183 total views
Silvio Micali, Turing Award winner and founder of Algorand, elaborated on the Algorand governance mechanism, incentive model, and early unlocking mechanism recommendations.
Original title: “Algorand founder Silvio Micali: Proposal on Algorand’s Decentralized Governance Program”
Written by: Silvio Micali, Algorand founder and Turing Award winner
On November 24, 2020, Algorand founder Professor Silvio Micali released a proposal on Algorand’s decentralized governance, soliciting feedback from the global community. Silvio said that Algorand has successfully achieved decentralization at the protocol level, and the next step will be to achieve decentralization at the governance level. The full text introduces Silvio’s understanding of on-chain governance in detail, as well as suggestions on Algorand’s governance mechanism, incentive model, early unlock mechanism, etc.
Status : Currently, in Algorand, every account can participate in network consensus, but cannot participate in governance (Governance). Governance is the ability to determine non-consensus tasks, such as funding for Grants project proposals.
As an important step in Algorand’s decentralization, we hope to introduce a mechanism so that any ALGO account can not only participate in consensus, but also participate in ecological governance.
Goal : The goal of this proposal is to realize Algorand’s decentralized governance and combine network rewards with governance models. More precisely, the mechanisms and incentives it proposes ensure that the governance model is the same as our consensus protocol, which is decentralized, safe and effective.
Quick summary : Participation in governance will be completely voluntary. The account that you choose to participate in, that is, the governance account, or more simply, the governor, needs to lock the tokens held by it for a certain period of time (currently tentatively set to be one year) . Rewards should be given to governance accounts that contribute to the ecology. We expect that the rewards for governance accounts will be higher than their current network ecological rewards.
We propose to gradually start the implementation of the governance mechanism, starting with the funding of the proposal of the “250 million ALGO Reward Program” that the foundation is responsible for, to achieve decentralization.
Consensus mechanism VS governance model
Before proposing the guiding principles of participation in governance, let’s review the principles of participation in consensus and highlight the differences between these two participation modes in Algorand.
The consensus process includes selecting the next block in the blockchain and verifying that all its transactions are valid. Like other blockchains, reaching consensus is a purely algorithmic process. But in Algorand, the consensus mechanism has some unique characteristics. That is, the consensus of participation in the Algorand network is:
Indeed, on Algorand it does not require any important computing resources (Translator’s Note: such as renting CPU or memory), and it can actually be executed by a normal laptop in the background.
(We strive to ensure that everyone can participate in consensus!)
Algorand accounts can freely choose whether to participate in the consensus mode.
Since only a small threshold of computing power is required in the Algorand consensus mechanism, it does not require any separate block reward as compensation.
Since there is no special reward for participating in the consensus process, it is not necessary to check whether an account actually participates in the consensus. Therefore, this mode of participation is not subject to special monitoring.
(In essence, in Algorand, participating in consensus is a “very easy form of service.”)
The above guidelines have been valid since the launch of our network. In our consensus agreement, we have procedures for approving and deploying upgrades. We have successfully deployed three major consensus upgrades (for example, the Layer 1 smart contract). But we have never changed the characteristic principle of our consensus agreement. These principles will continue to apply to this proposal.
However, participating in governance is completely different and requires different principles.
The governance mechanism is not an algorithmic mechanism like the consensus mechanism. It takes personal energy and time . Therefore, we propose that in the Algorand network ecosystem, participation in governance is:
Algorand accounts can freely choose whether to participate in governance.
Governance accounts will be rewarded, as detailed in section 3 below.
The governance account will be monitored to confirm its true participation in ecological governance.
Need to lock up
The decisions made by the governors will profoundly affect the development of the Algorand ecosystem. In order to ensure that they will also be affected by the results of their own voting, the management account will participate in the service for one year, and the position will be locked during this period.
However, as we will detail in Section 5, managers can withdraw cash from their locked accounts in advance, and there are clear restrictions and penalties for early unlocking. However, the compensation rewards obtained by the manager are not locked! In fact, they will be directly deposited into the unlocked account chosen by the manager.
Let us now explain how these principles affect governance and incentive mechanisms.
The governance mechanism clarifies the voting process for jointly making governance decisions. This process only involves the foundation and the governor, and their roles are very different:
“The foundation is responsible for assistance, and the governor is responsible for decision-making.”
This clear and diverse governance mechanism can and should be considered and explored. But all of these should follow the following guidelines:
The voting meeting requires managers to vote on the details of each group of proposals (for example, a group of reward plan proposals submitted to the foundation), including:
- Its name/ID;
- A short summary;
- A link to jump to its detailed information;
4. Yes and No options (will produce different broadcast results on the blockchain)
People need time to think. We recommend that the foundation provide 30 days to think about voting. Voting published on D-day has a deadline of D+30, and only votes made before the deadline will be considered valid.
