←back to thread

544 points josh2600 | 1 comments | | HN request time: 0.212s | source
Show context
leifg ◴[] No.26714074[source]
What's the consensus function you use on the mobile coin blockchain. Couldn't see if it's proof of work or proof of stake.
replies(1): >>26714124 #
josh2600 ◴[] No.26714124[source]
We use a modified version of the Stellar Consensus Protocol that we reimplemented from scratch using Rust.

https://github.com/UkoeHB/Mechanics-of-MobileCoin/blob/maste... << Page 81 is where you want to go.

replies(3): >>26714459 #>>26714837 #>>26715172 #
eMGm4D0zgUAVXc7 ◴[] No.26715172[source]
So this does seem to use some voting mechanism instead of scarcity through PoW.

How do you prevent the sybil attack then if there is no scarcity?

I.e. what prevents an individual from spinning up 10 million nodes to get more voting power?

replies(1): >>26715255 #
josh2600 ◴[] No.26715255[source]
Stellar is essentially a liquid democracy system similar to DNS. The core proof at the heart of David Mazieres paper is that in a densely interwoven graph, any set of nodes in a quorum slice reaching consensus is graph consensus. Regarding Sybil, nodes define their own trust relationships, so membership is open but not automatic.
replies(2): >>26716428 #>>26716543 #
eMGm4D0zgUAVXc7 ◴[] No.26716543[source]
> Regarding Sybil, nodes define their own trust relationships, so membership is open but not automatic.

Thanks for clarifying!

So you're expecting people to

- manually add peers to be able to use the network.

- manually monitor the said peers for if they do malicious transactions, and manually ban them if yes.

Right?

How is this supposed to work considering that:

- most users won't care about manually adding peers. They'll just add EVERYONE who offers to be added so they can be done with it and use the system.

- most users probably won't even understand what a malicious transaction is in the first place.

- the few who do will for sure not have the time to monitor a network which does dozens or in the future even thousands of transactions per second.

This seems just humanly impossible, there's by no means sufficient human time available to manually monitor a P2P network's content if the said content is super boring and complex.

If it were a distributed social network you could expect people to e.g. manually flag spam because using a social network implies reading the posts contained in it.

But manually reviewing money which strangers send to each other is boring as hell, who will do this?

replies(1): >>26721018 #
1. comex ◴[] No.26721018[source]
As far as I can tell, "nodes define their own trust relationships" is true only when considering the protocol at an abstract level, or if you're building the code yourself. As a concrete end user using the Signal client, you don't actually get to define your own trust relationships; the app is just hardcoded to trust a specific list of servers, albeit somewhat protected by SGX. Please correct me if I'm wrong.