C’est sorti aujourd’hui, techniquement ça a l’air intéressant ( https://squawker.org/technology/blockchain-just-became-obsolete-the-future-is-hashgraph/ ).
Ca s’appuie sur des votes (algo Bizantine) + un gossip of gossip. Je cite les développeurs :
Note that we are our own consensus algorithm. While Ethereum is looking at PoS with Casper, our algorithm uses something called Virtual Voting – its a voting system – without having to do the votes. Hashgraph uses a protocol called “Gossip about Gossip” to achieve consensus. Gossip is a well known computer science term, which can be defined as calling any random node and telling that node everything you know, that it does not know. In distributed ledger technology the “baseline” or minimum bandwidth required is that the transactions go to every node. Gossip about Gossip refers to attaching a small additional amount of information to this Gossip, which contains the last person we talked to, hence, we are gossiping about the information we gossiped. Using this information, we can build the Hashgraph. Once we have the Hashgraph, it is extremely easy to know what a node would vote, because we know what each node knows, and when they knew it. We now can use the data from the Hashgraph as an input to 30 year old voting algorithms, and achieve consensus essentially for free. These 30 year old voting algorithms have strong math proofs- they are Asynchronous Byzantine Fault Tolerant, which means we know when we will achieve consensus, guaranteed, and our math proofs make no assumptions about the speed of the internet, due to firewalls, ddos attacks, viruses or botnets. In addition, because of gossip about gossip, Hashgraph is extremely fast, (250,000 transactions/sec), and we also get fair ordering and time stamping on every event.
Par contre, pour l’instant :
- Ya un brevet déposé sur la technologie
- Ya juste un SDK
- Code non libre
Bref, pour l’instant ça a surtout l’air très marketé. Mais c’est un peu curieux de lancer le truc sous une forme privatrice pour quelque chose qui est censé remplacer la blockchain…
Le concept est intéressant, affaire à suivre pour notre R&D de consensus sans proof-of-work. Je laisse le lien vers leurs whitepapers : http://www.swirlds.com/whitepapers/