Any member of the community can initiate Session Items, but they must be posted on the blockchain before they can be included in the meeting.
The foundation is responsible for conducting preliminary research on all topics that need to be voted on, and will be the first to vote, and will release its own voting results at the same time as the conference is released.
Suppose a governance proposal has been broadcast for 6 months, but it is still not included in any voting session. Then, the governor can publish through on-chain transactions and indicate that at least 25% of the governance stake supports this initiative, and it is mandatory to vote on the single project.
The fee charged for the broadcast of a project proposal is higher than the Algo fee for a general transaction. In fact, ordinary transaction processes are processed through mathematical algorithms, while governance projects must be processed by humans. Therefore, we must prevent someone from wantonly wasting the precious time of the foundation and the governors due to low costs.
A voting meeting should have a “VWF-Vote with the Foundation” option. Governance accounts can be freely selected:
- Ignore the VWF option and vote separately; or
- Choose the VWF option to match the foundation’s vote.
Vote Validity, Vote Weight, and Outcomes
Needless to say, each voting session should specify what makes the voting effective and how effective voting is translated into results. Only the governance personnel are allowed to vote, and the voting weight needs to be proportional to the number of tokens locked.
For the sake of simplicity, assume that the lock-up period of the governance account always starts on the first day of each month. Now suppose that at the beginning of any month M during the lock-up period, there are T tokens in the account, and the governance return reward rate obtained is r. Then, at the end of M month, the account will receive one of two modes of rewards:
- Receive a token reward of rT/12 (Translator’s Note: The average number of reward tokens per month for 12 months), provided that it has voted on all issues published during the lock-up period, and the vote is closed Within M
- Otherwise, you will receive half of the token reward, which is rT/24 reward tokens.
In the latter case, the unreceived rT/24 tokens will be transferred to a special account of the foundation to reward future managers.
Due to the transparency brought by blockchain, monitoring governance participation has become simple and clear.
Governance voting should allow mobile wallets to participate.
Separate governance keys
Governance voting must be digitally signed by governance-participation keys, which are separated from the keys for transaction and consensus participation. This makes the governance key temporary and will expire after voting. It also allows users who retain their governance voting rights in their wireless wallets to transfer their governance accounts that remain locked to the escrow service provider.
All in all, the goals of the above guiding principles are actually very simple, namely:
“The foundation should promote participation in governance, not control governance.”
Examples of voting types: Voting-Type Examples
We need to consider different voting methods. Here are just two examples.
K-out-of-N Voting (K-out-of-N Voting)
Specifically, for example, a topic of 11 out of 20. This requires voting on 20 project proposals: for example, 20 award proposals, each with its own introduction information, “YES” and “NO” buttons. For such an issue, valid voting is (for example, by clicking YES) to select 11 of them.
In response to such a vote,
- The foundation can actually prioritize all 20 project proposals in order to provide more information to the governor
- The foundation’s vote is implicit in the ordered list provided: the first 11 proposals in the selected list.
Therefore, in the voting for K in N, the foundation provided more information than its own voting. If a manager can only decide to choose—say—7 proposals, then the ranking list provided by the foundation may help him choose another 4 proposals and cast a valid vote.
In the above-mentioned 20-out 11 voting, in addition to (a) choosing 11 proposals individually and (b) voting with the foundation (VWF), we envision that the governor can choose the third option: (c) Select the VWF option, plus the 7 proposals originally determined by the governor. By choosing the latter option, our manager automatically selected 11 specific proposals: 7 clear choices, plus the first 4 proposals that appeared in the foundation’s ranking list but were not explicitly selected by the manager.
The result of such a vote might include—for example—appropriating the 5 proposals with the highest selection weight. (If the manager has T locked tokens, then it will contribute weight T to each proposal it chooses.)
One advantage of voting for K among N is that a governance account with a large number of coins cannot fully concentrate its voting weight on a single proposal-for example, an account itself submits a financing application! If it wants to vote for its own proposal, it must also vote for 10 other proposals. Therefore, the administrator needs to spend time to select 10 solutions that he believes are the best for the Algorand ecosystem.
Greedy Budget Voting
When each proposal requires roughly the same amount of funds, the voting effect of N out of K is indeed very good. But when this is not the case, we need a voting system that takes budget parameters into account. Here is an example of voting.
The foundation specifies the budget and sorts the proposals. The vote consists of the highest-ranked projects, which can be allocated without exceeding the budget.
A manager can:
- Choose to vote with the foundation (VWF) option; or
- Choose a group of projects that can receive all financing within the budget.
In the latter case, the governor’s vote is automatically calculated by checking the foundation’s list in an orderly manner and selecting a proposal that has not yet been selected and the allocation does not exceed the budget.
