Inflation has a direct impact on the security, token distribution, and incentive design of any blockchain. When it comes to EOS specifically, we think the inflation debate is oversimplified. Inflation can have good results, bad results, or not much, depending on where it goes and how it is used.

The purpose of this article is to explore the problem of EOS inflation to confirm:EOS inflation that is not given directly to validators (consensus participants) may have obvious benefits, and inflation that is given only to these participants may have obvious negative effects.We hope this discussion helps drive community conversation about this issue, and would very much appreciate feedback from voters to further discuss it.

1. The centralization effect of cryptocurrencies

Properties of cryptocurrencies: easily analyzable, mathematically designed micro-economy with different properties, often pre-defined and irreversible. For example, Bitcoin will only issue 21 million before stopping inflation. This makes the analysis very unique - we can analyze these closed economic systems accurately and with great precision, looking at properties such as velocity and distribution of money as a whole, with all information publicly available. In this post, we will try to analyze changes in centralization characteristics.

There are many factors that contribute to the concentration of power that a cryptocurrency system can produce - in addition to those we usually think of, such as wealth accumulation.Using Bitcoin as an example, we have observed that miners concentrate on choosing large mining pools. In PoS (and DPoS), we see a similar phenomenon; due to a small number of technically capable potential consensus participants, the ability to achieve decentralization is limited by these capable participants. There are debates like this going on around Bitcoin all the time.

Supporters of "small blocks" believe that large blocks will always place too much data burden on new miners, while supporters of "big blocks" believe that technology is improving, and this burden will decrease with time and technological progress. In reality, the right balance might be somewhere in between.

The investment in the startup phase and the accumulation of long-term investment will undoubtedly increase the centralization of Bitcoin, because new miners cannot afford to verify the system and the number of potential consensus parties is reduced. However, care should be taken to find the right balance between technology costs that don't burden new miners and a usable system that doesn't drive them away.

In a PoW system, current rewards will not affect future rewards, because the ability to achieve PoW is an external factor. This is what makes PoW systems "permissionless" - their consensus mechanisms are not closed systems and require external factors (in the form of electricity and 'work').

For DPoS systems, their reward structure is very different compared to PoW. In many DPoS systems, the token supply is not fixed but inflationary. This means that current rewards actually affect future rewards. If we reward those responsible for the consensus with more influence on the consensus, then the degree of decentralization of the system will be affected in the long run.

To understand this problem, we need to take a step back to understand more clearly how a property like inflation works in a closed system. As an example, suppose there are two people, Alice and Bob, who each have 100 units of currency. If Alice and Bob each get another 100 units, now each has 200 units of the currency, does that make a difference? In a closed system, we do nothing - Alice originally owned 100 of 200 units (50% share), now owns 200 of 400 units (still 50% share).

The same level of inflation produced no change other than the number of units of account. In terms of "money supply percentage" reference, inflation completely changes the percentage. If we give Alice 100 but not Bob, now Bob still only has his original 100, but Alice has 200. We dilute the value of Bob's holdings with "inflation" - his holding ratio has changed from 1/2 to 1/3 now, and Alice's share has changed from 1/2 to 2/3.

It is worth noting here that we must understand inflation as a change in the proportion of the total supply.Inflation makes one party gain, and there must be another party that suffers.

So, how does this actually affect blockchain systems? In DPoS, control of consensus is determined by voting volume. In simple terms, those who hold the most will effectively control the consensus. However, what we reward to those consensus participants is to dilute the reward with inflation to those who do not control the consensus.

Imagine that Alice, Bob, and Charlie each own 1 unit of currency with a total supply of 3. If they elect Alice and she increases her supply by 1, and rewards herself with it, she now has two votes (compared to Bob and Charlie). It is now impossible for Alice to vote for herself to be considered invalid.When the consensus is controlled in a closed system fashion, eventually, the control of the current system rapidly develops into the centralized control of the future.

We might call this a form of "asymptotic convergence", whereby the aggregated supply by a party (as a fraction of the total issued supply) increases asymptotically in a closed system (i.e. , which gradually gets closer and closer to the entire issued volume). Although we are only using inflation as an example, this effect can also occur with fixed or variable fees and variable inflation, although the calculation will change slightly.

3. Gradual concentration of consensus

When token supply is important to consensus, and inflation is given to those who currently dominate consensus, then over time inflation can lead to centralization of the system. This is Asymptotic Consensus Convergence.

Of course, real-world systems are far more complex than our simple examples. Consensus rewards are not always all hoarded for backup. To improve understanding, let's use a physical analogy. Suppose we have a box in the middle of the warehouse. If a force is continuously applied to the box in a single direction, the box will move in that direction. If this force is counteracted, the box will not move. In this analogy, progressive centralization of consensus is when a fixed force is exerted on the currently elected consensus participants. This force can be counteracted, but it needs to be done voluntarily.This force is completely neutralized if consensus participants always sell all their rewards.

However, as long as there are consensus participants collecting rewards, then the box will move a little bit. The more you do this, the more the box moves. This strength can be defined in terms of rewards:

As an example, if we assume a 1% annual reward (via inflation), then this would produce a curve like this:

(The curve can be drawn by entering "(1.01^y -1)/(1.01^y) for y=0 to 100" in https://www.wolframalpha.com/input/)

From the graph we see that the percentage of control over the system (that is, control over the newly created supply compared to the total new supply) went from zero to less than 10% in 10 years, and 100 years later About 60%. This means that, in such a time, the maximum rate at which the box can move in our analogy is at worst that number (only by this applied force).

