On the Differences between PoW Consensus and PoS Consensus


In Secret Ape Classroom 16-25, we introduce two parts of the technical implementation of building an encrypted economic network, namely Cell model and CKB-VM virtual machine. In this issue, we will enter the consensus part of one of the three pillars behind the construction of encryption economy (the three pillars of encryption economy: technology implementation, consensus agreement, economic model). In this section, we will explain why the Mystery Ape Technology Team firmly chose PoW as the consensus protocol for the underlying public chain CKB after thorough study of PoW and POS and other consensus.

Secret Ape Science and Technology Block Chain Classroom No. 26

There are too many misunderstandings about PoW and POS. It’s a very difficult task to make clear in an article. So here we just discuss two points. One is whether the promises given by PoS are reliable in economics, the other is that PoS is essentially a licensing mechanism. The reason for these two points is that many discussions are very entangled with some technical details, ignoring these two fundamental issues.

Cost Paradox

Block chain technology provides de-centralized security, which can be measured by attack cost. How high the attack cost of a chain is, how secure it is.

Consensus nodes in block chains provide security products, and consensus nodes are producers of network services. We also know that there is a basic law in economics. In a fully competitive market, the marginal income of producers in equilibrium is equal to the marginal cost. That is to say, if the consensus algorithm of block chains creates a fully competitive market, how much security is provided, how much production cost is needed:

Attack Cost = Security Degree = Production Cost

How much marginal security is to be provided depends on how much marginal cost is to be paid. Now PoS tells us that we can provide the same amount of security, but it only costs 10% or even 1%. Don’t you think it’s strange? Why is it so easy for PoS to break the economic law?

In one case, the actual cost of a PoS with the same level of security is the same, only part of it is hidden. Take DPOS as an example. If you want to be a Validator, you need to collect enough votes to vote for yourself by some means and go to the list of the first X Validators. What is the cost of collecting enough tickets? Is the cost the same for different people / organizations? If not, is it a good thing or a bad thing? Is this cost an open information? If it’s not a public information, is it a good thing or a bad thing?

I like a sentence very much:

Financial institutions make people feel safe by hiding risk behind
layers of complexity. Crypto brings risk front and center and brags
about it on the internet.

It is PoW’s characteristic to expose the cost to everyone in the clearest way.

PoS is not really Permissionless

In another case, PoS did not create a fully competitive market.

In PoS, future consensus groups are determined by today’s consensus groups. Any new node wants to participate in the consensus through at least one transaction (e.g., mortgage, voting, etc.). Whether the transaction is handled or not is determined by today’s consensus group. They can handle the transaction or not. If they do not handle the transaction, the new node will always do so. It is not possible to participate in consensus. At the same time, “not dealing with transactions” is easy to disguise and difficult to punish. I have not seen a solution to this problem in the consensus agreement. PoS often allocates block proportions according to the weight of the Stake (regardless of the number of nodes). Considering the centralization of stakes in most systems, this is a very practical problem.

PoW is a thorough Permissionless. Whenever you want, you can buy miners and electric power to join the ranks of the producers, without the need for today’s miners to give you any form of permission. You might say, I still need to buy mine machines and electricity. Is this a form of license? Yes, on a lower level, it’s also a license. Unfortunately, among all roof of XXX, this is the most decentralized form of license. After all, the degree of decentralization of mining machinery production and power resources is much higher than that of Tokens. We should always pursue the decentralization as far as possible, otherwise we should use the centralized system.

Perhaps another person asked: If you design a PoS that can participate in consensus as long as Token is there without sending any transactions to the chain, doesn’t that solve the problem you are talking about?

This will lead to two problems:

  1. If Token is there to participate in the consensus, it means that users have no commitment to participate in the consensus, no cost and penalty (users may not be online at all). Such a consensus is very difficult to design.
  2. There will be a Nothing at Stake problem. I won’t say much here.

