[PROPOSAL #12][ACCEPTED] Are validators charging 0% commission harmful to the success of the Cosmos Hub?

I urge that 0% commission is harmful.
(though I think 0% commission is a good strategy for small validators to promote themselves)

However, I think security is not the main point of 0% commission problem.
Even 0% commission is forbidden, there are lots of incentives which makes validators invest security less before we solve the asymmetric information between validators and delegators.
(So I think the proper solution of security problem is slashing.)

Nevertheless, I think there are 2 possible negative effects.

One is that could make the non-technical community members ignore costs of node operation. (It is the problem which @syncnode mentioned )
I think this problem is not a big deal, because it could be solved by simple nudging, for example, prompting (community tax rate + commission rate) in place of (commission rate).

The other and more important problem is that 0% commission is very strong disincentive that (especially whale) atom holders start to operate their own new validators.
As @kwunyeung says, inflation is a punishment to the atom holders not bonding their atoms to the network. Likewise, commission could be a good punishment to the (especially whale) atom holders not operating their own nodes.

1 Like

IMO we should not penalize validators who are operating legally within protocol boundaries. The free market should decide which validators survive and which ones should be eliminated. There are many workarounds even if in-protocol restrictions are set for 0% commissions.

We should look more into how we can incentivize delegators to not just look at fees, though I think that most delegators will just opt to delegate to validators who participated in GoS and/or part of Tendermint team because they are deemed safer. Middle-range validators who did not participate in GoS are probably in the worst spot, but thatā€™s another topic lol.

Iā€™m particularly fond of stake-based slashing (where higher ranked validators get slashed more than proportionately as compared to lower ranked validators, suggested by @sunnya97) and stake-based unbonding (where higher ranked validators have a higher unbonding period as compared to lower ranked validators, suggested by @asmodat):

  • It forces validators to be more transparent about their setups and security policies in order to attract delegation;
  • It forces delegators who values liquidity to think about bonding to smaller validators with proper setups (yes, there is delegation vouchers in place for staked ATOMs but they might have to take a steep discount in exchange for liquid ATOMs).
3 Likes

Valid points, I basically calculated this security budget for Cosmos and Tezos for this post we published in April. The stake weighted average commission rate at that time was 9.7% for the Cosmos Hub, it probably went down since then. With $750mn market cap and 71% staked (resulting in 10.13% reward rate), this means the yearly security budget of the Cosmos Hub is probably around $5mn ($5.23mn taking the exact numbers provided).

If we consider the security budget for the 100th validator with a stake of currently 40k Atoms (~$160k) taking the weighted average commission rate, current reward rate and market cap, it means that if this validatorā€™s stake is coming only from delegations, his security budget is $1,572 per year. This obviously probably bearly covers the infrastructure cost for a basic setup and in no way justifies any human capital spent on security etc. Using the weighted commission rate here makes sense imo because it is reasonable to assume that this is the maximum a smaller validator can charge and still get delegations.

Taking this perspective, 0% fee validators clearly are harmful as they bring down the security budget for the network. There are also some implications for what a reasonable validator set size is. In the end it requires education around these concepts and potentially goodwill from delegators (or more and stricter slashings?) to understand the implications of delegating to 0% fee validators. Hopefully in the future higher market caps, refined incentive designs, and other differentiation possibilities will help change this game and the current trajectory.

5 Likes

Iā€™ve read all of the comments here and have given the issue some careful thought.
Here are my opinions:

