,
9 tweets,
4 min read
Read on Twitter
@SarahJamieLewis 1\ If I understand correctly (please correct me if I am wrong), you assume that the local nature of the decision making process would lead to parts of the network making different decisions than other parts (pretty much like the "pockets" in your simulations) ...
@SarahJamieLewis 2\ ... and that adding byzantine nodes would only make this situation worse? Intuitively this sounds like a reasonable assumption - after all it would be "easy" to just "surround" parts of the network and split the opinions, right? Interestingly this is only true for ...
@SarahJamieLewis 3\ ... low dimensional spaces like your simulated 2d-world. The higher the dimensionality of the network, the harder it becomes to achieve "stable domain walls" that prevent the system from converging. In fact 8 neighbors and upwards, seem to be enough to prevent this ...
@SarahJamieLewis 4\ ... from happening completely and it also becomes increasingly difficult to "surround" certain areas of the network because the network essentially becomes infinitely-dimensional (assuming a verifiably random peering). In fact, such a network is even more resilient ...
@SarahJamieLewis 5\ ... against attackers that try to prevent it from converging, than a completely random querying like in avalanche. This is due to the fact, that even in a 50/50 initial distribution of opinions, you will always have small fluctuations locally, that will form local maxima ...
@SarahJamieLewis 6\ ... which will cause a tipping of the network in one of the directions (first locally and ultimately globally in a chain-reaction like process). It is pretty much impossible to keep the network undecided. In addition, the local deterministic behavior of the protocol ...
@SarahJamieLewis 7\ ... allows you to not only speed up the voting (because you don't need to query others), but also to verify the honesty" of nodes using our "Proof of Opinion" mechanism, which will boost the byzantine resistance substantially. This is why I asked if you have ever simulated ...
@SarahJamieLewis 8\ ... a multi-dimensional space. Human intuition doesn't work very good when it comes to "chaotic / random" systems like these and I would really encourage you (and everybody else) to simply "try it out", rather than relying on a gut feeling alone. Another important aspect ...
@SarahJamieLewis 9\ ... that we have not published, yet is the sharding approach, which will allow us to rely on "local consensus" rather than a global one but it doesn't make much sense to talk about this before we have revealed how this will work.
