> I did NOT invent 99% fault tolerant consensus; Leslie Lamport did. I just wrote an explainer and adapted it to a blockchain context. Please fix, Trustnodes.
Damn that link really makes it look like Vitalik is running a money scam, but seems like it’s just a scammer who’s replying to him with near identical name talking about a scam.
Is twitter really such a Wild West for scammers or is that just the cryptosphere?
Cryptocurrency is absolutely overrun with scams. It's hard to find anything with actual merit in the vast ocean of spam that is the digital currency world.
It won't stop until the bubble crashes or enough people wise up that the scammers no longer make money. As it stands low effort shitcoin token scams using digital variants of the "big store" and Ponzi schemes are still cashing out pretty good.
That's at the higher effort end of the scam spectrum. For every one of those there are 50 search and replace coin forks and fake tokens for things that don't even pretend to exist.
Note that the title implies Vitalik invented this algorithm. Vitalik was very clear (as a paragraph in this article does disclose), that he was summarizing Leslie Lamport’s algorithm from 1982, and showing how it could be applied to the Blockchain context.
I'd have to read the paper, but I'm extremely skeptical as I've read most of Lamport's work and I don't recall a relevant solution.
Also, Journalists: please stop repeating propaganda about a 51% attack, it's a <33% attack. Not all chains are created equal on this, but all chains can be attacked by a 33% dos attack.
I did, it's for "investors" rather than developers. I'm reading Vitalik's blogpost now. I see what he's talking about now. I'm going to finish it and decide, but I see which part of Lamport's work he's referring to.
A network using this technique needs a reasonable upper bound on network latency for validators (hard to get) and needs to keep the set of active nodes relatively small (which forever dooms a traditional mining pool to be somewhat concentrated).
The former problem can be substantially improved by:
1) Using custom time hardware a la spanner.
2) Causing the network to halt entirely in unfavorable network conditions.
This has in fact been known for a long time; Leslie Lamport’s famous 1982 paper “The Byzantine Generals Problem” (link here) contains a description of the algorithm. The following will be my attempt to describe and reformulate the algorithm in a simplified form.
Vitalik's explainer doesn't address Sybil attacks at all. Deploying a constellation of nodes all controlled by the same attacker seems like a way to defeat even this protocol.
This was the line of thinking that lead to proof-of-work as described in the Bitcoin white paper.
I wondered about this too. Vitalik is real, so I'd be curious to read a deeper analysis.
With sybil attacks its important to realize there is no upper bound on effort or sophistication. A sybil can use botnets, social engineering, machine learning, etc.
There's an assumption that there's some proof of work or proof of stake (or heck, proof of authority)-based sybil resistance mechanism that prevents anyone from launching a huge pile of nodes.
I tend to (and think all Byzantine fault tolerance theorists should) view consensus set selection (where sybil resistance is a key goal) and consensus algorithms as largely separate problems.
You and I know that, but the press and the public don't and they will run away with crazy headlines. This caveat should really be the first line of the blog post (same for Avalanche).
> In a highly complex paper, Ethereum’s co-founder Vitalik Buterin has proposed a new kind of consensus algorithm which requires just 1% of nodes to be “honest”.
You know, it's probably not a good sign if the authors of a site literally called blockchain mania consider a paper summarizing a classic on consensus algorithms "highly complex".
I am sure this must be handled somewhere, but why can't malicious observers in this scheme flag honest behavior, creating a vector for denial of service. I suspect that observers must report something that is actually true, but my gut is still saying am missing something more fundamental here.
The original article [1] referenced is interesting, but I'm not sure if the proposed algorithm is useful in a completely unauthenticated and decentralized network, since it makes assumptions about identity of nodes and the latency between them. Are these chosen "observer" Nodes doing any PoW? Otherwise I assume this is a nice vector for a sybil attack.
PS: Could the link be changed to the original paper [1]? The linked article doesn't add much and shows ads.
Also what happens in a split brain scenario e.g. if some massive undersea cable cut or BGP screwup splits USA and EU for 5 minutes (or causes a few minutes of massive packet loss).
21 comments
[ 16.7 ms ] story [ 52.1 ms ] thread> I did NOT invent 99% fault tolerant consensus; Leslie Lamport did. I just wrote an explainer and adapted it to a blockchain context. Please fix, Trustnodes.
[1]: https://twitter.com/VitalikButerin/status/102797212659301580...
Is twitter really such a Wild West for scammers or is that just the cryptosphere?
It won't stop until the bubble crashes or enough people wise up that the scammers no longer make money. As it stands low effort shitcoin token scams using digital variants of the "big store" and Ponzi schemes are still cashing out pretty good.
Here is a recent howler I spotted: http://metahash.org
That's at the higher effort end of the scam spectrum. For every one of those there are 50 search and replace coin forks and fake tokens for things that don't even pretend to exist.
Also, Journalists: please stop repeating propaganda about a 51% attack, it's a <33% attack. Not all chains are created equal on this, but all chains can be attacked by a 33% dos attack.
A network using this technique needs a reasonable upper bound on network latency for validators (hard to get) and needs to keep the set of active nodes relatively small (which forever dooms a traditional mining pool to be somewhat concentrated).
The former problem can be substantially improved by:
1) Using custom time hardware a la spanner. 2) Causing the network to halt entirely in unfavorable network conditions.
https://vitalik.ca/general/2018/08/07/99_fault_tolerant.html
The link points to Lamport's original formulation of the Byzantine Generals Problem:
https://people.eecs.berkeley.edu/~luca/cs174/byzantine.pdf
Vitalik's explainer doesn't address Sybil attacks at all. Deploying a constellation of nodes all controlled by the same attacker seems like a way to defeat even this protocol.
This was the line of thinking that lead to proof-of-work as described in the Bitcoin white paper.
With sybil attacks its important to realize there is no upper bound on effort or sophistication. A sybil can use botnets, social engineering, machine learning, etc.
I tend to (and think all Byzantine fault tolerance theorists should) view consensus set selection (where sybil resistance is a key goal) and consensus algorithms as largely separate problems.
You and I know that, but the press and the public don't and they will run away with crazy headlines. This caveat should really be the first line of the blog post (same for Avalanche).
You know, it's probably not a good sign if the authors of a site literally called blockchain mania consider a paper summarizing a classic on consensus algorithms "highly complex".
PS: Could the link be changed to the original paper [1]? The linked article doesn't add much and shows ads.
[1] https://vitalik.ca/general/2018/08/07/99_fault_tolerant.html
https://vitalik.ca/general/2018/08/07/99_fault_tolerant.html
If 90% of the nodes decide to go to ETH2 and 10% of the nodes decide to stay in ETHC, then which side wins or there is a deadlock?