tl;dr
- Merge progress – minor spec updates, engineering overall progress đźš‚
- No progress in client diversity. Be selfish and run a few clients!
Merge updates
First of all, fantastic work by all the engineering teams on the Kintsugi sprint, culminating in the launch of the Kintsugi Merge testnet. It’s amazing to see a total of 3 running clients and 5 consensus clients. 15 Different pairs operating from a unified front.
Kintsugi🍵As the first long-running Merge testnet, , was not without excitement. that much #TestingTheMerge The effort cluttered the testnet with transactions, bad blocks, and other confusing inputs, and introduced several bugs in state transitions, synchronization, etc. We expect to find these bugs in the early testnets, but with each iteration the client becomes more and more stable.
Reboot the kiln 🔥🧱
The team identified a critical issue a few weeks ago. This was a discrepancy. Engine API (How the PoS consensus layer drives the execution layer) Meaning related to how the execution layer client actually works in practice. The bottom line is that in some situations the consensus layer has inadvertently caused unexpected load on the execution layer.
Engineers realized that the two layers could work more harmoniously if the engine API semantics were a little more flexible. This resulted in a subtle but important modification. Engine API and related Destruction Specifications released.
today Kiln specifications🔥🧱 It’s released and engineers are busy rolling out the changes. At the end of this sprint, the team aims to bring a production-ready implementation to a new testnet for public consumption. Keep an eye out for ways to participate.
From there, the team will convert the public testnet to proof-of-stake before preparing the mainnet.
Client Diversity Indicators
Michael Sproul announces a new wave. Client Diversity Metrics It uses his new fingerprinting mechanism. Unfortunately, the client deployment of validation nodes has remained flat over the past six months.
The diversity of consensus layer client implementations allows Ethereum and its users uniquely strong resilience to software errors and attacks. Users achieve some degree of resiliency by using a small number of clients regardless of network configuration, but the network itself achieves resiliency across several key validator deployment thresholds.
For single client:
- Not exceeding 66.6%, errors/bugs from a single client can’t finish
- Not exceeding 50%, errors/bugs in single client fork selection The head of the chain cannot be ruled.
- No single client exceeds 33.3% errors/bugs. Finality cannot be disturbed
Judging by the fingerprinting mechanism, Prysm still holds above ​66.6%.
I want to give a big shout-out to the teams, individuals and communities who take the diversity of their customers seriously (Exhibit A, Exhibit B). Running a small number of clients is not only healthier for the network, it is also safer for individual users’ funds.
Be selfish (reasonable)! Running a few clients 🚀