On the problem of delegation centralization


#1
  1. It is economically forced for big validator with large delegations to become continuously bigger. “Economy of scale”

  2. Although there exists several technical and incentive oriented attempts to make delegations more decentralized, spinning up new node can sidestep most of the attempts.

  3. Therefore I think the problem of delegation centralization is more about governance issue.

  4. Although we can solve it when it happens in mainnet, it will be a messy chaos discussion when it actually happens because it is directly affect each validator’s business quite severely.

  5. Therefore I suggest a healthy discussion here about ground rules of prevention/post-happened-solution of delegation centralization before we become too greedy in mainnet. Strong and healthy consensus on this topic will help us guide to a better calm way to deal with it when it really happens.


#2

Can you outline, why 1. holds true?

I understand that it is a likely scenario, since delegators give commission to validators. However in a perfect market, where MR = MC (marginal cost equals marginal return), the commission would be just so high, that it covers the expense of running the validator. So in reality of course the return is slightly higher than the cost, but it might be that it takes some 100 years until this makes a noticable difference. But within such a timeframe there are heirs who don’t identify with cosmos and just sell their stake and buy lambos etc. So at some timeframe this will be mixed up anyway.
Furthermore there is another scenario, where big validators don’t charge any commission at all, because for them it is nice to have more voting power, since it turned out that many just delegate and don’t use their individual voting rights.
So I’d like to hear more why it necessarily must happen this way. For me it is possible that someone who wants to validate, because his/her own stake is big enough so that it makes sense, then well allowing delegation with a commission that matches the running cost of the validator nodes, means that the costs are paid by the delegators but there is no increase of stake in relation to other validators.

I totally agree to 4. It makes sense to think about this early, if it is a problem and there must be a solution that cuts off profits from someone, then it might cause a lot of trouble.
What are possible solutions?
a) force validators to sell atoms
b) force validators to donate atoms
c) slash validator atoms
d) lower the profits of the validators commission

So to me option b) makes the most sense, especially if the validator is allowed to decide what kind of donation can be done, for example fund some cosmos based startups. In that case the validator has an economic reason to do that, because he increases the value of the product he is highly invested in. c) and d) are mostly disincentivizing validators, a) can be sidestepped by new nodes as mentioned in your point 2. Of course the donation thing can also be sidestepped, but the question is if you want to do that, once there is a consensus for this solution to delegation centralization. You would lose trust of the community and that is very important to keep.


#3

Thanks for expressing opinion.

  1. Why economy of scale happens?

Good reputation / great skill / better marketing ability / enough capital to spend on / possessing good platform(other mining or exchanges or other blockchain services)

Those will bring more delegation, result in more profitable business, hence more capital to reinvest, richer services and possibly less fees. So it is an infinite loop of better competitiveness and better profit interation.

It is not a bad thing in the process. It is very nature of capitalism. I am focusing more on the side effects, especially more critical in dPoS network. It is happening in most IT service industries in real world so I am not sure why you doubt it to be happened?? Do you know any IT service industry without a goliath?

One counter arguement is that delegator needs risk diversyfication.

First of all, a lot of investors are not wise enough to actually execute that idea.

Also, even though wise investors want diversification, they don’t diversify over 100 validators. They will probably diversify among top 5 validators.

Therefore oligopoly is what we can naturally expect, and existence of a cartel of those top validators is a reasonable assumption when we consider real world situation.

In blockchain, we don’t have any legal, diplomatic method to resolve such problems. And it is difficult to disincentivize by solely technology, so in the end, fork is the last possible choice, which is very painful to execute.

  1. About possible prevention described.

First of all, we should have some self resolve period when a big delegation owned validator can be warned so that they can reduce the delegation themselves without any forced punishment.

Second, it is quite easy to spin up a new validator to avoid certain punishments caused by over-concentrated delegations. Like the case of bitfish, even though it was quite clear, it can be much more unclear which validators are operated by single entity.

Therefore, to prevent such centralization, we need a strong consensus in validator community that if an entity hide their identity and use several validator nodes to gather “harmful” amount of delegation under control, merciless forkout should be executed like GoS case in condition of the agreement of major number of validators in Cosmos. It should be fearful enough so that any entity cannot dare to even think about possessing big percentage of delegation in Cosmos.

In addition, forking out is not neccessarily meaning confiscation of their assets. It is just a unbonding and delete-validator procedure with maybe a little bit of financial punishment on their own asset.(bitfish event was a exceptional case because in GoS, most of delegations are self delegation) Therefore the delegator will not lose any of their assets from the forkout event.


#4

Oh sorry, I assumed you were referring to another effect.

