,
13 tweets,
5 min read
Read on Twitter
1/ A bit late to reading this, but this is an absolute *gem* by @RyanTheGentry, @KyleSamani & the @multicoincap team. Highly recommended read.
I agree w/ Kyle more efforts should be focused on understanding this aspect of BFT consensus: separating time & state.
I agree w/ Kyle more efforts should be focused on understanding this aspect of BFT consensus: separating time & state.
2/ If the prior distributed consensus rendition was focused on better understanding security, this one by the Multicoin team focuses on better understanding scalability. @RyanTheGentry does a fantastic job breaking this down for us - thankfully, we have him to learn from!
3/ Recall distributed systems lack a global clock (covered more in the below thread). Thus, whether by partial synchrony or asynchrony, the global state of a blockchain is locked until the order/sequence of events is determined & the next block is added.
4/ Ryan summarizes this nicely by saying, in order to increase throughput, a protocol must either increase the block size or decrease the block time. Pick one. In other words, time and state have been coupled since the creation of Bitcoin.
5/ Enter sharding. Everybody is familiar with the term "divide and conquer". Sharding breaks the global state into a bunch of smaller pieces, each w/ its own independent clock & each w/ the ability to progress independent of the others.
6/ Sharding helps split a network into smaller sets of chains where each chain has its own state in an asynchronous (e.g., randomly ordered) fashion, allowing network participants to process a lot more transactions than they could normally & gain a lot of throughput as well.
7/ Sharding represents an important step taken to separate time & state, but still doesn't quite accomplish it. Even though an underlying beacon chain provides a global clock, each shard only syncs their local clock w/ the beacon chain's periodically.
8/ This creates a challenge of inter-shard latency. As Ryan puts it, "in order for Alice to send tokens from shard A to Bob in shard B, validators in shard A must prove that they burned Alice’s tokens before validators in shard B will mint an equivalent amount of tokens for Bob."
9/ In other words, the global state of a blockchain is locked until there is communication between shards to determine the order/sequence of events. Thus, even sharding can potentially run into some scalability limitations.
10/ Thus, the challenge remains how to accomplish faster clock & state updates. This goes back to separating state & time. @RyanTheGentry & @KyleSamani highlight that this is just what @Solana introduces: a global trust-minimized clock that progresses w/o updating global state.
11/ As discussed more in the post, this is done by hashing the hashes continuously intra-block rather than chaining hashes together every block. In other words, every node is forced to encode the passage of time in a data structure, called Proof-of-History (PoH).
12/ It remains extremely early to see how well this mechanism could work, & what degree of security it can achieve (vs. existing PoW & current PoS experiments). Nonetheless, this piece was worth highlighting for any1 trying to better grasp the current state of layer 1 innovation.
13/ Thank you, @RyanTheGentry & @KyleSamani for open-sourcing this valuable knowledge!
Disclosure: Vision Hill is not invested in Solana. This was highlighted for informational purposes only.
Disclosure: Vision Hill is not invested in Solana. This was highlighted for informational purposes only.
