I got interested into the hashgraph algorithm quite early and wrote a toy implementation in python (in fact discovering an error in the paper in the process). Unless something changed it's entirely useless for open-membership internet-scale consensus. As I remember, the processing time of a message at a node is linear in the number of nodes and same for the local storage at a node, meaning it's not very scalable. Moreover, the nodes must somehow a priori agree on the list of participants of the consensus process, again something which is not realistic for internet-wide consensus. The protocol is quite neat and not too hard to implement but it's similar in scope to paxos/raft: consensus inside an organization, where some things are a priori agreed upon.
No comments yet.