I think this is a very hasty generalization. It is not necessary to generalize IT industry and then answer specific questions for Cosmos, because we know the mechanics of token economy in Cosmos and we can also have influence on it. So let’s do some calculations:

First scenario: zero commission, no running cost of validators
In this scenario everyone gets 7% (or 20%, does not matter) inflationary reward and some transaction fees as well, let’s say it is 3% on average. This value can change quite fast, but it does not matter, it is always distributed proportional to stake, so it can always be calculated as a percentage of what you hold and acts the same way as inflation it is just more volatile. So lets assume there are 3 different actors, a whale with 900 atoms, a medium validator with 90 atoms and a small delegator with 10 atoms. So at the beginning the proportions of ownership are:
whale 90%
medium validator 9%
delegator 1%

Now let’s apply the 7% + 3% of interest they gain, then they have:
whale 990 atoms
validator 99 atoms
delegator 11 atoms

Now we calculate the new proportions:
whale 90%
validator 9%
delegator 1%

This is the important part of the simple scenario (no commission, no running cost) and it means that it does not matter at all how many atoms someone has, the proportions stay the same, regardless if someone gains interest with a lot of atoms or just a few. The rich are not getting rich faster than the poor, all are getting richer at the same speed. If you look at the absolute numbers, yes the rich is getting more atoms in absolute terms, but regarding the share of the whole network, everything stays the same.

Now if we move on to a more realistic scenario, which has running costs of a validator. Then it can no longer be the case, that no commission is being charged. Let’s say a validator has a fixed cost of running the validator, then he has to charge a commission and sell the atoms from the commission to pay the running costs. If the validator doesn’t, then it makes no sense to run a validator, because atoms are being lost in the process. So for simplicity let’s first assume all validators charge the same commission. This means for the delegators it does not make any difference which validator they pick, assuming they don’t get slashed. Instead of 10% from the previous example, they now only get 9.9% or 9%, depending on the commission. The validators now get more than 10%, depending on their own stake to delegation ratio. So if we assume the smallest validators just pay their running costs with commission and the biggest validators have a net profit from commission after running costs are subtracted. Let’s look at the numbers:

whale 900 atoms -> 990.5 atoms
medium validator 90 atoms -> 99 atoms
delegator 10 atoms -> 10.5 atoms

The medium validator grows by 10% as before, since what is earned from commission is sold to pay running costs. The whale does not need to pay the whole commission and gains more than 10%, in this example 10.05%, whereas the delegator does no longer have 10% growth, but rather 5%, so this is a heavy cut. I know that 0.5 atoms do not exist, this is just an example, you can double all numbers and it stays the same though. The proportions of the whole network have also changed, the delegator now has less than 1% and the whale has more than 90%. So here we see a scenario, where the rich are getting richer and the poor are getting poorer. This is what I was referring to in my initial post.

However this scenario is also hypothetical and in a more realistic scenario there will be different commissions among the validators. So then the delegators can switch to lower commissions and thus drive a rallye for lowest commissions. In a perfect market there is a rule of MR = MC (marginal return equals marginal cost), which mostly means that the price will be lowered until the point is reached, where there is a balance between production cost and return. Perfect market mostly means that there is no hidden information. So in the beginning this might not be true, but after some time everyone knows how to run a stable validator and so on. So every validator can lower the commission until the running costs are just paid and in the end all proportions stay the same. Summing it up, global stability is possible but we will see now, why this is not a nash equilibrium, thus won’t be stable if single entities act economically reasonable. If all validators have commissions according to MC=MR, then the biggest validators, with the most delegators have the lowest commissions and the smaller ones have higher commissions, because running cost of a validator does not depend on the number of delegators and is thus fixed. In such a scenario each delegator is incentivized to switch to bigger validators driving delegation centralization, which is bad for the whole network, but makes sense as an individual decision. In game theory such a situation is often referred to as a prisoner’s dilemma. Since the validators don’t want to become to big, since they are afraid to be forked out being a cartel or similar actions, they can actually do something against it and make people delegate somewhere else and this is raise the commission. With higher commission delegators will move to other validators, where they get the same uptime but for less commission. However then we are back to the rich are getting richer problem, which makes the big validators bigger over time.

What I wanted to point out here is that there are 2 economic mechanisms competing with each other. One solution I wanted to point out is that the popular validators (=> high commission possible) really tune up their commission but then donate much of their overcommission instead of maximizing growth of atoms. Of course they can also sell it and buy lambos and yachts, which also prevents them from becoming single controlling entities over time. But if they want to reinvest into cosmos, they can fund projects that want to connect to the hub which they validate.