  1. Enforcing any kind of minimum or maximum at the protocol level will only harm the free market economy that exists. Itā€™s too easy to ā€œchargeā€ one fee via protocol and then settle out-of-band to get around any kind of ā€œenforcedā€ minimumā€¦ very much like how itā€™s ineffective to ā€œenforceā€ that no one validator receive more than x staked ATOMs. (They can just create two validators and divide the stake between the two to get around any kind of protocol enforced limits.)
  2. It is a logical fallacy (non sequitur) to conclude that validators who offer 0% fees do not have a secure/redundant setup (or conversely: a validator with high fees is secure/redundant. We saw Polychain (with 20% fees) go offline for long periods of time.) Perhaps a 0% validator makes their money another way that allows them to offer 0% fees. Itā€™s up to each validator to figure out their own economics, their own overhead costs, their own security models and I think itā€™s a mistake to have the Cosmos protocol meddle in individual validatorsā€™ business; theyā€™re independent for a reason. If companies want to pay their own money for the hardware and labour involved in operating a validator while receiving no direct fees from their delegators, thatā€™s their prerogative. It is not the Cosmos ecosystemā€™s job to identify profit/loss of each individual validator or to otherwise meddle with their operations.
  3. The security models of each validator is different, and each incurs a different cost to the operators of those validators. Over time, as validators miss blocks, get slashed, or otherwise misbehave, the market will react in kind by delegating away from them. ā€œEvery action has an equal and opposite reaction.ā€ The protocol already has been designed with this in mind, and the incentives of the Cosmos protocol work just fine. If someone wants to operate a validator in their basement with little-to-no redundancy, maintenance, or care - all the power to them. When they go offline or get slashed, the market will react and theyā€™ll find fewer delegates - perhaps knocking them off the top 100 (125?) making room for a new entrant.

TL;DR:
The fees collected by validators do NOT say anything about how secure the validator is.

2 Likes

I need to mention that the free market could be threatened by negative externality.
When it comes to negative externality, governance or regulation could be a good solution.

I agree that some regulation which has perverse incentive could harm the efficiency of free market.
However, we should aware that the free market itself is not always efficient.

I would really like to see some numeric evidence with regards to 0% fee validators doing any observable harm to the system. My cursory inspection of this forum tells me that a 0% fee validator such as Sikka achieves their low running cost by having an access to university server. I am okay with that. I also work at university and I can tell you that securing your spot on university cluster is really competitive for a large university, now that everyone wants to try out their poorly optimized machine learning algorithms. I am sure Sikka worked hard to earn their spot. Also, does a tall person deserved to be called out for making it into the league-A basketball team ā€œjust because the person happens to be taller than othersā€? No.

The argument only valid on validators who are not eager to invest their resources on security. Commission is independent to the whole validator setup. In your perspective, do you mean validator charging 100% commission has the highest security setup?

Instead the flexibility of the current protocol leave us room for creativity, just like our investors support us to run the #PWYW campaign. And Iā€™m happy that our investors and delegators understand 0% commission have no negative implication to the security of our validator setup.

:+1:

You are of course right that commission rate is no indicator of a secure setup. My argument is that by charging a low/0 fee other validators are forced to also charge lower rates to compete, thus lowering their security budget, and in turn what the network is spending on security as a whole.

1 Like

If a system succeeds because of good will of the participants then it is a game-theoretically fragile system. The cheaters will eventually invade. Ideally the system should work regardless of the participants being kind to one another or are at neck and neck to each other. A robust system should be indifferent to this. Thereā€™s only a scenario but no evidence that a particular set of validators are harming the system. They are not cheating. If their setups are fragile then they will get slashed.

I donā€™t agree with this. I donā€™t see this happening in many existing markets. And I donā€™t see how you make the direct relationship between commission and security budget. The commission should not be the only income of a validator (currently is as there is low tx fee) and the security budget is one portion of the total budget of a validator. Knowledge on protocol, security setup, contributions to the community are investments before being a validator. The budget should be there before taking up any commission. This is a commitment of a validator. If a validator rely on the variable commission to define their budget, do they change security budget on every delegation change?

In our daily lives, we see how large entities squeeze the small parties to die by using pricing strategy. On Cosmos, I hope it would still be a free market to encourage competition while we think ways to help small parties to survive with their initial small budget. When they become stronger, they can improve their setups and contributions to become a bigger player. Iā€™m one of those who got a lot of help from the community and eventually become confident as a validator.

ā€œThe commission should not be the only income of a validator (currently is as there is low tx fee)ā€

Comission is also applied on the tx fee, so where is the difference to now when tx fees come in? Also what else do you see as the income of a validator? I think itā€™s pretty dangerous to rely on income from other activities, as that will mean that parties such as e.g. exchanges that can subsidies 0% commission through trading fees will end up maintaining the network.

