ไม่สามารถเล่นวิดีโอนี้
ขออภัยในความไม่สะดวก
Michael Sproul - Update on client diversity and blockprint
ฝัง
- เผยแพร่เมื่อ 18 ส.ค. 2024
- Michael: / sproulm_
----------
AI summary:
TLDR: The Blockprint project is focused on measuring and improving client diversity on the consensus layer to prevent network damage from bugs, with a goal of promoting the use of different clients and sharing the workload to reduce stress on developers.
00:00 📊 Blockprint project measures client diversity on consensus layer to prevent network damage from bugs, aiming for equitable split between clients and analyzing block building process to improve diversity, with Main net client diversity close to danger threshold for prism.
04:49 📊 Blockprint algorithm development and client diversity update, including accuracy metrics and outlier Lighthouse, using Python machine learning library.
08:56 📊 Client diversity has not improved with the new model, leading to a campaign to promote the use of other clients and share the workload to reduce stress on developers.
11:34 📊 Blockprint development focuses on improving accuracy and adding features for researchers, while also onboarding more developers and infrastructure operators to make it a core part of Ethereum's diversity infrastructure.
14:25 📊 The speaker addresses recent reorgs and client breakdown, focusing on Lighthouse client issues and the importance of a dashboard, while also discussing online and offline ATT testers and the Y AIS monitoring validators for bugs, with the Blockprint project working on client diversity and accuracy.
16:46 📊 The system classifies blocks from different clients, develops an execution multiplexer for cost reduction, and efficiently builds execution payloads.
19:33 📊 The speaker discusses client diversity and blockprint development, highlighting the need to address misclassification for minority clients and the collaboration with Migal Labs for further development.
24:37 📊 The speaker plans to onboard more people, build infrastructure, and focus on block print instances, while addressing the challenge of fingerprinting execution clients and considering using a unique identifier instead.