flâneur

The Goldilocks consensus problem - Varun Srinivasan

varunsrinivasan.com · 1,410 words · saved by 1 readers

Imagine that you wanted to build a sufficiently decentralized Twitter — a social network in which no single person or company is in control. How would you build something like that? A social network is decentralized if it has many servers run by different people. The data would be spread across servers, so you’d need a simple way to answer questions like “What are the latest posts?”. This is where consensus comes in. We developed a new consensus model for Farcaster, our decentralized social network. Existing models had user experience and decentralization tradeoffs, making scaling a social network difficult. Our model, called a deltagraph, uses blockchains and CRDTs to scale to millions, and hopefully billions, of users. Federation and blockchains, which are common models in untrusted networks, have tradeoffs that aren’t ideal for a social network. Federation is simple but leads to centralization, while blockchains prevent this but are slow and expensive. In a federated system, anyone

The Goldilocks consensus problem - Varun Srinivasan The Goldilocks consensus problem Imagine that you wanted to build a sufficiently decentralized Twitter — a social network in which no single person or company is in control. How would you build something like that? A social network is decentralized if it has many servers run by different people. The data would be spread across servers, so you’d need a simple way to answer questions like “What are the latest posts?”. This is where consensus comes in. We developed a new consensus model for Farcaster , our decentralized social network. Existing

related reading