What you will read here is a preview of what could be implemented in the Allfeat Mainnet and is subject to change in the future.
Symphonie Testnet doesn't include any governance system as the network is maintained and managed by Allfeat Labs.
Allfeat uses a sophisticated governance mechanism that allows it to evolve gracefully overtime at the ultimate behest of its assembled stakeholders. The stated goal is to ensure that the majority of the stake can always command the network.
To do this, the network brings together various novel mechanisms, including an amorphous state-transition function stored on-chain and defined in a platform-neutral intermediate language (i.e. WebAssembly) and several on-chain voting mechanisms such as referenda with adaptive super-majority thresholds and batch approval voting.
All changes to the protocol must be agreed upon by stake-weighted referenda.
To make any changes to the network, the idea is to compose active token holders and the council together to administrate a network upgrade decision. No matter whether the proposal is proposed by the public (token holders) or the council, it finally will have to go through a referendum to let all holders, weighted by stake, make the decision.
To better understand how the council is formed, please read this section.
Referenda are simple, inclusive, stake-based voting schemes. Each referendum has a specific
proposal associated with it that takes the form of a privileged function call in the runtime (that
includes the most powerful call:
set_code, which can switch out the entire code of the runtime,
achieving what would otherwise require a "hard fork").
Referenda are discrete events, have a fixed period where voting happens, and then are tallied and the function call is made if the vote is approved. Referenda are always binary; your only options in voting are "aye", "nay", or abstaining entirely.
Referenda can be started in one of several ways:
- Publicly submitted proposals;
- Proposals submitted by the council, either through a majority or unanimously;
- Proposals submitted as part of the enactment of a prior referendum;
- Emergency proposals submitted by the Technical Committee and approved by the Council.
All referenda have an enactment delay associated with them. This is the period between the referendum ending and, assuming the proposal was approved, the changes being enacted.
Referenda is considered baked if it is closed and tallied. Again, assuming the proposal was approved, it would be scheduled for enactment. Referenda is considered unbaked if it is pending an outcome, i.e. being voted on.
For the first two ways that a referendum is launched, this is a fixed time of 28 days. For the third type, it can be set as desired. Emergency proposals deal with major problems with the network that need to be "fast-tracked". These will have a shorter enactment time.
Proposing a Referendum
Anyone can propose a referendum by depositing the minimum amount of tokens for a certain period (number of blocks). If someone agrees with the proposal, they may deposit the same amount of tokens to support it - this action is called endorsing. The proposal with the highest amount of bonded support will be selected to be a referendum in the next voting cycle.
Note that this may be different from the absolute number of endorsements; for instance, three accounts bonding 20 AFT each would "outweigh" ten accounts bonding a single AFT each.
The bonded tokens will be released once the proposal is tabled (that is, brought to a vote).
There can be a maximum of 100 public proposals in the proposal queue.
Unanimous Council - When all members of the council agree on a proposal, it can be moved to a referendum. This referendum will have a negative turnout bias (that is, the smaller the amount of stake voting, the smaller the amount necessary for it to pass - see Adaptive Quorum Biasing).
Majority Council - When agreement from only a simple majority of council members occurs, the referendum can also be voted upon, but it will be majority-carries (51% wins).
There can only be one active referendum at any given time, except when there is also an emergency referendum in progress.
Every 28 days, a new referendum will come up for a vote, assuming there is at least one proposal in one of the queues. There is a queue for Council-approved proposals and a queue for publicly submitted proposals. The referendum to be voted upon alternates between the top proposal in the two queues.
The "top" proposal is determined by the amount of stake bonded behind it. If the given queue whose turn it is to create a referendum that has no proposals (is empty), and proposals are waiting in the other queue, the top proposal in the other queue will become a referendum.
Multiple referenda cannot be voted upon in the same period, excluding emergency referenda. An emergency referendum occurring at the same time as a regular referendum (either public- or council-proposed) is the only time that multiple referenda will be able to be voted on at once.
Voting on a referendum
To vote, a voter generally must lock their tokens up for at least the enactment delay period beyond the end of the referendum. This is in order to ensure that some minimal economic buy-in to the result is needed and to dissuade vote selling.