Layer 1 must use PoW

Layer 1 is the foundation of the cryptographic economy, and the block chain that assumes the role of Layer 1 must be a network without permission, because Layer 1 must be a globally shared, neutral facility, like the Internet, and the “need for permission” means that it is controlled by a group of people, which is fundamentally in conflict with this goal. From the above analysis, we can draw a conclusion that if we want Layer 1 to be as decentralized and secure as possible, Layer 1 has only one option, PoW. PoS is not unavailable, but not suitable for Layer 1.

More discussion

These two days, in the “Orange Book” micro-messaging group, there was a great discussion between PoW and POS, which was very exciting. From the discussion, I saw many valuable points of view.

Since Wechat Group itself is a very inappropriate tool for discussion, after discussion, many replies were ignored, many questions were repeatedly raised and said that the efficiency is too low, so here I will sort out the questions raised in the discussion, reply once in a unified manner, to avoid duplication of work.

A: The capital gains made by the big POS household sitting there are insurmountable divides for the small scattered household, and they cannot restrict their rights.
B: It’s a normal social phenomenon. It’s easier for the rich to earn money than the common people. It’s unfair for them to invest so much money and earn less than others. Now we get the same proportion. If the big households invest more money, they will get more returns.

This is a common conversation in PoW and POS discussions. A raised two questions:

  1. (Fairness) PoS households sit and make profits, but ordinary people do not;
  2. (No restriction) The rights of large PoS households cannot be restrained. B answered 1.

With regard to question 1, I agree with question B. It is reasonable to invest more and return more, whether it is PoW or PoS. Block chains are tools to help achieve process fairness rather than result fairness. We still see Matthew effect in the block chain. The consequence of attempting to solve the result fairness with block chains is that the connotation of block chains governance is infinitely expanded, and all the problems that should be solved by the agreement (as you will see later) are put to the vote to solve.

However, both history and theory have long told us that there is no perfect group selection system in the world. Block chain is a tool that can record data, ensure that data is not tampered with, and provide data for all. It has been a very big progress to help us achieve process fairness better through such a tool.

Question 2 is a key issue that has been neglected in the discussion. It’s natural to invest resources in return, but only if you want to invest. For example, is it possible for any VC to invest in the financing of a star entrepreneurship team? PoW has a very good openness, so that subsequent consensus participants can always join the consensus group, while PoS does not have such openness. Referring to the discussion here:

In PoS, future consensus groups are determined by today’s consensus groups. Any new node wants to participate in the consensus through at least one transaction (e.g., mortgage, voting, etc.). Whether the transaction is handled or not is determined by today’s consensus group. They can handle the transaction or not. If they do not handle the transaction, the new node will always do so. It is not possible to participate in consensus.

At the same time, “not dealing with transactions” is easy to disguise and difficult to punish. I have not seen a solution to this problem in the consensus agreement. PoS is often based on Stake
Weights are used to allocate block proportions (regardless of the number of nodes), which is a very practical problem considering the centralization of Stakes in most systems.

PoW is a thorough Permissionless. Whenever you want, you can buy miners and electric power to join the ranks of the producers, without the need for today’s miners to give you any form of permission. You might say, I still need to buy mine machines and electricity. Is this a form of license?

Yes, on a lower level, it’s also a license. Unfortunately, in all roofs of XXX
This is the most decentralized form of licensing. After all, the degree of decentralization of mining machinery production and power resources is much higher than that of Tokens. We should always pursue the decentralization as far as possible, otherwise we should use the centralized system.

“Can you invest” itself can be regarded as an Option, which has very high value. PoW can have such a nature, because the calculation of workload proof is a kind of calculation that does not depend on history. Whether you buy calculating power to join the calculation at any point in time, you and other miners are on the same starting line. This is a very unique and counter-intuitive nature, which makes the latter always be able to break. The advantages of early participants are also discussed.

