great to look back and see all that I’ve accomplished thus far. But I’ll go into more detail about this in my final development update.

This week, I started on my final presentation and also began analyzed an interesting family of protocols, Total Order Broadcasts with graded agreements. These were previously only able to be used when participation was static, like in traditional distributed consensus systems, but a paper by Momose and Ren created the concept of dynamic quorums by using the time-shifted quorum technique. This basically concedes that two nodes may not agree on the current level of participation, but can agree on the participation level of a prior time, and uses this as a basis to create transferable graded agreements in the dynamic participation model.

The original attempt to create a consensus mechanism out of these, MR, had issues with a real-world implementation, primarily due to its high latency and lack of support in asynchrony. After a few more iterations (MMR, MMR2), D’Amato and Zanolini, authors of RLMD-GHOST, released a paper titled [2310.11331] Streamlining Sleepy Consensus: Total-Order Broadcast with Single-Vote Decisions in the Sleepy Model](https://arxiv.org/abs/2310.11331).

This paper uses a similar method to RLMD-GHOST, increasing the vote expiry period, to allow for bounded periods of asynchrony. It also uses some clever tricks to reduce the latency and only require one graded agreement per view/slot.

I’m currently reading through additional notes about this protocol to see how it could be the eventual replacement for LMD-GHOST and allow for single-slot finality to be brought to Ethereum. At the moment, my understanding is that we can introduce a similar “fast confirm” phase that RLMD-GHOST has to reduce latency in ideal conditions. We should also be able to use an acknowledgment subnet, along with an additional slashing condition, to induce finality with opt-in recognition.

In the following days, I’m going to conclude my research into this protocol, create my final update, and plan my next steps going forward. At the moment, the most likely option is for me to formally write up some more info about this protocol’s SSF promise, as suggested by its authors. I’m also going to finish my presentation with this new information (it’s difficult to trim 4 months’ research into 5 minutes!), and prepare to meet the other EPF fellows and present to them in Istanbul!