It is possible to vote without locking at all, but your vote is worth a small fraction of a normal vote, given your stake. At the same time, holding only a small amount of tokens does not mean that the holder cannot influence the referendum result, thanks to time-locking. You can read more about this at Voluntary Locking.
No with 10 AFT for a 128 week lock period => 10 x 6 = 60 votes.
Yes with 20 AFT for a 4 week lock period => 20 x 1 = 20 votes.
Yes with 15 AFT for a 8 week lock period => 15 x 2 = 30 votes.
Even though combined both Logan and Kevin vote with more AFT than Peter, the lock period for both of them is less than Peter, leading to their voting power counting as less.
Depending on which entity proposed the proposal and whether all council members voted yes, there are three different scenarios. We can use the following table for reference.
|Public||Positive Turnout Bias (Super-Majority Approve)|
|Council (Complete agreement)||Negative Turnout Bias (Super-Majority Against)|
|Council (Majority agreement)||Simple Majority|
Also, we need the following information and apply one of the formulas listed below to calculate the
voting result. For example, let's use the public proposal as an example, so the
Super-Majority Approve formula will be applied. There is no strict quorum, but the super-majority
required increases with lower turnout.
approve - the number of aye votes
against - the number of nay votes
turnout - the total number of voting tokens (does not include conviction)
electorate - the total number of tokens issued in the network
positive turnout bias, whereby a heavy super-majority of aye votes is required to carry at low
turnouts, but as turnout increases towards 100%, it becomes a simple majority-carries as below.
negative turnout bias, whereby a heavy super-majority of nay votes is required to reject at low
turnouts, but as turnout increases towards 100%, it becomes a simple majority-carries as below.
Majority-carries, a simple comparison of votes; if there are more aye votes than nay, then the proposal is carried, no matter how much stake votes on the proposal.
To know more about where these above formulas come from, please read the democracy pallet.
Assume we only have 1_500 AFT tokens in total and that this is a public proposal.
- John: 500 AFT
- Peter: 100 AFT
- Lilly: 150 AFT
- JJ: 150 AFT
- Ken: 600 AFT
Yes for a 4 week lock period =>
500 x 1 = 500 votes
Yes for a 4 week lock period =>
100 x 1 = 100 votes
No for a 16 week lock period =>
150 x 3 = 450 votes
- approve = 600
- against = 450
- turnout = 750
- electorate = 1500
Since the above example is a public referendum,
Super-Majority Approve would be used to calculate
Super-Majority Approve requires more
aye votes to pass the referendum when turnout
is low, therefore, based on the above result, the referendum will be rejected. In addition, only the
winning voter's tokens are locked. If the voters on the losing side of the referendum believe that
the outcome will have negative effects, their tokens are transferrable so they will not be locked
into the decision. Moreover, winning proposals are autonomously enacted only after some enactment
Allfeat utilizes an idea called
Voluntary Locking that allows token holders to increase their voting power by declaring how long
they are willing to lock up their tokens, hence, the number of votes for each token holder will be
calculated by the following formula:
votes = tokens * conviction_multiplier
The conviction multiplier increases the vote multiplier by one every time the number of lock periods double.
The maximum number of "doublings" of the lock period is set to 6 (and thus 32 lock periods in total), and one lock period equals 30 days. Only doublings are allowed; you cannot lock for, say, 24 periods and increase your conviction by 5.5. For additional information regarding the timeline of governance events, check out the governance section on the Allfeat Parameters page.
While a token is locked, you can still use it for voting and staking; you are only prohibited from transferring these tokens to another account.
Votes are still "counted" at the same time (at the end of the voting period), no matter for how long the tokens are locked.
Adaptive Quorum Biasing
Allfeat use a concept, "Adaptive Quorum Biasing", which functions as a lever that the council can use to alter the effective super-majority required to make it easier or more difficult for a proposal to pass in the case that there is no clear majority of voting power backing it or against it.
Let's use the above image as an example.