PoS is different, because Stake is an endogenous asset in the system, the ownership of assets is determined by the system history, and the order of transactions is determined by the system history. Therefore, whether it is a POS that needs to mortgage Token or a PoS that can participate with Token, its Validator set is the system’s own history. History decides that PoS and PoW are totally different in terms of “whether to participate in consensus depends on history”. It’s easy to see that this is essentially different, and no matter what kind of upper design can solve this problem.

Therefore, the design of PoW and POS is fundamentally two different ideas, reflecting two different concepts:

  1. (PoS) The system should give the pioneer a natural advantage
  2. (PoW) The system should not give the pioneer a natural advantage

Note that the subject is “system”. Consensus participants’advantages outside the system can not be solved by any protocol design. If you think 1 is right, you should naturally support PoS, and vice versa, you should be more in favor of PoW. CKB chooses PoW in its design because its design goal is Layer 1, a common infrastructure all over the world. We hope that it can run for a long time and neutrally. To achieve this, the system should not give the pioneers (including the designers themselves) natural advantages.

A: Isn't PoS buying Staking just putting it in?

It’s investment, but the “return” in the “return on investment” here has changed. We need to look at the return on investment separately. One kind of return is Token’s return. The recent PoS basically supports Delegate, so basically everyone has this Option. The other is the right to participate in consensus. According to the answer above, it can be monopolized by the existing Validator, and most people do not have Option. There are many kinds of rights, and dividend is only one of them, and not the key one.

So why is the right to consensus important and what is its use? It means that you can sort transactions, and the order of transactions determines whether your transactions can be linked up in time when the chain is crowdsourcing, whether your orders can be completed in time when trading in DEX, and so on. Do DeFi on PoS? You need to think carefully about what Validator itself does, and whether it’s related to the transaction you send?

We know that DeFi’s transaction is likely to be a very high value transaction. There’s a huge bill in e.g. DEX. What if Validator arranges a queue jump? Just the different order of transactions can result in huge profits, let alone consensus nodes can do much more than that?

The smart one would think, “Isn’t that the same problem with PoW’s miners / pools? “Yes, PoW miners have the same rights, but PoW has two advantages that can weaken the problem:

1. Output nodes and key users in the ecosystem are decoupled. Key users in the ecosystem refer to service providers such as exchanges, wallets and so on, which provide high-quality services for a large number of users and aggregate a large number of users and transactions.

In PoS, because of the large number of users and transactions, Stake will naturally concentrate on key users in the ecosystem and form a natural take Pool. Therefore, the business advantages of key users can be transformed into the advantages of consensus and governance (which are already evident in some chains), making the advantages of pioneers. More intensive.

In PoW, the group of miners and the exchange/wallet are independent. They have different professional division of labor. They get rewards through different professional knowledge and different ways. Key users can not turn their business advantages into advantages in the agreement, and miners can not turn their advantages in the agreement into superior business. Advantages. In PoW, there are checks and balances among developers, users and consensus nodes.

2. PoW’s consensus is open (see above), full of fierce competition. Maybe a miner can do this in a short time, but because new miners are always free to join, it is very difficult to do this for a long time. Intense competition will create an increasingly fair and fully competitive market, which takes time (30 years)? Maybe.

Conversely, in PoS, due to the natural advantages of the pioneer, as well as the binding of business advantages and consensus advantages, the advantages of the pioneer will only grow larger and larger, competition will gradually disappear, and finally form a monopoly or oligopoly. In infrastructure-level agreements, we should try our best to avoid monopoly.

Block chain itself is a large queuing machine, and the right to decide the order is the most critical right in this system.

A: (Bitcoin) Fifty-one-hour attack cost $443,000...

What A wants to say is that Bitcoin is not safe because it can be attacked for only $443,000, and then PoS Token is limited, no attacker can buy enough Stakes from the market to attack.

