PARSEC is claimed to be a great breakthrough, but another heavyweight team in crypto just declared their best consensus algorithm(since bitcoin invention) a few days back, take a look at their whitepaper and press report, the project is vested by Emin Gun Sirer, who is a very respected figure in the crypto land ( the guy first discovered the DAO bugs).
The new algorithm is also claimed to be better than Algorand and build on top of DAG and their claimed to be a leaderless BFT as well (which is what PARSEC claimed to be), is the SAFE team ever looking into their paper, I think they bear lots of similarity and I like their whitepaper, which include lots of actual test stats.
https://ipfs.io/ipfs/QmUy4jh5mGNZvLkjies1RWM4YuvJh5o2FYopNPVYwrRVGV
2 Likes
I’d like to see a reply to this by someone who understands these things.
2 Likes
We did look into this paper, and referenced it in PARSEC.
The main difference that disqualifies it for our use is that it’s synchronous.
The idea is based on some kind of “multi gossip” where each node gossips to k random nodes at any time step. They then use history (gossip DAG) to favour values that have been represented more.
I think it’s quite interesting because it’s more original than some other ideas out there, but comparing it to other synchronous consensus algorithms, I find it less elegant than “Byzantine agreement, made trivial”. The approach they follow has more of a brute force feel to it.
The complexity is O(kn log(n)), where k << n, and k is a security parameter.
For us, the complexity is O(n log(n)), which is marginally better.
My personal opinion is that our approach is also simpler to understand, but I may have a bias here.
So I would say that PARSEC is better than avalanche for these three reasons:
- highly asynchronous
- more elegant
- marginally more efficient
7 Likes