If a publicly submitted referendum only has a 25% turnout, the tally of "aye" votes has to reach 66%
for it to pass since we applied
Positive Turnout Bias.
In contrast, when it has a 75% turnout, the tally of "aye" votes has to reach 54%, which means that the super-majority required decreases as the turnout increases.
When the council proposes a new proposal through unanimous consent, the referendum would be put to a vote using "Negative Turnout Bias". In this case, it is easier to pass this proposal with low turnout and requires a super-majority to reject. As more token holders participate in voting, the bias approaches a plain majority carries.
Referring to the above image, when a referendum only has 25% turnout, the tally of "aye" votes has to reach 34% for it to pass.
In short, when the turnout rate is low, a super-majority is required to reject the proposal, which means a lower threshold of "aye" votes have to be reached, but as turnout increases towards 100%, it becomes a simple majority.
All three tallying mechanisms - majority carries, super-majority approve, and super-majority against - equate to a simple majority-carries system at 100% turnout.
To represent passive stakeholders, Allfeat introduces the idea of a "council". The council is an on-chain entity comprising several actors, each represented as an on-chain account. On Allfeat, the council currently consists of 10 members.
Along with controlling the treasury, the council is called upon primarily for some tasks of governance: proposing sensible referenda, and cancelling uncontroversially dangerous or malicious referenda.
For a referendum to be proposed by the council, a strict majority of members must be in favor, with no member exercising a veto. Vetoes may be exercised only once by a member for any single proposal; if, after a cool-down period, the proposal is resubmitted, they may not veto it a second time.
Council motions which pass with a 3/5 (60%) super-majority - but without reaching unanimous support - will move to a public referendum under a neutral, majority-carries voting scheme. In the case that all members of the council vote in favor of a motion, the vote is considered unanimous and becomes a referendum with negative adaptive quorum biasing.
A proposal can be canceled if the technical committee unanimously agrees to do so, or if Root origin (e.g. sudo) triggers this functionality. A canceled proposal's deposit is burned.
Additionally, a two-thirds majority of the council can cancel a referendum. This may function as a last-resort if there is an issue found late in a referendum's proposal such as a bug in the code of the runtime that the proposal would institute.
If the cancellation is controversial enough that the council cannot get a two-thirds majority, then it will be left to the stakeholders en masse to determine the fate of the proposal.
A proposal can be blacklisted by Root origin (e.g. sudo). A blacklisted proposal and its related referendum (if any) are immediately canceled. Additionally, a blacklisted proposal's hash cannot re-appear in the proposal queue. Blacklisting is useful when removing erroneous proposals that could be submitted with the same hash.
Upon seeing their proposal removed, a submitter who is not properly introduced to the democracy system of Allfeat might be tempted to re-submit the same proposal. That said, this is far from a fool-proof method of preventing invalid proposals from being submitted - a single changed character in a proposal's text will also change the hash of the proposal, rendering the per-hash blacklist invalid.
The council, being an instantiation of Substrate's Collective pallet, implements what's called a prime member whose vote acts as the default for other members that fail to vote before the timeout.
The prime member is chosen based on a Borda count.
The purpose of having a prime member of the council is to ensure a quorum, even when several members abstain from a vote. Council members might be tempted to vote a "soft rejection" or a "soft approval" by not voting and letting the others vote. With the existence of a prime member, it forces councillors to be explicit in their votes or have their vote counted for whatever is voted on by the prime.
The Technical Committee(TC) is composed of the teams that have successfully implemented or specified either a Allfeat runtime or Allfeat Host. Teams are added or removed from the TC via a simple majority vote of the Council.
The purpose of the TC is to safeguard against malicious referenda, implement bug fixes, reverse faulty runtime updates, or add new but battle-tested features. The TC has the power to fast-track proposals by using the Democracy pallet, and is the only origin that is able to trigger the fast-tracking functionality. We can think of the TC as a "unique origin" that cannot generate proposals, but are able to fast track existing proposals.
Fast-tracked referenda are the only type of referenda that can be active alongside another active referendum. Thus, with fast-tracked referenda it is possible to have two active referendums at the same time. Voting on one does not prevent a user from voting on the other.