On-Chain Governance

Basics

• Is a term used to describe governance protocols built into a blockchain. Usually done through staking, voting and funding mechanisms. These protocols can vote on new upgrades or on how to spend their funding.
• From the Daily Gwei (27-7-2020):

"When it comes to governance in the context of blockchains, there are two core types - binding on-chain governance and off-chain governance. “Binding on-chain governance” is when a blockchains rules can be changed by token holders (such as in Polkadot) and off-chain governance is when those rules cannot be directly changed by token holders - rather, they are changed by an off-chain process (that can also use on-chain processes as a signal)."

• From The Defiant (29-4-2020):

"By transferring the administrative control from single teams to a distributed network of token holders, DeFi projects can realize their value proposition as “decentralized” financial protocols and applications. The theme of decentralized governance will become increasingly important as these protocols become more robust and less reliant on the core teams."

• As explained by this post (7-2018):

"Off-chain Governance: A governance process whereby decision-making takes place on a social level (off-chain) and then, afterwards, is acted on by developers who encode it into the blockchain protocol. Examples include Bitcoin and Ethereum.

On-chain Governance: A governance process whereby rules, voting, and choices are hardcoded (on-chain) into the blockchain protocol. Examples include Tezos and EOS. Voting in a democratic system relies on a “one person, one vote” system. On-chain governance, on the other hand, is often considered plutocratic (A plutocracy or plutarchy is a society that is ruled or controlled by people of great wealth or income. — Wikipedia, 17th May 2018) because of the way voting is carried out. The on-chain voting systems rely on the number of native tokens (corresponding to the protocol being governed) that a given address holds."

• This blog post from Richard Red explains it further (6-2018):

"On-chain governance is not a well-defined term, but it indicates that changes to consensus rules are not made in the conventional way (nodes switch to a new software version). On-chain governance implies a formal system operating on-chain that determines if and when changes to the consensus rules are activated. Projects that embrace this approach (or plan to) show significant variation, these projects are not a homogeneous set.

Coin-holder voting is a method of assigning votes to participants based on the number of coins (the project’s native asset) they hold. This is a misrepresentation of how most of the projects that have formal vote-based governance actually work. Typically holding coins is not enough to entitle one to vote, users must take some action to stake their coins. For Decred, one must lock DCR to buy tickets to participate in governance (and so the more precise term is ticket-holder governance). For Dash, one must hold enough DASH then run a masternode to participate in governance (and so the more precise term is masternode governance).

Stakeholder governance is accurate in the Proof of Stake sense, in that holders who have staked their coins are the governors. I don’t think the term stakeholder governance is great for this piece because it is often used to imply a variety of types of stakeholder. In stake-based governance the wishes of different stakeholder groups are reflected through the unitary dimension of how much they have staked. Stake-based governance actually seems like quite an accurate term.

Internally, off-chain governance is also an ever-present alternative, in that a hard fork that removed or nullified the on-chain governance of a blockchain would effectively be reverting to this more conventional approach to blockchain governance. No governance, at the protocol level, is like a default setting that is always available. On-chain governance will have to continuously show that it is a better choice."

Protocol-level versus Application-level Governance

• As explained by this post (7-2018):

"One must also distinguish between two kinds of blockchain-based governance: protocol- and application-level governance.

I) Protocol-level governance – this is governance over changes to a given blockchain protocol such as Ethereum or Bitcoin. This type of governance has historically been off-chain, but there have been more recent efforts at on-chain protocol-level governance.

II) Application-level governance – this relates to the inner workings of blockchain-based applications. For example, users being able to vote for the content they see on a decentralized video sharing application, or the voting mechanisms of a Token-Curated Registry.

It is useful to draw the distinction since both of the negative accusations of on-chain voting as ‘plutocratic’ (as well as the assumption that plutocratic governance is bad) most often hold on the protocol-level, but not necessarily for the application-level."

How Does It Work?

• Again from This blog post from Richard Red (6-2018):

"There are two types of decision-making for which stake-based governance is being used right now:

1. Deciding whether to adopt changes to the consensus rules of the network. This is primarily what on-chain governance refers to. Decred is the only project I know of that’s already doing this. There is a formally defined process for changing the consensus rules, it is defined within the protocol itself. That process involves 75% of voting tickets saying Yes to the change within a voting period of 8064 blocks (~4 weeks).
2. Deciding how to spend a project fund. There are quite a few blockchains now that put some proportion of the block reward into a fund for developing the project. Most of these projects (and all of the ones I’m interested in) aspire to decentralizing control of how these funds are spent. Dash has been doing this with a fairly simple DAO setup, operated by the votes of MasterNodes, for almost three years now."

How It Should Work

• From this blog (16-1-2020) by bZx, which at the time of writing is transitioning towards a DAO model for governance:

"Properties of Decentralized Governance

Restricted Mutability

The protocol should only be upgradable through the consensus of representative stakeholders. The original team which developed the protocol should not be able to arbitrarily upgrade the logic of the smart contracts. Requiring changes made by administrators to pass through a timelock is an improvement on completely centralized trust models, but still leaves the protocol vulnerable to regulatory risk and malicious operators.