#5

I really carefully read your post but I think you are 100% agree that 1 will happen without fork…

Is it right? If not, could you sum up just counter argument of 1? Most of your argument actually support 1.

Donation is our hope, which will be not likely to happen. It is not a “solution” or “prevention”.


#6

We will see top performers of GoS attracting the ICO atom holders, probably as the atom holders become a larger group newcomers in turn are more likely to want to delegate to the most popular(in this case larger) validators. How can this be stopped ?

  1. Technical “warnings” in Voyager that a Validator is reaching a threshold , as mentioned this can be circumvented.
  2. Validators refusing to take more delegations , bad actors probably won’t follow this path.
  3. Delegators making informed decisions about the security of the network.

What will influence a delegator ?

  1. Validator has skin in the game
  2. Top performer (Initially measured via GoS)
  3. Validator actually contributes to eco-system

Of course a validator can have all of the above qualities and more.

Most delegators will diversify but this diversification will probably be aimed at the “top” ranked validators.

Delegation will be the key to the success of this system just as much as validation is.
Validators have proven their technical prowess. What do delegators have to prove ?

Can a quiz be implemented into Voyager ? If a delegator fails the quiz they are unable to delegate and must return to retake the quiz in x days. Again with a change of IP (if we collect such details) it may be possible to bypass this and retake immediately.

In an anonymous system how can we identify if a delegate participates in governance/voting and if they do not vote can they be punished ?
However by voting we cannot automatically assume they make educated decisions when voting or selecting a validator.


#7

If we believe in decentralization, this is inevitable we need to deal with the evils of capitalism, which had been sufficiently covered by Karl Marx and proven by the Wall Street.

I suggest independent validators/delegators need to discuss a “gentlemen’s agreement” publicly to conclude:

  1. A set of code of ethics or best practice of being a professional validator and delegator on PoS blockchain
  2. A set of measurable red flags of which if appear, the group of these independent validators and delegators will initiate and support a hard fork

#8

Thanks for the opinions. Possible solutions discussed so far are below.

  • education and quiz(?) on delegators

  • warning to delegators when there exists a big possibility of oligopoly by certain big validators or some other threats to the network

  • proper effective communication channel from united validators to delegators(including detail information on governance issues)

  • bring awareness of each validator’s contribution to ecosystem

  • gentlemen’s agreement on code of ethics for those validators who are willing to join into the group of “good and honest cosmos validators unity”(there should be a better name of course)

Please share more ideas on suggested solutions or new ones.


#9

IMO you guys are being way too logical, too rational, and not nearly realistic enough…
So I’m going to throw some monkey wrenches into this conversation.

In GoS, Bitfish provided a near-perfect example of a what a nefarious validator could look like and who you all should plan for…

They werent communicating, ever. So a “gentleman’s agreement” is cool in the short-term but likely useless in the medium- to long-term as validators quit/change/grow/move on/merge/etc.
And sure the lack of communication from bitfish might have made forking them out easier to do, but what if they were a massive exchange, with a sudden dominant network share of staked Atoms? What if they were also a well-known exchange, with the most daily fiat-to-Atom purchases? And what if their Atom owners were legit crypto noobs who just wanted to earn interest? What if their staked users never needed to touch the Cosmos Wallet, hence wallet warnings are useless because the exchange does it all? Are you all ready to attempt to fork out someone like that?

Continuing on this…
What if the exchange charged 0% commissions for staking? Meaning they now have a massive competitive advantage over most independent validators.
Perhaps they can afford to do this because they know they’ll make it up in trading fee commissions and perhaps they used accounts’ staked Atoms as a sort of ‘sweeping money market’ facility which automatically earned stake for any user account opening Atoms custodied on the exchange.

Perhaps this exchange then starts a trend among all other exchanges, who also automatically sweep any Atom-owning account into the exchange-run validators to earn stake.
Are you all ready to fork out all these exchanges?

I don’t have a clue as to what the correct answer is in all this, but it’s my opinion that there is a massive first-mover economic advantage to whichever exchange is first to implement automatic staking of users’ Atoms under custody. And when, not if, that happens we could see some major strife amongst validators, delegators and potentially AiB and Interchain if we don’t figure it out now.


#10

what if they were a massive exchange, with a sudden dominant network share of staked Atoms? What if they were also a well-known exchange, with the most daily fiat-to-Atom purchases? And what if their Atom owners were legit crypto noobs who just wanted to earn interest? What if their staked users never needed to touch the Cosmos Wallet, hence wallet warnings are useless because the exchange does it all? Are you all ready to attempt to fork out someone like that?

