Post
Topic
Re: BYTEBALL: Totally new consensus algorithm + private untraceable payments
by
portabella
on 16/02/2017, 10:21:45 UTC


That's a valid concern and freezal got it right: the network had to start centralized and no one's showed up yet with the needed reputation and credibility.  I'm going to use every opportunity to relinquish control over 11 out of 12 witnesses in favor of reputable individuals/organizations/companies that are willing to be part of the network.  That would make the network only stronger thanks to both their involvement and the network becoming more decentralized.

You could have an election for 12 witnesses like Komodo did for their 64 notary nodes.


Why bringing politics in when there's no need for it? Human intervention shouldn't be as strong as this. I mean Bitcoin works because there's very little human intervention. There's a lot of politics outside the protocol (big blocks, small blocks, BU not BU) but that doesn't harm the continuation of the network.
I'm confused on this one  Roll Eyes

If there is a better decentralised way to select 12 witnesses than a vote by coin holders then we should do that, but is there?

look at eigentrust+ a peer2peer technology which is used to support a trust relationship in decentralized networks
NEM implemented it to create tzogether with their POI (prove of importance) a blockchain consense mechanism that can create a high secured network without any energy waste (POW) and without any human interaction needed (witnesses lists manual set by wallet owners)

https://en.wikipedia.org/wiki/NEM_(cryptocurrency)#Node_Reputation_System
"Although the NEM client is open source and available on github,[7] the NEM server based component, the NIS, is closed source and the binary is obfuscated to presumably prevent decompilers from revealing how it works."

That raises many red flags and sirens.

Another interesting algorithm for P2P networks is FairShare, it improved bittorrent seeding/leeching ratio massively, speeds up initial transfers and is very difficult to attack - since that algorithm doesnt need to ask other peers anything, is only in the client and affects the way client initiates and picks whom to talk to.

EigenTrust looks very nice and interesting, thanks!