Har Preet Singh

Wanted to point out that the advantages of RLNC over RS have been recognized for a long time now. I've published several peer-reviewed papers analyzing the two codes and RLNC is by far the more performant option for both data movement and storage purposes. This simulator is a great way to see the power of RLNC in action. Research for reference: mit.edu/~medard/mpapers/netc… arxiv.org/html/2603.20426v1

Ethereum– We Need To Talk About Erasure Coding In @VitalikButerin May 3rd blog post “Simplifying the L1” he called out the need for a unified erasure code across three major components of Ethereum: - P2P broadcasting (aka gossip or block propagation) - Data availability sampling (DAS, used in L2 blob posting) - Distributed history storage (how the entire chain history is kept) While erasure coding is on the radar of the top brass at Ethereum, it hasn’t gained much narrative recognition across the wider community. Time to change that. Proper erasure coding will unlock more L1 and L2 scaling potential than anything else on the Ethereum roadmap, a huge step towards operating as a true world ledger. And the wheels are already in motion. One code to rule them all: Random Linear Network Coding RLNC is the product of two decades of my research at MIT, purpose built for handling decentralized data storage and communication as fast and efficiently as possible. So how exactly does RLNC turbocharge P2P broadcasting, DAS, and history storage? P2P Broadcasting- Optimizing Ethereum’s consensus layer starts with making block propagation fast and efficient. We’ll spare you the technical details of slot anatomy, but the core challenge here is that a lot of messages need to reach a lot of validators all over the world in the shortest time possible while using the least amount of bandwidth possible. RLNC is extremely good at this. It parses messages into smaller chunks and encodes them into math equations. Receive enough equations and you get the full message. The key feature– It doesn’t matter which equations you get, only the amount of equations. This vastly reduces the need to send redundant data, reducing latency and bandwidth usage by orders of magnitude. Fast propagation→ fast consensus→ fast chain at scale. Just last month @get_optimum mump2p recorded 6x faster propagation than Gossipsub on Ethereum Hoodi testnet. Data Availability Sampling- L2s settle to Ethereum by grouping transactions into blobs, which are then included in L1 blocks. DAS solves the problem of verifying that the L2 transaction data is available without downloading all of it, instead taking random samples to give high confidence the dataset is in fact available. The type of erasure code used to encode the blob data makes a huge difference in the amount of samples needed to hit that high confidence level. RLNC needs 50x fewer samples than Reed-Solomon codes (used in PeerDAS & Celestia) to achieve the same confidence level. Less samples→ less time & bandwidth used for DAS→ more blob scaling. Distributed History Storage- The goal of distributed history storage is simple– with enough peers you can recreate any past state of the chain. Rather than keeping everything on every node, data is split up amongst nodes to reduce storage costs. Erasure coding historical state data lowers storage costs and reduces bandwidth usage when sending the data around. RLNC excels here for the same reasons discussed on the topic of propagation, but especially important for history storage is its unique recoding capability. When nodes churn we want to be sure the pieces of data they were storing do not get lost. Since RLNC turns data into equations, nodes are able to generate new equations from those they receive, and since a full data set can be recovered from any combination of its equations this creates a system where data is reliably preserved during node churn. When RLNC in Ethereum? Sooner than you might think. Optimum is developing products tailored to each of these three components, with the first being our fast propagation protocol– mump2p. It’s currently running on Ethereum’s Hoodi Testnet with some incredible initial results I'll link below. Mainnet target is H1 2026 where we aim to hit the ground running with our 20+ major node operators partners. Our DAS and storage products are in the R&D phase, so no firm timeline for launches as of now. We’ll be releasing research on both projects along the way, and in fact there’s already a paper we recently published linked in the tweet below for my fellow data availability nerds out there. Erasure coding will turbocharge Ethereum and RLNC is the code for the job.