Inexorability

An unstoppable protocol has no central party capable of activating a kill-switch. This means that in the event of legal or regulatory threat, no party can be compelled to prevent or censor activity on the protocol. If a kill switch exists, it must be under the control of a diffusely distributed set of representative stakeholders.

Self-Perpetuance

The demands of a protocol are constantly evolving, as is the Ethereum execution environment. Future upgrades are not guaranteed to avoid breaking changes to existing contracts. A DAO must have a source of funds and a process for allocating those funds for development in a way which is secure. By far, this is the most challenging property for a DAO to achieve."

Problems of Governance

Participation

Current governance schemes record extremely low participation rates. There are several proposed explanations for why this might be the case. The leading explanation is that it requires substantial effort to determine when to vote and how best to cast that vote while the effects of any one participant’s votes is unlikely to be determinative. In light of this, voters, especially smaller ones, rationally abstain from voting. Some have expressed that this is a feature, not a bug, and that voting participation reflects an organic form of holographic consensus.

Consider the case of a stability fee vote in the Maker ecosystem. If DAI is selling for higher than the peg, voters will generally agree to lower the stability fee (or now decrease the DSR). If only a few participate, it doesn’t reflect voter apathy or a failure of the voting system, it reflects the fact that most voters agree with the course of action. If it is true that voters are more engaged than their participation would suggest, we would expect high turnout for controversial proposals and low turnout otherwise.

It is likely that both explanations have merit. There is some degree of holographic consensus built in to turnout, but it’s also the case that a large percentage of voters do not feel adequately incentivized to vote. Reasoning about incentive structures that can promote voter engagement is a useful endeavor because the larger and more engaged the voter base, the more expensive an attack becomes.

Shadow Voters

A shadow vote is a vote cast by a token holder with no economic stake in the protocol. This can be accomplished by borrowing a governance token, voting with it, then returning it to the lender. In the worst case, a shadow vote can be virtually free. The attacker executes a flash loan, votes, and returns the loan within a single atomic transaction, incurring no capital carrying costs or interest payments. In more ideal cases, the attacker is forced to bear capital carrying costs, to pay interest for an extended period, or to expose their collateral to margin calls and penalties.

Plutocracy

Without identity, governance structures collapse into plutocracy. Identity ideally imposes a cost greater than the reward for an attacker to act as more than one entity. Existing forms of identity, both centralized and decentralized, are lacking, failing to impose substantial costs on attackers. Low quality proofs of identity such as Twitter accounts can sell between a few cents to a thousand dollars depending on the age of the account, the number of followers, the quality of the followers, and other factors. Proofs of identity at the highest end, e.g., government issued passports, can range between a few dollars for a digital scan to low five figures for a hard copy. If a sufficiently valuable dApp was to implement quadratic voting, the rewards of sybil attacks would far outweigh the costs.

Consider the case of MakerDAO. It was recently published by Micah Zoltu that an attacker could steal all the funds insider Maker with only 20MM (40,000 MKR). If quadratic voting were in effect, an account with 40,000 MKR would only have the influence of two hundred individuals holding 1 MKR, meaning that 20MM in the hands of a single person would have the same influence as 100,000 USD in the hands of two hundred people. If Maker had implemented a robust KYC system requiring multiple identity documents, it would cost an attacker at most 15,000 USD - 20,000 USD to register a second identity. By splitting MKR holdings into two piles of 20,000 MKR, the attacker would possess voting power equal to 282 people, a 41,000 USD increase in voting power. This example should demonstrate just how ineffective even current identity schemes are for the purposes of implementing quadratic voting.

Lastly, a major risk of having a few large stakeholders along with linear voting is the possibility that they will collude amongst each other to drain funds from the DAO into their own wallets.

Cartels and Bribes

Representative democracy is often deployed to counter a lack of voter enthusiasm and expertise. If representatives are incentivized, and they are able to vote on their own compensation, then this opens the door to cartels and rent-seeking. There are two possible equilibria for a set of representatives: the Nash equilibrium and the Cournot equilibrium. Under the Nash Equilibrium, each potential representative competes with each other, leading to a situation where no rent is extracted and the representative offers their service at the marginal cost of doing so. The alternative is the Cournot equilibrium, where representatives cooperate with each other to extract rent. Since cooperation is the most profitable long term strategy for representatives to pursue, this equilibrium can arise spontaneously without explicit coordination during infinitely iterated games."

It then follows to explain how they themselves will create a DAO with a Trias Politicas in mind.

Critiques

• As of 1-2020 most on-chain governance has still seen little turn out and voter apathy.
• Quite some people have said blockchain governance leads to plutocracies. Most notably Vitalik and Vlad.
• Some have argued against this view. From this before mentioned counter post:

"Vlad Zamfir and Vitalik Buterin have each written articles on blockchain governance and the perils of plutocracy which you can read respectively here and here. The two articles offer two fundamental arguments against ‘plutocratic’ blockchain governance systems:

1. Coin-holder and blockchain user interests are not naturally aligned (Vlad).
2. Plutocratic/On-chain governance schemes open the protocol up to attack vectors — such as bribery and vote-memeing (Vitalik).

The writer later self counters with:

The key to preventing an on-chain governance system from descending into a plutocracy is to ensure there are checks and balances in the form of loosely-coupled voting systems — with Decred being an excellent example of this."