These arguments are what I want to discuss in here. I am aware of those propositions very well. As a honest Cosmos ecosystem contributor, we need to discuss which value is more important to us.

First, we need to clarify the forkout scheme. When we did it in GoS, we burnt all coins possessed by the cartel. But, definitely it will not be the case in mainnet because most of the delegations from big validator are from other delegators(not self-delegation). Therefore, non-self-delegation atoms will be ONLY UNBONDED. So, it is not kind of some extreme situation. We just remove the problematic validator and go on the blockchain. We might decide how much of the self-delegation of big validator has to be burnt as a punishment. It is subjective matter which needs to be discussed.

Commission rate does not matter in this argument. Whatever the rate is, the important thing is the centralization caused by single entity. It is the only thing we should care about. 0% commission does not justify centralization.

Blockchain has a characteristic of decentralization as a necessary proposition. It is different from other traditional internet industry, which does not have decentralization as a must. Therefore imagining economic justice by comparing traditional internet business is not very advised imo.

In short, commission rate, honest willing, business advantages, all of these are irrelevant when we decide the centralization caused by an entity is harmful or not. If it already caused it, it is already harmful no matter what reasons behind it.


#11

Interesting. Explain these ramifications more. Wouldn’t they just rebond? Or would they forever be unbonded? And if that’s the case did you just create 2 types of Atoms, those which can and can’t stake?


#12

They can rebond, it is possible. We need to discuss deeper about how to deal with this. We might give all the delegator little bit of punishment, like 1% or so. But it is just an idea which should be decided by consensus of the community.

I assume this topic is not a simple problem to solve. We need proper group with many honest validators which can give information and rationales about decentralization of the network and its threats. It will be a long journey to the new normal of decentralized blockchain economy. Nobody never been there yet.


#13

Bharvest, I love that you are a worry wart, and always take the worst case scenario in mind when building or suggesting ideas, but in this case, there is quite honestly little we can actually do to prevent the above from happening.

There -are- things we can do to help minimize them.

  1. Yoyager will unlikely issue warnings against X validator, it’s simply to “hands on” for a decentralized network.

  2. a Wall of rules, or a quiz, will be skipped no different than a terms of service on a software install.

  3. Greed all to often outweighs good deeds.

With that said, I think that there are a few good things that could come of this discussion. While I think that we should build a group of technically capable validators, bring them together in a public facing channel (where public can see the chat, but only verified validators can actually communicate). From there and with some mainnet experience under our belts, we can build an expectation of the validator community.

I personally think governance to decide if someone can or can bond atoms should be a last resort, and that we should use governance to not -remove- a validator, but rather as a recommendation of commission rate to discourage further delegations. I’m sure we could develop a sliding scale, and if all of our websites has an explaination of the problems of centralization, and the expectation of increasing commission values based on network voting power we can “train” the delegators in what to watch for, why to watch out of it, how to redelegate etc.

Many of the people in our validator chat are devs/people of technical backgrounds and are rooted in the industry and the concept of what we are doing. The biggest mistake I’ve seen in the crypto community is developers thinking the people that are appealing to understand/care/comprehend what they are doing and why they are doing it (which is fueled by their immersion in a field of other highly technical people). Unfortunately, a large percentage of our clientele barely have the ability to understand circulating supply much less an understanding of BFT consensus.

There will be a large number of people who will have read just enough to know that more atoms = more proposals and therefore more rewards, and will be thinking "if i am getting 7-20% from returns, and only risking 1% to slashing (when the market can swing 5-10% a day) , who cares if the largest validator has 23% or 23.02%.

I know the statement is patently false, and I don’t have to point out the errors in the assumptions to you guys, but I assure you a certain percentage of people delegating shares will have the above mindset…

TL;DR You can’t educate someone who doesn’t want to be educated, but I think building expectations of behavior from each other as validators will help to clear up potential issues down the road, and instead of someone waking up in the morning finding out they’ve been cliffed/voted out, they find a message that we all fee it prudent they increase their margins. It is far easier to teach 100 like minded individuals to act in the networks best interest, than it is to teach potentially millions of people how they should delegate.

Remember, if the expectation going in is that I’ll be voted out at X percentage, then my goal going is not going to be increased transparency.


#14

And I agree. I was more painting out a scenario/path whereby it’s relatively easy for an exchange to suddenly centralize something that’s meant to be decentralized.

It seems to me that there little solutions here and we’re almost entirely reliant upon the exchanges purposely choosing NOT to run a dominant amount of validators. Like even if they somehow have 2/3’s of all Atoms in their accounts’ wallets, we’re just going to hope that they choose to run less than 2/3’s of the available validator nodes.
Not good.