DON consensus question

[Anonymous No.53456429]

If 9/10 nodes colluded to lie about data for some contract, and I raised an alert, how would that be resolved? If it’s raised with another DON, the same issue is just getting passed along from DON to DON, no? What stops the next DON from colluding and saying my alert is false?

[Anonymous No.53456442]

>>53456429
Absolutely nothing. the whitepaper completely missed this obvious hole in the product.

You must be a genius. Amazing you are asking these questions when you understand how chainlink works so thoroughly. Literally operating on a different plane of existence.

[Anonymous No.53456444]

>>53456429
If you’re right you get all the collateral. So you’re incentivised to tell the truth otherwise you go poor and become unrepeatable user wise. This is completely from what we have today btw. Atm you’re incentivised to shut up and take, and if it ever comes out something bad happen go into damage control and if that doesn’t work bail. Look at SBF and the rest of the crypto space. And regulations is going to fuck it even more

[Anonymous No.53456455]

>>53456429
one of those node have a reputation of rugging people, so how will anyone trust them now?

[Anonymous No.53456470]

>>53456442
Mate, I’m just trying learn over here. Stop being so defensive
>>53456444
Ok, but what determines if I am right. My claim that other nodes are wrong is brought to a second group of nodes, a tribunal of some sort right? How does that determine what is right and what isn’t when the same thing can happen - they could collude

[Anonymous No.53456477]

>>53456470
Read the fucking whitepaper if you want to learn.

[Anonymous No.53456567]

>>53456429
>What if 9/10 nodes colluded to lie about data for some contract
What if 6/10 Bitcoin miners agree to split Satoshis wallet between them?

[Anonymous No.53456595]

>>53456429
Because at every step, there is always a greater incentive to tell the truth than there is to lie. You should read up on Link's superlinear staking, in their blog they explain in detail how idshonest behaviour is disincentivized.

[Anonymous No.53456600]

>>53456429
>>53456595
Here, I even went and foudn the pertinent info for you.

[Anonymous No.53456693]

>>53456429
That's a legitimately good question because the 2nd whitepaper wasn't very clear on how this problem is resolved. The reason it wasn't clear is that the answer is "it depends". Chainlink's approach to decentralization is that it's up to the users (data consumers) to decide what kind of setup and SLA they're interested in. They just offer infrastructure with guarantees it'll operate as intended, but it's up to you to set it up as centralized or as decentralized as you'd like.

So to answer your question more directly, in use cases where huge sums are in play and you want maximum security you could set it up like this:
- 1st layer DON = 20 nodes using OCR2 for consensus
- if an alert is raised by any of the participating nodes OR an external alerter, 2nd layer is asked for verification
- 2nd layer of security is a random selection of 10 high reputation nodes (out of thousands or tens of thousands available in the network), they vote with a simple majority; 2nd layer nodes voting against majority will get slashed

While it's feasible for more than half of 1st layer DON nodes to collude to produce fabricated response, the chances of more than half of all high reputation nodes to do that (and that would be required in case of random selection for the 2nd committee) is almost nonexistent

[Anonymous No.53456698]

>>53456429
Bro just ask chatgpt. It knows alot about link usecases and shit like this.

[Anonymous No.53456890]

>>53456693
Thanks, this is really helpful. The key point I was missing was the random selection of high reputation nodes in round 2. I assumed it was just some stagnant group of nodes, equal to any other node.
So essentially super linear staking occurs in the first round among regular nodes, which requires a decision from elsewhere on what the truth is.
And to determine what is true in the first round, it’s down to democratic vote by a randomly selected jury of trusted nodes (with penalty for being in minority)

[Anonymous No.53456977]

>>53456429
>chesed to meet you
https://en.m.wikipedia.org/wiki/Chesed