We use the force analogy for a very important reason: counteracting the push just delays the immediate result, it doesn't prevent future effects - for example, consider fighting someone who keeps kicking boxes with their feet: although Doing this might temporarily stop them, but it won't stop them from constantly animating the box.

While that doesn't sound like an issue today, the numbers should be cause for concern. In EOS, the number of votes for headnodes is currently only about 12% of the total supply (Note: when translating this article, the tokens voted for the top 10 nodes accounted for 17% of the total supply). However, votes in EOS have a number of 30 that has a multiplier effect: if a group of 30 people agree, then they can collectively pass this number to increase their vote count (via voting transactions).

In this case, a collective can gain control for up to 5 months by voting for a different collective by gaining more supply (and hoping that the new collective will voluntarily and selflessly oppose Concentration of power) can consolidate its position. (Note: Each account can vote for 30 nodes)

In fact, the numbers get quite complicated when multiple competing collectives vie for a single dominant control of consensus. With some effort, though, it can be shown that the group with the most power (in particular, the group with the most votes from the currently dominant group) will eventually win. For those not in the most powerful collective, this presents a prisoner's dilemma - if you choose to join the collective, you may be "rewarded". If you say no, and they are able to join forces with others, then you get nothing.

4. Let the centralization of power fail

If you were worried about players with concentrated power securing their positions due to inflation, then this confirms the thesis. However, we should not only think about how this problem exists, but also how to solve it.

There are some extreme measures that can be taken. Such as completely removing inflationary rewards for consensus participants. This does solve the problem, but it does not incentivize nodes to work for the network in the first place. Another attempt is to separate token properties: voting tokens (fixed supply), and utility tokens (inflationary supply).

In theory, consensus participants can only be paid using utility tokens, so they don't gradually accumulate more voting power over time. However, this won't actually work! Because as long as they can exchange these tokens in the market, this method can be invalidated.

However, completely eliminating the expansion of voting rights is not the only solution. In fact,If there is a second source of inflation added, an equal but opposite force, then centralization of power cannot be achieved (or even leads to eventual decentralization)!

In the equation above, rewards refer to the amount of supply consensus participants earn relative to the total supply. If we were to increase the total supply in another way but not give it to the consensus participants, then their control over the system would not increase. Look again at the example of Alice, Bob, and Charlie from before: Suppose everyone has 1 unit of currency, and Alice gets an extra 1 as before.

But now, suppose Charlie also gets an extra unit, so Alice's extra voting power is offset. For other purposes, this will be in the form of additional inflation. As with all inflation, there are winners and losers: in this case, Bob's voting power is "diluted" and loses - but perhaps he is also the least or no participant in the system.

However, it is more difficult to find a good reason to introduce a second source of offsetting inflation. In the original inflation problem, if we only consider the percentage of ownership and not the arbitrary number of units, then inflation is just a tool to adjust the ratio between each other at the expense of one party in favor of the other party. Therefore, we need to elect a winner (elect new rewards to some people who are not part of the consensus participants) and new losers. You might think of an obvious solution - we can pick those who are not BP to be the winners! But this will be a precise offset, which is equivalent to not giving BP rewards in the first place.

If you have been following this logic, then you know exactly what the next solution is: the Worker Proposal. In fact,The solution to this problem is actually a second source of inflation, which is diverted by token holders to non-block producing parties.This may include funding for applications and services, grants and scholarships, and other costs that may be incurred to increase the value of the overall network.

It is important to strongly encourage block producers not to double down on these funds, because if they do, then the approach is moot. With this second source of inflation, the force of centralization is counteracted and may actually be a force for decentralization. If the inflation going to the community is greater than the inflation given only to consensus participants, then, while consensus participants may still be rewarded for their efforts, their status is no longer the most attractive factor for obtaining rewards.

The worker proposal system can work in a number of ways. There is a very simple approach: replicate the structure of the block producers, but only assign wages by weight, not pay for the size of the block. This second group will be chosen from non-BPs, not participate in the control of the consensus, and can get their own inflation - determined by token holders vote.

This would be great for organizations that aren't very technically capable or good at governance, and they can then spend their time doing what they're good at; whether it's community outreach, marketing, scholarship, or other EOS business projects. In fact, with this structure (or a similar one), consensus participants become more obligated to community decisions, as more voting power is transferred to the community over time.

5. Summary

In this post, we highlight three important points to consider when using inflation in a closed consensus system such as DPOS.

1. Inflation itself is just a tool to create winners and losers based on ownership of total supply.

2. If we give rewards for having more influence over the consensus to those who are responsible for it, then we end up giving more power to those who currently hold power.

3. It is very important to prevent power consolidation, and one of the best ways to avoid this is to apply corresponding constraints - ensuring that a single party or group is not the only recipient of the reward.

As the EOS community continues to discuss the pros and cons of various configuration options for EOS inflation, it is important to remember that changes to it can have long-term and second-order effects. While directly reducing the amount of inflation may seem beneficial to token value, doing so ensures the long-term health of the network is achieved. We believe that short-term operations can create changes in market prices, but only by considering long-term development can real value creation be promoted. In order to achieve this, we must carefully consider how changes to important parameters will affect the centralization of the network.

-END-

Disclaimer: This article is the author's independent point of view, and does not represent the position of the Blockchain Institute, nor does it constitute any investment opinion or suggestion.

Disclaimer: This article is the author's independent point of view, and does not represent the position of the Blockchain Institute, nor does it constitute any investment opinion or suggestion.

Original link: https://blog.greymass.com/eos/@greymass/inflation-centralization-and-dpos