Governance reward mechanism
Let the foundation directly determine the compensation incentive for the governance account is centralized, simple and straightforward. But this may be counterproductive. If the wrong reward benchmark is selected, the foundation is likely to discourage the participation of the governor.
Consistent with Algorand’s principles, we propose to select the governance reward rate of governance accounts in a decentralized manner. To be more precise, we propose that within the parameters determined by the foundation, the governance account itself decides through Dutch auctions.
For the sake of simplicity, first assume that once a year, the Dutch auction of the governance reward rate will be conducted on January 1. The account that wins the auction this year is the governance account for this year and remains locked throughout the year. Once the lock-up period is over, they can freely participate in next year’s auction.
At the end of this section, we will explain how to deal with any number of auctions that start during the year. For a more precise description of Dutch auctions and their implementation on the blockchain, please refer to . Here, we only want to provide content related to the Dutch auction of governance reward rates. The following description is only descriptive, but hopes to express how the governance rate of return is “determined by the account itself.”
For the annual auction, the foundation will specify:
- Reward pool-taken from the inventory, it is prepared to allocate the total amount R of governance reward tokens within one year .
2. The number of auction stages, k
3. The increasing rate of return between (0 ~ 1): r1<r2 <…< rk, where r1 = R/T, T represents the number of tokens in circulation at the beginning of the year
For example, R=8⋅108, T=1010, k=100, r1=8%, r2 =9%,…, r100=107%
The rate of return r1 in the first stage is the smallest possible governance reward rate generated by the auction, and the rate of return rk in the final stage is the largest.
The actual governance reward rate (non-compounding) is determined by the bid (Bid) of the account at each stage of the auction.
Therefore, let us briefly describe (a) what price an account can offer at a stage, and (b) how the price offered determines the outcome of the auction.
(A) At stage i, the account can bid for tokens of a. The bid “commits to lock in (extra) tokens when the governance return rate is greater than or equal to ri”.
These tokens are bound:
(1) Cannot be used or traded during the auction period, (2) If the bid “wins”, the token will be automatically locked for one year.
Of course, the bid is valid only when the account still has enough tokens of a
(b) The auction is terminated at the first stage i, and Ci is used to represent the number of tokens bound so far, riCi ≥R
If such a stage does not exist, then the auction ends at stage k.
Let the auction end at stage t, then,
- Governance reward rate is rt
- All tokens that have been bound so far will be locked for a full year (in a separate account)
- Accounts with these lock-up tokens become governance accounts
The termination condition is set as riCi ≥R, which is actually quite natural. In fact:
- The foundation is unwilling to use more than R token reward governance accounts,
- When the auction ends at stage t, the reward rate is rt
- Under the rt governance reward rate, all locked tokens Ct in the governance account are rewarded, and the foundation needs to spend rtCt tokens at the end of the year.
Keep the above conditions consistent, let k=100, r1=8%, r2=9%,… ,r100=107%; let the auction end in the 13th stage; and assume that an account x with 900 coins only bid (Bid) Three times: the first time 100 coins were submitted in stage 1; the second time 200 coins were submitted in stage 5; the third time, another 200 coins were submitted in stage 10. Please note that all three bids are valid.
Then, the auction sets the governance return rate to r13=20%, and account x becomes a governance account with 500 lock-up tokens. (The other 400 tokens of account x continue to be in a non-locked state, in a separate account controlled by x himself.)
x The governance participation of the account will be monitored. If x does participate in governance, then the number of rewards received throughout the year will be 500⋅20% = 100 coins, which will be issued from the reward pool at the end of the year.
Account x should be satisfied with the auction results: at the first bid, it announced that if the governance return rate reaches 8% or higher, it will be happy to lock up 100 coins. In the end, the governance reward after the Dutch auction was 2.5 times higher than originally expected! For the same reason, account x was also satisfied with its second and third bids. (Translator’s Note: Dutch auctions are auctions in descending order, that is, if the final transaction yield is higher than the yield when you bid, then all the bids you bid at the low yield will be successfully sold at the highest yield.)
Yearly to Pipelined Dutch Auctions
For simplicity, we assume that a single governance reward rate auction is conducted every year.
However, it is important to be able to conduct such auctions more frequently (for example, once a month) while maintaining the lock-up period of the governance account for one year.
The reason for maintaining the one-year lock-up period is to continue to ensure that the governance account has sufficient “benefit binding”. The reason for this special auction more frequently is that new tokens will continue to enter the Algorand ecosystem. Therefore, we must allow these tokens to participate in governance as soon as possible, without waiting until the next year.
However, we must also realize that holding Dutch auctions that govern the reward rate several times a year may complicate the strategic thinking of bidding accounts.
For example, an account won the Dutch auction on February 1 and locked its tokens, and when the Dutch auction on April 1 produced a higher governance reward rate, it may regret it. Faced with these complex strategies, governance accounts may react in unforeseen and counterproductive ways.
