Gossipsub may not be the best option for Eth 2 - Networking - Ethereum Research Ethereum Research Gossipsub may not be the best option for Eth 2 Networking jamesray1 February 20, 2019, 11:10pm 1 See https://github.com/w3f/messaging/ for an alternative with Whisper v2.0. For reference, this is the gossipsub spec. Pros: see the Protocol requirements section of the above, as well as the Motivation Cons: it’s not specified or implemented yet MihailoBjelic February 21, 2019, 5:45pm 2 What are the advantages of the alternative (W3F’s) approach? jamesray1 February 21, 2019, 11:13pm 3 I’ve edited my post with pros and cons and made it a wiki. jamesray1 March 10, 2019, 5:08am 4 Gossipsub has a different purpose to Whisper v 2.0. Following is a reply from Jeff Burdges of the Web3 Foundation which cleared up misunderstandings on my part: We have four unrelated “messaging” projects in the works: There is gossipsub work for libp2p being handled entirely internally at Parity. It’s nothing to do with “messaging” in any conventional sense, just about replacing p2p broadcasts with a more efficient strategy. You should ask the rust-libp2p developers for more information on this. In Polkadot, we have an interchain message passing protocol (ICMP) for polkadot that is 100% about blockchains authenticating blocks, and nothing to do with normal “messaging”. We also have a more directed chunk distribution process used in the availability game for polkadot, where nodes must send very different messages to all other nodes, so basically the extreme opposite of gossipsup but still all libp2p work. Finally, the w3f/messaging repo you found exists for cordinating our work on mix networks. At first blush, mix nets are only for messages between two individuals, or from an individual to a blockchain, as they do not obviously support group messaging. In fact, we do want to support group messaging in the mixnet, but never via gossip of course. As a rule, gossip protocols are never anonymous because they cannot provide cryptographic unlinkability.* We’ve encountered problems with adding important functionality, like header encryption, to the current best-of-breed group messaging cryptography like MLS: https://messaginglayersecurity.rocks We’ll therefore focus on the 1-to-1 case this year, as that presents enough problems all by itself. Home Categories FAQ/Guidelines Terms of Service Privacy Policy Powered by Discourse, best viewed with JavaScript enabled