This view ignores the problem that at some point, the power of the earth is also limited. If a PoW chain has only 10% SHA 256 power, that’s fine. But if Bitcoin already concentrates 90% (estimated value) of SHA 256, where do you get another 90% (estimated value) of SHA 256? Quantitative change will cause qualitative change, and the change of arithmetic status will affect security.

It’s not PoW that’s unsafe, it’s a PoW chain that doesn’t get enough power. Block chains using PoW will encounter larger start-up problems than those using PoS, but it is such a real and cruel test that can prove the safety of block chains. Otherwise, I also run a PoS chain, 99% of Token belong to me, 1% want to fry in the market how high, security is not more than Bitcoin in minutes? Bitcoin has been running for 10 years. It carries so much value. Why hasn’t the attack happened? On the contrary, some chains have been running for less than a year, and intelligent contracts with hundreds of thousands of dollars in value areas have been stolen countless times… Theories need to be constantly revised with facts. When theories do not match facts, they must be wrong.

Similarly, there is an upper limit. The upper limit of resources needed to participate in the consensus in the PoW chain changes with time. The progress of science and technology, human enterprising spirit and fierce competition are constantly pushing forward, and the acquisition of natural resources needed is completely decentralized. The upper limit of resources needed in the PoS chain is stipulated by the agreement. Isn’t it familiar that the additional Token is all in the hands of the existing Validator and distributed through the sale of Validator in the market or the dividend of Staking Pool?

A: Both of them are abstracted to the end. PoW calculates the cost of capital with capital expenditure and Opex; PoS calculates the cost of capital with the market value of collateral. Both costs are irreversible.

I don’t agree with this logic. Ignoring the middle process to the essence just ignores the key. Process is the key, process will produce friction, process will produce loss. Even if it is all capital, the liquidity of capital and the speed at which monopoly is generated are different. Whether agreement can be recovered from monopoly is also different. See the preceding article.

A: I don’t think the holders of PoS will keep selling.
A: Who will contribute to the ecosystem if we concentrate too much?
A: If 90% of the chips are in your hands, the ecosystem will not be able to continue.
A: For you, a high concentration of money is of no value.
A: You all monopolize 90% of tokens. No one pays you the rent.

Monopoly can also build ecology. Tencent and Apple are examples. Whether it is historical or economic principles or “From 0 to 1” clearly tells us that monopoly can only make windfall profits.

At the same time, the existence of monopoly does not mean that monopoly exists, you know. Token is the most liquid capital in the world. Even if I own 90% of Token, I will diversify it into 100 Staking Pools instead of focusing on one Staking Pool. Monopolys don’t like to jump out and say, “Hey, I monopolize this system! “

A: Potential evil monopolists will sell Token for short-term benefits.
A: To do evil is for short-term benefit.
A: That is, potential monopolists seeking short-term interests will sell Token for short-term interests.
B: The head mortgager is clearly seen. If the mortgage is released or the money is withdrawn to the exchange, it will cause the price to fall. It is already reflected that he hasn’t broken it yet.
A: That’s what I mean. Isn’t it safe for monopolists not to do evil?

Here the concepts of monopoly and evil are confused. Evil refers to an explicit attack, such as a double-spending deal. Evil can be observed, and the system or ecosystem can react accordingly. Monopoly is implicit. Monopoly does not need or will not attack the system, but it can still use its consensus rights to gain more benefits. As mentioned earlier, as long as you can control the order of transactions, you can control everything.

Manipulating the order of transactions is undetectable. In PoS system, manipulating transaction sequencing also means manipulating future Validator sets, which means that monopoly status can be easily maintained. This is the inevitable result of ensuring system security according to system history, which does not exist in PoW.

We haven’t yet found a way to eliminate all monopolies at all times, but PoW has given us at least a design that makes monopolies difficult to exist in a longer time dimension, which I think is important for Layer 1.

