The bootstrap phase in the lifecycle of a blockchain is a very delicate one, not only is it vulnerable as was explored in a previous article but there is evidence of how decisions are made having a lasting consequence.
We have seen how the Validator landscape has evolved where a small number of Validators have accumulated power through offering zero commission to delegators and thus further concentrating voting power. In a market economy there is nothing wrong with that, right? The people with the better business plans and execution, will in a level playing field, do better.
The issue in the blockchain world becomes where the power is concentrated, then an important element, namely, decentralisation becomes diluted. Even with five fiercely independent Validators, holding 50% of the voting power, the concentration of power could lead to the interests of the chain being steered by a very small interest group.
So how did this happen in chains like Cosmos? How can it happen to a new chain? Handing out a bunch of tokens to the staff, or deeply discounted tokens to early investors, gives them a big leg up in setting up Validators.
In the heady days of launching testnets, having the team involved is good, and why shouldn’t they have a run at it. The staff are not the only ones with an advantage, there are “whales”, who maybe early investors who can throw cash at the opportunity knowing they can squeeze out the others. All is fair in business except that in a decentralised world it matters.
Bonded proof of stake examined
Taking the Cosmos model of Bonded Proof of Stake and introducing some non-linear reward curves to discourage Validators building big stakes is a starting point.
The challenge is to find a mechanism that is hard to “game”. Let’s consider a scenario where there is a threshold beyond which there are no further rewards, the obvious “work around” for a Validator would be to set up 10 separate ones, as the unit cost is potentially lower for each one and the Validator would continue to exert influence and power to the possible detriment to the chain.
The solution to this is to create a link from the blockchain to a “real world” identity or even better a decentralised identity verification mechanism (other chains or self-sovereign). The linking to identities would make the setting up of multiple Validators more difficult.
Wait this is Proof of Authority? Not really as there are decentralised approaches to link identity to an address, and unlike PoA where reputation is staked it is recognised that a bonded stake has a role in rewarding good behaviour and punishing bad behaviour.
Let’s reconsider the purpose of the Validators within the ecosystem and what the goals of the chain are. For the long term success of all the interested parties, partnerships between developers and Validators is a key element of this. A good relationship built on engagement between the developers building and contributing code to the chain and the Validators in securing and running the chain will foster a strong platform.
In considering all parties, it is important to consider the checks and balances between the various parties, so that the long term goals are aligned with everyone. The Validators, developers and users can veto decisions, for example the developers and Validators could fork the chain but if the users are not on-board the market cap collapses.
Having a verified identity and recognising that there are long term goals of a chain, what tweaks do we need to put in place? Remember we began by mentioning non-linear reward curves. There needs to be an extra dimension to align the goals of long term commitment and a human element. We have covered the human element by adding the link to an identity, but what about the long term commitment and why is that important?
It could be distracting to focus on pure economics and is important to keep in mind that the objectives of a chain are to get a depth of Validators to ensure that the chain is decentralised, is secure and the Validators have clear sight on their ROI which motivates them to run a node.
Having a lot of Validators running nodes and paying the same to everyone, good bad or indifferent may lead to a less secure chain as they know they get paid regardless of what they do so will make less effort to invest in their infrastructure and maintain an active participation..
The revenue generated on chain is through transactions and fees, and it may be stating the obvious but the chain must provide a good platform for organisations to build their businesses on. What makes a good platform? A good feature set coupled with fair and transparent fees? A growing chain with a robust revenue stream is the foundation for bringing in Validators and it thus becomes a virtuous circle.
The key missing element is “engagement” of the interested parties to ensure the long term success of the chain.
Proof of Engagement
Starting with the basic premise that the consensus model aims to reward Validators who are active in securing and maintaining a strong chain, we consider how to fairly reward the Validators for the work they do.
Other consensus mechanisms have shown that there must be a combination of rewards and with that comes responsibility which is normally where slashing comes in. Of the active chains in use today, Cosmos is one of the best designed mechanisms, there are however some small changes that would address the issues around the centralisation and security which are linked to size of stakes. For the Cosmos Hub, the group is strong and stable enough that change is not urgent. However, for smaller chains, this is imperative.
