Chain-off, designed for developers and students, Insider!
This article from the chain off the block chain Q & A community , refused reproduced without permission.
Mechanism for consensus on the block chain structure can be understood as who is going to block, and to preserve the unity of the block chain, mention this issue when he would mention the Byzantine fault tolerance.
Byzantine generals problem: This story can be traced back to the year of the Roman Empire during the war, the Imperial has a large army, each army guarding their own turf, the distance is relatively far away. At that time the communication unlike now, there are micro-channel, buttoning a shirt, and a telephone. They command armies of transmission is through so-called messenger to transmission. Prior to the war all the army must reach a consensus, seen the costume drama friends all know, inside the army may occur traitor or an enemy spy, etc., resulting in the army can not move at the same time, they failed to reach agreement. This is the Byzantine generals problem in reality.
But now the continuation, on the issue of the Byzantine general, it presents a Byzantine fault tolerance, Byzantine fault tolerance is a model of assumptions about the real world, in a distributed network system point to point, the node due to hardware failure, or malicious attacks and so on, leading to unpredictable consequences of a total system. The Byzantine fault-tolerant protocols must deal with these unknown error.
Consensus mechanism block chain in a process of evolution, the beginning of time Nakamoto Daniel did not use BFT, used in Bitcoin is formerly known as articles written: proof-of-work. POW workload proof and longest chain mechanism, is the ultimate expression of Bitcoin Bitcoin longest chain in the consensus mechanism. With the development of block chain technology, they are designed to work by consensus algorithms to ensure system uniformity.
In fact, the consensus mechanism is no solution in a distributed system, because the communication between the many nodes, there must be reasons for unreliable networks, fault host, attacked it, and so on. It is not guaranteed to achieve full consensus. This conclusion as early as 1985, when he had raised. They are three: Fischer, Lynch and Patterson. They proposed the principle of a FLP impossible: under the premise of a reliable network, any node fails, the asynchronous model system to minimize one or more of them, there can be a deterministic algorithm to solve the same problem. The three later papers Dijkstre also won the award. Interested parties can Google it.