,
16 tweets,
6 min read
Read on Twitter
1/ This thread is my attempt to explain @hashgraph in 20 tweets. As Erik points out in this video, blockchain is a Version 1 DLT. It is fundamentally incapable of scaling securely in permissionless environments because it is leader-based and current implementations lack finality.
2/ PoW is used to secure the network but also to (1) drive probabilistic consensus through a lottery system that randomly picks "leaders" to propose new blocks and (2) throttle the network so it doesn't inadvertently fork the database every time a new block is added to the chain.
3/ @hashgraph is a fundamentally different paradigm altogether. It has no leaders and reaches finality within seconds. This means Hedera's DLT can process 100,000's of t/ps, pre-sharding (potentially millions of transactions or more with a fully sharded database). Let me explain:
4/ To begin with, let's establish that it is extremely difficult to build a byzantine algorithm (contrary to popular belief, Bitcoin is not Byzantine fault tolerant). This problem has been solved, however, and there is a body of academic literature on it that goes back decades.
5/ Some of the strongest implementations were built using voting protocols (computers casting votes on yes/no questions in order to come to agreement on the state of a shared database). The problem is they don't scale. You can't practically deploy such a system in the real world.
6/ So, we know voting algo's achieve the gold standard in security (asynchronous Byzantine fault tolerance), but everyone knows they don't scale because of computational complexity, right? Wrong. They do, and @leemonbaird figured out how to scale them by providing an answer to...
7/ ...the question: "What is the minimum amount of information needed to run a voting protocol?" It turns out that solving consensus is a problem of information. @leemonbaird's great insight was to recognize that almost all the information needed to run a voting algorithm...
8/…is already on each and every computer that is participating in maintaining the shared database and that the only missing information takes up a couple of bytes of data. When this additional information is shared during the normal course of network communication, each node...
9/ ...can independently run a voting algorithm locally by representing all the other nodes virtually, without having to cast any votes across the Internet. Until Leemon invented @hashgraph, voting protocols were order N^X algo's. They were logarithmically complex.
10/ The @hashgraph consensus protocol is an "order 0" algorithm when it comes to executing a voting protocol. I cannot overstate enough what this means. It is a revolutionary innovation. @leemonbaird has solved consensus. Full stop. This is why a governing council of some of...
11/...world's largest multi-national organizations have agreed to shepherd the network's transition to a permissionless environment (see their "Path to Decentralization" in tweet #19 for further details). It is why @hashgraph can outcompete credit cards in latency and throughput.
12/ ...The council members that they’ve already announced are companies like Nomura, Deutsche Telekom, DLA Piper, etc. But the biggest names haven't even been announced yet (they will be announced as part of Open Access). You can read more about that here: hedera.com/council
13/ The @hashgraph consensus protocol has been independently validated by what is known as a Coq proof (conducted by Karl Crary at @SCSatCMU), which is a formal proof verification method conducted by computer. It is the future of computer science in terms of how code is checked.
14/ Karl is one of the best in the world at this, and even in his case, it took several months to translate the formal math proof into machine language. You can read about his and Carnegie Mellon's Coq proof of the aBFT @hashgraph consensus algorithm here: hedera.com/blog/coq-proof…
15/ This is why everyone involved w/@hashgraph who understands the technology is so excited about it. It is why Hedera has worked closer and more carefully with regulators in order to be compliant than any other DLT or permissionless database in crypto that I know of, including…