What can be observed is the tokenomics which favour large stakes, where there is a concentration of stake and voting power in favour of a small number of Validators. If we consider IrisNet, Terra and Binance, these chains base security on the fact that the organization developing it holds a major portion of the tokens, avoiding manipulation by whales and speculators only by effectively centralizing the control.
The question of finding a balance between a fair reward and a mechanism so that the Validators who go beyond the basic block metrics (up time, voting participation etc) have a reward mechanism.
The link for engagement to reward
We all know what good is, right? Sort of, well Alice’s view may differ from Bob’s who may differ from Anil’s. In the decentralised world, a nice set of rules that are measurable is what is needed, but in a softer definition of engagement we need to build a picture of what is “good” and tie that to reward.
How is engagement defined?
This is the Validator that reaches out to others, shares best practice, is active in the group chats and forums, and the one that writes some nice open source code tools to help run and monitor the chain for good health.
This highly subjective and mostly off-chain, now we know what good looks like we can start forming profiles of Validators. This must be community led and ties in with the close collaboration between the Validators, the development community and the token holders.
What happens with dis-engagement?
We have all seen the scenario where a new person comes in, is very active, contributing to forums, building useful tools and then the enthusiasm fades and the engagement slows down and they begin to cruise.
In the case of a Validator accruing engagement rewards, it is not right that the Validator should continue to enjoy the benefits if their engagement slows.
To ensure that a Validator remains engaged there is a “half-life” mechanism where the engagement rewards decay unless the Validator continues to engage and accrue rewards.
We have established a subjective method for evaluating engagement, and looked at the engagement rewards “half-life” for dis-engagement. The idea of converting engagement to rewards points, so that a Validator can accrue tokens in acknowledgement of their engagement.
The engagement rewards should be proposed and awarded by the electorate and the engagement points are issued based on the outcome of the vote.
Long term commitment and innovation
It is desirable to have the Validators aligned with the long term vision of the chain and build engagement tokens, however, there is a risk that this could end up with a situation that leads to “time served” Validators building a big advantage over newcomers.
There must be a balance between long term supporters and newer Validators who may bring fresh thinking and innovation.
A number of initiatives must be baked in to encourage and showcase innovation such as hackathons and bounty programs which reward innovation. This is not necessarily restricted to newcomers, established Validators can also participate, as long as they can demonstrate innovation.
When defining engagement rewards, we need to look at the following key factors:
- Initial distribution of the rewards
- Distribution of new rewards
- Rate of decay of old rewards
When we consider the long-term evolution of engagement rewards, we need to evaluate two distinct points:
- The Macro level which is the change in centralization, with a focus on the number of holders of engagement rewards and the slope of Pareto Distribution.
- The Micro level with respect to Turnover. Do the same individuals retain their top positions, or does this change often? Balancing Stability vs. Innovation
Initial modelling showed the rate of decay should be a flat percentage applied to all engagement points, such that top reward holders who cease to actively participate should quickly lose rank, faster than lower reward holders.
Furthermore, the distribution chances should be divided between new and old, such that say 50% of engagement points are distributed to existing, high-performing Validators based on their voting power. And the other 50% distributed to community members (excluding say the top 50% of Validators by engagement).
If the influx via reward and the outflux via decay are balanced, then the top Validator position will naturally be in flux, since they are subjected to decay at 100% of their power, but eligible for rewards proportional to only 50% of their power.
Pulling it all together
Bringing together the stake, engagement and the Validator power and applying a sigmoid curve we begin to see a fairer reward. Using a sigmoid curve provides a smooth ceiling on max stake, and incremental bonuses to power for those with high engagement.
The interplay between stake and engagement to the overall rewards can be seen in the chart above where the x-axis represents the amount staked and the y-axis the engagement points. It can be observed that engagement eases in newcomers and equally reduces staking requirement for established Validators. By applying a sigmoid curve the distribution is smoothed.
The distribution gives lower rewards for new-comers, however once established, the Validators can get similar rewards, and the staking requirement goes down as engagement goes up.
Were do we go from here?
A draft whitepaper has been written which further details the economics and a possible implementation. Martin and Ethan would be happy to discuss more and get feedback from Validators and new zones, both on the economic design as well as a potential proof of concept implementation.
Thanks to Ethan Frey for co-authoring the article