3 Likes

I do mean the income includes commission and tx fee in the above context. I think validators should rely more on tx fee in a long run as the inflation is not a reward and getting more commission from the ATOM holders is diluting their asset.

A suggestion on disencouraging(not enforcing) ultra-low commission rates by top rank validators

  • min_commission_mandatory_rate : a commission rate that any delegator of a validator should pay(either to the validator or to the community fund)

  • min_commission_mandatory_rate = 1.0 * validator_power / total_power

  • block_reward * commission_rate : goes to validator

  • block_reward * max[0, (min_commission_mandatory_rate - commission_rate)] : goes to community fund

When ā€œmin_commission_mandatory_rate <= commission_rateā€, the distribution is same as now.
When ā€œmin_commission_mandatory_rate > commission_rateā€, the difference amount will be donated to community fund.
1.0 can be a genesis parameter so that governance can decide. We can think that current value is 0.

Example) Validator A has 7% of total power, and commission rate of 3%. Then, delegators pay 3% commission to validator A, and delegators donate 4% amount to community fund, resulting in paying 7% as total.

Good things about above approach,

  1. It is not an enforcement on commission_rate of a validator. It is a strong disincentive given to validators having ultra-low commission rate. Validator can still stay 0% commission rate(But his/her delegators will pay some donation to community fund instead)
  2. Higher commission cost responsibility for delegating higher power validator, which has a good impact on decentralization.
  3. Still it allows quite good degree of freedom for small-mid validators to have ultra-low commission rate because min_commission_mandatory_rate is relatively lower than high rank validators.
  4. The protocol change does not have impact on the economy of most non-ultra-low-commission-rate validators

I admit that this approach still have sybil-attack vulnerability. But because the topic is not about ā€œmaking more moneyā€, the risk of manipulation by sybil-attack is relatively/significantly lower than other topics.(sybil-attack can become a tax-avoidance manipulation at best)

4 Likes

@sunnya97 Itā€™s really great for you to put up the proposal,cosmos uses inflation to both punish unbonded stake and incentivise bonded stake.From a delegator perspective I stake because I get rewards and it saves me from being punished.So,lot of delegators would definitely think of getting more rewards and this only increases with atom price going down.In a free market where we see everyone aligning to more monetary incentives,you canā€™t expect delegators to distribute stake to validators with 10% commission when they have a 0% commission fee validator from cosmos team who has better knowledge.
tl;dr 0% commission may not directly impact network security but over time will make other validators(whose only income is from staking) leave the network thereby making network centralised and weaker

I like the idea, but it still susceptible to the ā€œrebate attackā€:

  1. Validator sets commission to min_commission_mandatory_rate
  2. Validator rebates delegators whatever they paid in commission

Yes I agree. The rebate attack is very universal concept related to most design for distribution. @meherroy introduced a solution something like https://kleros.io/
I think we can research further on this topic(dispute resolution) independently from commission topic, to make each problem modular.

1 Like

I feel 0% fee validators cause the network to become less secure over time. Iā€™m not sure I would impose a minimum fee floor though, as doing that feels more autocratic than democratic.

A global minimum isnā€™t the only option.
The most important part I would like to add is a suggestion of a minimum fee proportional on the amount of ATOM a validator is in control of. So a global minimum fee is not necessary and validators with less than a certain amount of ATOMs staked, letā€™s say as an example, 1%, would be still able to offer 0% fees.
A linear function or an exponential function could work with a lower boundary at 1% and the maximum amount at 33% which we could all agree itā€™s bad. The function could look like this.minimumFees
This is just an example of course, parameters can be changed, but this alternative approach can offer a sensible middle ground between no minimum fees and a global minimum fee.

We hosted a podcast discussion around the 0% fee topic with Michael Perklin from Shapeshift, Sunny from Sikka, Chris from Chainflow and Brian from Chorus One. Give it a listen if you are interested in their viewpoints: https://chorusone.libsyn.com/7-validation-economics-and-the-impact-of-zero-fees

Also, as a side note: the proposal passed today with 63% of ATOMs participating. 59% voted yes, 0% fee validators are harmful to the network (https://hubble.figment.network/cosmos/chains/cosmoshub-2/governance/proposals/12).

1 Like