A: First, why can’t the latter of the PoS chain participate? Buying money is much lower than buying miners. I disagree with the conclusion that there are monopolists in PoS for a long time, and there is no economic scale effect. Secondly, comparing the cost of attacking PoS, acquiring Stake is only one aspect, and Reputation system. The nodes in PoS are very concerned about their Reputation, which is cost-effective. The point is huge.

In the first question, the question of monopoly has already been answered. On the threshold issue, I think it’s a mistake that many people, even protocol designers, often make. The primary goal of block chains is security and de-centralization, and the threshold to ease of use is not the goal of block chains.

Discussing threshold issues in block chain protocols is like saying, “How do you let ordinary people construct TCP request packets”. It confuses different levels of goals. To lower the threshold and improve usability, we can do a lot of work at the top, do wallets, do cloud mining, design various financial products. Why should we consider the threshold in the block chain agreement?

Nervos pursues a layered protocol architecture because it sees that ease of use and security issues must be considered separately. There is a contradiction between ease of use/threshold and security in essence. Together, we will get nothing. In the future, the direct participants of block chain protocol must be professional users, who can obtain the support and benefits of ordinary users in the ecosystem by building (trusted) services, lowering the use threshold and providing ease of use.

Second, Reputation is something that cannot be quantified or judged by block chain protocols. Placing the security of block chains on Reputation will only bring block chains back to the old way of existing trust system. At the same time, Reputation also has the characteristics of non-transferable. Would it be familiar to build a security model based on a non-transferable thing?

A: Actually, there are many implicit assumptions. If a PoS chain has a Cartel-controlled Token of 1/3, the community can fork out cartel through a hard bifurcation. I don’t think a PoW chain is controlled by 51% computing power and faces the same serious problem. In addition, I think the node hides for the last long time. There is little possibility of majority control.

Pushing the unsolvable issues of the agreement to out-of-chain governance (note that when Cartel controls 1/3 of the state and poses enough threat to the community to want Fork, on-chain governance is no longer useful) and hard bifurcation can really solve all the problems, but this should be a costly last resort and should not be the last resort. As a tool for casual use, block chain protocols should avoid falling into this scenario as much as possible.

Using out-of-chain governance and hard bifurcation is equivalent to acknowledging the inadequacy of the protocol, which requires someone to take over. I agree that block chain ecology ultimately needs human governance, but I think the lower the frequency of human intervention, the better. If we do not pursue this point, why do we need block chain? Only by reducing the frequency of participation, the cost of collaboration automation can be reduced, and the trust foundation needed for collaboration can be reduced.

“It’s not possible for nodes to hide for a long time to form majority control” – as long as the time is long enough, no matter how small the probability of time will occur. The black swan in the financial market has told us countless times. May our memory be more than seven seconds.

A: The chips of XXX are dispersing.
A: XXX has more than 1000 wallet addresses from ICO, less than ten thousand in a month.

According to the address of Top 100 or the degree of dispersion of Top 20 Staking Pool, it is impossible to prove that Token is decentralized. The reason is simple. We’re talking about systems that don’t require permission. Addresses represent only a public-private key pair, not an identity. Generating addresses is almost cost-free, and the number of addresses that hold them does not represent the different users who hold them. Don’t confuse addresses with users.

In PoS, the number of Validators in the Validator set is irrelevant and does not represent Stake dispersion. As mentioned earlier, if you have 99% of your Suites, you should spread them over 100 or even 1,000 Validators.

PoW also has the problem of concentrating its power in the hands of large miners, but because PoW’s openness and system do not create disadvantages for later generations (see above), such concentration will only be temporary, and the power will continue to shift from one hand to another in the fierce competition. PoW praises competition. PoW is an open system. Only an open system can stay away from the end of thermodynamics and maintain long-term vitality.

PoS has its value and problems, so it can’t be applied to all scenarios. Layer 1’s block chain protocol must use PoW. Only Layer 1 of PoW can solve the problems we want to solve and realize the future we want.