Quote from: smooth on October 06, 2017, 10:21:01 PM

Have not deleted any comments. Don't remember whether it was PM or posted.

Grrr (to myself). Remembering some post but I cant find it. Its a nerdy form of the dog eating the car keys.

Ditto now on a research paper about trading off safety for liveness which I wanted to add to our discussion. I know I wrote a link to it some where, but even my writings are so voluminous I doubt I can find it if I do not remember the title.

Quote from: smooth on October 06, 2017, 10:21:01 PM

Quote

My economic point remains that the block producers will attack each other or be attacked by non-network-access means if ever a DPoS chain reaches $1 trillion mcap which I pointed out in my prior post. Actually @r0ach was the first person to make that prediction in 2015 or 2016 afair.

I don't know how block producers can attack each other. They don't need to be directly connected to each other. Unlike Dash masternodes they don't need to have any fixed or public location or IP address, just some mechanism to deliver their signed blocks. As far as non-network means, I don't rule that out and in fact find it reasonably likely.

By disagreeing on the consensus. They could double-spend transactions or refuse to acknowledge blocks. Requires a vote (or some centralized power) to remove the faulty block producers, but there is even a threshold where the voters can not objectively determine which ones are at fault. This is Byzantine agreement.

The entire point is that in an asynchronous setting, there must be adherence to the liveness and consensus thresholds, else nothing at all is objectively observable.

Perhaps it would fork or stall in chaos or the strongest centralized power would take control (if there is one that is much stronger than the rest). Also afaik DPoS is not a formal Byzantine agreement protocol and theres just an assumption of trust and correct synchronous ordering of block production, but afaik no actual protocol mechanism between the block producers to enforce it? I have never seen a formal specification of the DPoS protocol that block producers follow. IOHKs provably correct Ouroboros PoS protocol can be delegated, so DPoS could be changed to that to get a more formal result and then from there we could analyse the dependencies that are just undeclared and adhoc in DPoS.

There is an Issues thread some where on Github where we delved into this last year and even Dan came in there. But I dont remember where that thread is. Perhaps it was in Cosmos issue tracker.

The problem is that n00bs have no comprehension of this. They see Steemit and think wow it works!.