There is an "impossible triangle" in the underlying design of the blockchain. Security, decentralization, and performance can only be selected out of the three. Currently, the most common POW mechanism and DPos mechanism are the same. Can existing solutions break this impossible triangle? On September 5th, at the POD conference hosted by Odaily and co-organized by 36Kr Group Strategy, Tai, the founder of Nervos, gave his own answer.
Tai Chi believes that the Layer 2 extension of the blockchain is a good medicine to solve performance problems, that is, through the layering of the consensus mechanism, different levels can be used to meet different needs. The Nervos project it founded is focusing on dividing the underlying infrastructure of the blockchain into two layers, layer1 and layer2. layer1 does not focus on efficiency, but only on security and optimization for the upper chain, while layer2 focuses on efficiency, performance and ease of use.
In Tai's view, not all news needs to be globally agreed, so not all things can be done on Layer 1, and a lot of things can be done on Layer 2. "Consensus is hierarchical. Some consensus needs to be reached on a small scale, and some consensus needs to be reached on a global scale, and there are very few consensus that require a global scale."
Tai said that because the current Layer 1 is not very well adapted to Layer 2, it is difficult to implement the Layer 2 solution. And Nervos will develop Layer1 and Layer2 at the same time. Compared with the Bitcoin public chain, layer 1 of Nervos will start with the improvement of the transaction model, generalize UTXO (Unspent Transaction Output, that is, unspent transaction output), and add a data process.
secondary title
The following is the transcript of Tai's speech:
Hello everyone! I'm a little nervous. First of all, I'm very happy to be here to introduce our project, called Nervos. The pronunciation is very strange, and our own people often say it wrong. The topic of my speech is the next-generation blockchain network born for Layer2.
First of all, let me introduce myself. My name is Tai, and my name is rather strange. I am a programmer. I have been writing code for almost ten years. I have been writing code for a living. This year I write a little less.
This impossible triangle seems to be mentioned a lot, so let's start here.In distributed systems, there is a so-called three-energy problem: it is difficult for you to obtain security, performance, and automation at the same time.It is meaningless to lose a secure system, and as a blockchain system, it is difficult for you to compromise on security. The remaining question is, how do you make a trade-off between performance and decentralization? We all know that this must be a trade-off, not both.
Let's look at some trade-offs from some current projects. The POW I am talking about here, think about it in your mind that this is the blockchain made by Satoshi Nakamoto. What he gave up is performance, and he chooses extreme security and extreme security. Centralization, in this way, a blockchain with relatively poor performance is obtained, which is indeed very poor.
Then let's look at the right side, DPos, how does it make trade-offs? It gave up some decentralization in exchange for some performance, so it got relatively good performance, but gave up decentralization. Is this good? It's hard to say, but it's an interesting choice.
We are in a system, when you do consensus, how to do a perfect Train-Off? We find it very difficult, how do we do it? Solve this problem in a layered way:In Layer1, it only does verification and mature consensus. Layer1 chooses extreme security, so we use POW; Layer2 has many kinds, it can be centralized, it can be a mobile phone, no matter what, the state is generated, In Layer2, it can also be improved to very high performance.we think,Problems should not be solved at one level, but at multiple levels.
We can also look at it from another perspective, not all messages need to be globally agreed, so not all things are done on Layer 1, but a lot of things are done on Layer 2, for example, I will send you a message It's OK as long as you know it, not everyone.We believe that there are levels of consensus. Some consensus needs to be reached on a small scale, and some consensus needs to be reached on a global scale, but there are very few consensus that require a global scale.
Why do we have to do Layer1 and Layer2 at the same time? We think about it this way. Now all Layer1s are not considered for Layer2 when they are actually created, but this cannot be blamed on them, because this is due to time. When they were created, there were not so many Layer1s. . The Layer1 we are doing now is to design very well for Layer2, and we can design the protocol for interaction between them at the same time on the first day. In terms of program, this is of course eating our own dog food, so we simultaneously Can do Layer1 and Layer2, but will keep communicating with colleagues who do Layer2 at any time.
What does our Layer1 look like? Everyone look at this Layer1 CKB (Common knowledge Base). We start with a question, why is it so cheap for us to store a piece of data in a database, but so expensive to store a piece of data in a block? Because your data on the blockchain is a global consensus,Global consensus data is inherently very expensive. No matter how time goes by, its absolute cost will decrease, but its relative cost must be relatively high.
What is Common Knowledge? This is an economics vocabulary. It basically means that you know, I know, and I know you know that I know. This is a matter of consensus. You may have played many such games, such as "Pirates Divide Diamonds", "Green Eye Island" and so on.
So
SoCompared with the blockchain created by Satoshi Nakamoto, CKB has only taken a small step forward. It generalizes UTXO (Unspent Transaction Output, that is, unspent transaction output).UTXO has LOCK and AMOUNT, and CELL has one more data, which is only a very small step on the basis of UTXO. This is one of our ideas, that is to say, we think that Layer 2 will shine in the future. We found that there is currently no very good Layer 1 tailor-made for them, so we hope to make a very good Layer 1 to support Layer 2.
OK, that's all for my speech, thank you!
