4 ms·

A Block header includes a root hash of a Merkle tree of transactions. So yes, it is a tree, if you also consider the useful payload, not just block headers. Th

A Block header includes a root hash of a Merkle tree of transactions. So yes, it is a tree, if you also consider the useful payload, not just block headers. Then e.g. in Ethereum there's also a Merkle tree of the entire state as well. > there's no distributed consensus Distributed consensus is necessary if you want high availability and replication for your Merkle tree. If data in your tree is important, you want to replicate it, right? You want to be sure it's stored on multiple nodes. So this distributed consensus thing, Byzantine Fault Tolerance, it basically gives you a receipt that at least N nodes have this data stored. You didn't implement it because it takes considerable effort. But if you implemented Merkle tree with high availability, you'd get something very similar to blockchain.

thwayunion 4y ago
> If data in your tree is important, you want to replicate it, right? Yes! > You didn't implement it because it takes considerable effort. I'd wager it's because he was smart enough to read the man page for rsync and didn't over-engineer a non-solution to a non-problem in his particular use-case.
- killerstorm 4y ago
  Data integrity & availability generally considered to be important. So e.g. many databases implement synchronous replications, and Turing Award winning scientists Liskov and Lamport developed Byzantine Fault Tolerance solutions. I guess they were just too lazy to read the man page for rsync...
- hinkley 4y ago
  Something like that. The point was accuracy, not throughput.
hinkley 4y ago
> Distributed consensus is necessary if you want high availability and replication for your Merkle tree. No, distributed consensus is necessary if you're willing to sacrifice correctness for availability. >> public safety > So this distributed consensus thing, Byzantine Fault Tolerance, Show me a blockchain that actually implements BFT. Longest chain wins is not Byzantine Fault Tolerance.
- killerstorm 4y ago
  > No, distributed consensus is necessary if you're willing to sacrifice correctness for availability. ?!?! Do you think it's most correct to have only one replica? > Show me a blockchain that actually implements BFT. A lot of blockchains implement some variant of PBFT. Including the one I made: https://gitlab.com/chromaway/postchain/ https://gitlab.com/chromaway/postchain/
  - hinkley 4y ago
    Specifically, longest chain is Might Makes Right, which is not consensus, it’s leader following. If I makes you feel any better, I see Raft as only BFT during leader election, the rest of the time if the leader is a bad actor then everything is fucked. It’s a game of Secret Hitler. Remember, the Byzantine Generals problem is what the BFT term is acknowledging, and the problem with the generals was one of sabotage, not just missing messages. It’s only statistically unlikely that longest chain is created in earnest and not via malfeasance. But the bigger a system scales the more frequent counterexamples occur. That’s the lesson of every scalable system. Bad days that used to happen once a quarter now happen a couple times a week. Edit: > In a consortium blockchain, typically blocks must be approved (signed) by a majority of consortium members. Well, well. Adults have finally joined the blockchain community. Maybe there’s hope yet.

thwayunion 4y ago

> If data in your tree is important, you want to replicate it, right? Yes! > You didn't implement it because it takes considerable effort. I'd wager it's because he was smart enough to read the man page for rsync and didn't over-engineer a non-solution to a non-problem in his particular use-case.

killerstorm 4y ago
Data integrity & availability generally considered to be important. So e.g. many databases implement synchronous replications, and Turing Award winning scientists Liskov and Lamport developed Byzantine Fault Tolerance solutions. I guess they were just too lazy to read the man page for rsync...
hinkley 4y ago
Something like that. The point was accuracy, not throughput.

hinkley 4y ago

> Distributed consensus is necessary if you want high availability and replication for your Merkle tree. No, distributed consensus is necessary if you're willing to sacrifice correctness for availability. >> public safety > So this distributed consensus thing, Byzantine Fault Tolerance, Show me a blockchain that actually implements BFT. Longest chain wins is not Byzantine Fault Tolerance.

killerstorm 4y ago
> No, distributed consensus is necessary if you're willing to sacrifice correctness for availability. ?!?! Do you think it's most correct to have only one replica? > Show me a blockchain that actually implements BFT. A lot of blockchains implement some variant of PBFT. Including the one I made: https://gitlab.com/chromaway/postchain/ https://gitlab.com/chromaway/postchain/
- hinkley 4y ago
  Specifically, longest chain is Might Makes Right, which is not consensus, it’s leader following. If I makes you feel any better, I see Raft as only BFT during leader election, the rest of the time if the leader is a bad actor then everything is fucked. It’s a game of Secret Hitler. Remember, the Byzantine Generals problem is what the BFT term is acknowledging, and the problem with the generals was one of sabotage, not just missing messages. It’s only statistically unlikely that longest chain is created in earnest and not via malfeasance. But the bigger a system scales the more frequent counterexamples occur. That’s the lesson of every scalable system. Bad days that used to happen once a quarter now happen a couple times a week. Edit: > In a consortium blockchain, typically blocks must be approved (signed) by a majority of consortium members. Well, well. Adults have finally joined the blockchain community. Maybe there’s hope yet.