Fraud Proofs introduce a so-called Dispute Time Delay (DTD), once a proof is submitted in the L2, the validator has time in order to mark the proof as invalid; an invalid proof could have incorrect state transitions and thus result in penalizing the validator(s) involved; a rollback of the state to the latest valid snapshot will follow.
Example:
咱一帮人往sever加入large prime number,作为其中的一员
相信别人新加入的数字是prime number【有信任的】
对于一个新人,run prime number test, make sure it is same as our database【系统之间是不信任的,问题是新加入的人算力更强我们就很亏】
我们尝试下,让别人帮我们证明,可以对话和指控
咱让其他的peers给我们发送证明,证明我们数据库中的某个数字不是prime number 的话
这样不用每个prime number 都跑一遍test. 如果在超过了一定时间没收到fraud-proof那这个就是prime
至少一个peer在check 发现问题就可以在整个网络announce
证明快
可以惩罚欺诈的人
Instead of validating all the rules of the system, others can demonstrate that the current system state is wrong.
Light-nodes vs. Full-nodes
When a new block appears, we process its transactions and define the new current state. The nodes that do this work are called full-nodes
If a node runs on an Android device, it has limited resources, such as computing power, storage, and bandwidth. These nodes are named light-nodes since they only download block headers instead of all the data associated with it.