In order to avoid such regrets, Dr. Chen Jing and I designed an architectural approach that allows a governance account to use its locked coins in previous auctions to participate in any subsequent auctions it wants. If they win, their account can be “seamlessly switched”, and some or all of the locked coins can be selected and credited to the new lock-up period and the new governance reward rate.
We call this type of structured auctions Pipelined Dutch auctions .
For a more precise description of this type of auction, please see .
The pipelined Dutch auction allows the foundation to conduct an auction of governance reward rates at any time.
For example, at the beginning of each month, the foundation may:
- Calculate the number of tokens that entered circulation last month T
- Create a reward pool containing R=T⋅r tokens, where r is the annual reward rate chosen by the foundation
- Use the created reward pool (and the number of stages it wants and the stage interest rate) to conduct governance reward rate auctions.
Early unlocking mechanism: the reward and punishment mechanism for withdrawal from the governance account in advance
In order to encourage locked positions to participate in governance, we allow early withdrawal from the governance account, but with the following conditions:
The governor can withdraw up to 90% of his locked tokens within 30 days. In addition, for each extraction, a manager g needs to:
- Pay a penalty equivalent to 10% of the withdrawal token
- All governance rewards for the current month will be lost.
Half of the fines of g will be injected into the planned bonus pool for the next auction, and the other half will be distributed to the remaining managers at the end of the lock-up period. The distribution standard is based on the weight of the currency held by other managers at that time. The same principle applies to the governance rewards of the month that g loses.
These fines and losses make g reluctant to easily initiate early withdrawals from locked accounts. In addition, this will also encourage those managers who receive most of the fines and governance rewards of g to continue to participate in governance: the more early withdrawals, the more additional rewards the managers who continue to lock their positions will receive. Finally, the foundation can use g’s losses and fines to promote new governance auctions in order to attract tokens that have not participated in lock-up governance to participate in governance voluntarily.
Comparison of network rewards VS governance rewards
Today, all Algorand accounts, regardless of whether they participate in consensus, each ALGO in the account will receive rewards from the Algorand network. In 2020, the total number of network reward tokens will reach 6% (non-compounding average). According to the original plan, such a network’s annual reward return rate will reach zero in the next 3.5 years.
However, the foundation is planning to change the network reward program. By 2021, the reward program will expand to 7 years, correspondingly from an annual rate of return of approximately 8% in 2021 to 1% in 2028. In 2021, the foundation plans to distribute R2021 network reward tokens , R2021=8% ⋅T2021 , where T2021 is the number of ALGOs circulating in early 2021. The operation from 2022 to 2028 is the same.
If the same number of reward tokens are used by the foundation for governance instead of network rewards, what will happen? It will set the governance reward rate in 2021 through the Dutch auction on January 1, and there will be R2021 in the reward pool Tokens.
In such an auction, according to the basic rules, the annualized rate of return for stage 1 is r1=8%.
Therefore, as long as they are willing to lock these tokens and participate in governance, each account can ensure that its governance reward rate for all or part of the token is 8%.
In fact, all the account has to do is submit all the tokens in phase 1. By doing this, even if all other accounts have submitted all of their tokens in Phase 1, the account ensures that its bid is a winning bid. More generally, no matter how other accounts handle the 2021 auction, the final governance return will be at least 8%. In fact, the reward rate for the first stage is 8%, and as the stage increases, the reward rate will also increase.
Of course, the ultimate governance return rate enjoyed by the governance account may be higher than 8%. In the first stage of the auction in 2021, as long as not all accounts have submitted all tokens for participation, higher governance return rates will appear.
For example, suppose that all accounts do not want to lock T/2 coins (that is, half of the coins in circulation). Then, these coins will never be submitted in the 2021 auction. Therefore, no matter how the auction is conducted, the final governance rate of return is at least 16% . In fact, the auction cannot end at any stage s where the reward rate r_(s)<16%. Indeed, even if all other T / 2 coins are submitted in stage s, because R = 8% T, we must have T/2·rs
Perfection is the enemy of excellence.
True decentralization cannot be achieved in all governance fields at the same time.
We should start from an important area; learn; and adjust if necessary; then, with confidence and experience gained, expand decentralized governance to more areas.
We propose to start decentralized governance starting from the funding of the proposal of the foundation’s reward plan. Funding the right projects is critical to the development of the Algorand ecosystem. But finding these correct items is not easy.
Authorize communities so that they can participate in choosing which projects to fund-we will be more likely to find those projects that can bring the most value to the ecological development of Algorand.
Decentralization has always been Algorand’s primary goal.
We first achieved decentralization at the blockchain protocol level. In fact, we are proud of Algorand as the first centralized, malleable and secure blockchain to successfully break the impossible triangle of blockchain.
Now, we must achieve true decentralization at the governance level.