# EPF Cohort 5 - Week 11 Date: 26/08/2024 ### Updates 1. The spec was almost wholly refactored (it's only a few lines of code) but it highlighted the changes in the initial idea of rated list. * Nodes are scored based not on activity status but the number of samples they(and their descendants) serve successfully. This change would differentiate between (not directly) a supernode that serves only one sample versus a supernode that serves most of its designated samples * We added score use i.e. a `filter_nodes` function that filters out nodes based on scores. There can be various ways we can filter nodes and I think we are gonna simulate and find the best one. * There is a parallel database `ScoreKeeper` that maintains the scores. It logs every sample requested and every sample responded for a node in all of the nodes ancestors (parents, grandparents etc.). Since the list is only at T=2, only parents and grandparents. 2. Finished writing a blog on "node syncing and weak subjectivity". Read it [here](https://hackmd.io/@chirag-parmar/Bk7srGvcA) * TL;DR If a ETH1 chain also downloads the sync target from other nodes why is it not subjective? subjectivity does not exist because a state root is imported socially but because the trust put in the blockchain anchors on the socially obtained information. And since the consensus layer can verify all the other information except the target the beacon chain is *weakly* subjective. > You can find the specs of rated list [here](https://github.com/dankrad/rated-list-specs) > You can find the code of rated list [here](https://github.com/hopinheimer/rated-list-research) ### Next Steps 1. The above blog sets the stage for Part 2: "Weak Subjectivity Period" which is the actual goal. 2. Build a simulator for rated list because we decided to pause on the POC development. 4. Finish up on the snap sync blog from the initial weeks of the EPF.