You are viewing a single comment's thread from:

RE: There is NO 51% Attack in Steem. Is there? Help! Dan! :D

in #steemtron3 years ago

However, I did not know that Graphene blockchains use a Pos/PBFT hybrid. Are you sure about that? Do they gather two quorums? Is there a view change algorithm?

right, no not POS/PBFT thats the point! The POS in dPOS is not the same as POS in Proof-of-Stake or Proof-of-Work (which is POS), which both are no consensus algorithms but ways to ship around the FLP-Impossibility theorem.

Besides the function of leader election oracle and Sybil-resistance mechanism and incentive-estimator both are clocks in the first place. Instead of time Nakamoto-Consensus uses entropy of the memory-less Hashfunktion. This is why we have synchronicity assumption in an open hence asynchronous network. One does not simply assume synchronicity. Same goes for POS

the POS in dPOS is just Sybil-resistance mechanism. Not more not less. Not related to "real" POS. The hybridization I refer to is that Steem can additionaly incorporate a Proof-of-Work sheme (which it had until 2017).

Additionally, there are already protocols that scale traditional consensus linearly (HotStuff, Byzcoin, etc) that could scale easily too much bigger numbers of witnesses.

yeah or like Tendermint BFT but they are asynchronous? As far as I have understood the point with "Larimer-Consensus" is that it sacrifices open consesus setup for the property of "pending"-state in the case that the correctness of the algorithm gets violated.

and that it has no 100% finality but an probabilistic escape hedge like the lasvegas algorithms.


Yes I know what PoS here is used for compared to the "real PoS".
I still didn't know that Steem/EOS/Bitshare use a 2phase protocol that collects two quorums from the majority (2f+1) of witnesses (I will have to read up on this).
In the end that would make it a permissioned blockchain.

HotStuff is like PBFT, it has partial synchrony and is still linear (an adaptation of it will be used in Libra afaik).

In this paper they assume DPoS got 51% as well:

It is a permissioned Blockchain this is Vitaliks critique

Dan's EOS achieves its high scalability by relying on a small number of what are essentially master nodes of a consortium chain, removing Merkle proofs and any other protections that would allow regular users to audit any part of the system's execution unless they want to personally run a full node themselves. See for why I think this is undesirable.

Du meinst weil die Witnesses im Rundverfahren arbeiten und an den LIB ansetzen ist es kein 2Phasen Quorum? Warum skaliert es dann nicht in der Anzahl der Validatoren. Werde es auch nochmal durchlesen. Super Einwand


Hast du mir den Link zum whitepaper wo das beschrieben ist?

das ist nur ein reddit Eintrag von Vitalik Buterin welchen er auf seinem Blog ausführt.

Hier spricht Larimer selber über 2f + 1

We then introduced a new concept known as the Last Irreversible Block (LIB). This is a block which has been confirmed by ⅔ or more of the elected block producers. No node will automatically switch to a fork that isn’t built on top of the LIB.

Wo soll der Schwellenwert herkommen wenn es kein klasisches Quorum ist. Ich bin überfragt. Hoffentlich findest du was raus

Confirmed ist wahrscheinlich wenn ein witness ueber dem block nen anderen block produced hat. Dass ein block von weniger confirmed sein kann heisst das es kein quorum haben kann.

Ja stimmt das wäre dann ein solcher Minority-Fork. Aber hier produziert die Minderheit einen Block nur alle 9 Sekunden und verliert das Rennen um die längste Kette.

sry ich habe ganz vergessen es gibt ja ein missing Whitepaper von Larimer zu dPOS:

To help explain this algorithm I want to assume 3 block producers, A, B, and C. Because consensus requires 2⁄3 + 1 to resolve all cases, this simplified model will assume that producer C is deemed the tie breaker. In the real world there would be 21 or more block producers. Like proof of work, the general rule is that longest chain wins. Any time an honest peer sees a valid strictly longer chain it will switch from its current fork to the longer one.

das meint Gün Sirer mit hybrid es nutzt longest chain rule und ist kein Quorum im klassischen Sinne aber ein permissioned consortium welches im Round robin verfahren Blöcke produziert, dann ergibt sich doch 2f+1

Ja, im Endeffekt sehe ich aber nicht wie das quadratisch messages hat. Es gibt halt nen broadcast von jedem block producer per instance, aber das ist immer noch nicht quadratisch (per round). Ich denke das problem ist halt dass bei 100 nodes, der broadcast ne weile dauert und evtl nicht schnell genug bei #100 ankommt. Gossip wuerde das noch langsamer machen (allerdings mit less messages).

hast ja auch recht, habs editiert. Der Flaschenhals kommt wohl von der 3 Sekunden Blockzeit. Wäre die Frage ob die echt von dem Netzwerk Delay kommt (kenn mich da garnicht aus). Bei Bitcoin ist es ja die Difficulty als Zeitersatz die es auf 10 minuten konvergieren lässt. Klar 3 Sekunden klingt schon nach Übertragung wirst Recht haben. Problem gelößt danke 👍