Block producer DAO: Difference between revisions
m (→Pseudocode) |
|||
(13 intermediate revisions by 3 users not shown) | |||
Line 2: | Line 2: | ||
== Overview == | == Overview == | ||
A fork of [https://ethereum.org/ Ethereum] (or any other blockchain) with the original consensus mechanism initiates the new blockchain. The difference between an original blockchain consensus mechanism and a DGF variation is in governance. The fees are in the native currency token denoted $. Initially $ is ETH in the case of an Ethereum fork. Then REP is minted when a block is produced. Distribution of REP and $ fees follow [[DAO Governance Framework#DGF workflow|DGF flow]]: | A fork of [https://ethereum.org/ Ethereum] (or any other blockchain) with the original consensus mechanism initiates the new blockchain. The difference between an original blockchain consensus mechanism and a DGF variation is in governance. This section details [[executive governance]]. | ||
The fees are in the native currency token, denoted $. Initially $ is ETH in the case of an Ethereum fork. Then REP is minted when a block is produced. Distribution of REP and $ fees follow [[DAO Governance Framework#DGF workflow|DGF flow]]: | |||
# A block producer is a holder of REP. A producer signals their availability to produce blocks by encumbering their REP in the [[availability smart contract]] (ASC). | # A block producer is a holder of REP. A producer signals their availability to produce blocks by encumbering their REP in the [[availability smart contract]] (ASC). | ||
Line 8: | Line 10: | ||
# The selected producer follows the current protocol for block production which is encoded in the current [[work smart contract]] (WSC). | # The selected producer follows the current protocol for block production which is encoded in the current [[work smart contract]] (WSC). | ||
# When the block is finished it is posted to the [[Forum]] with the $ fees collected from the block, opening a [[Validation Pool]] (VPSC): | # When the block is finished it is posted to the [[Forum]] with the $ fees collected from the block, opening a [[Validation Pool]] (VPSC): | ||
## The fees include any transaction fees and any | ## The fees include any transaction fees and any newly minted $ from that block. These sources of $ are two types of production taxes to the system. | ||
## The VPSC mints new REP in proportion to the $ fees and encumbers the REP appropriately--default is 50% for the producer, 50% against the producer. | ## The VPSC mints new REP in proportion to the $ fees and encumbers the REP appropriately--default is 50% for the producer, 50% against the producer. | ||
## The Bench of experts evaluate the block and stake their REP on the outcome. | ## The Bench of experts evaluate the block and stake their REP on the outcome. | ||
Line 24: | Line 26: | ||
The entire high-level algorithm described above is automated as part of [[executive governance]]. Each vote in each validation pool is an explicit coordination action between the set of all nodes maintaining the blockchain. However, these votes are automated. | The entire high-level algorithm described above is automated as part of [[executive governance]]. Each vote in each validation pool is an explicit coordination action between the set of all nodes maintaining the blockchain. However, these votes are automated. | ||
If the node's client software recognizes a block as being properly constructed (lines 53/54), then the client votes to validate. If not, the client votes to reject. If a block is invalidated, the REP encumbered by the block producer in the ASC is slashed. This has the effect of motivating block producers to run the client software unhacked. | If the node's client software recognizes a block as being properly constructed (lines 53/54), then the client votes to validate. If not, the client votes to reject. If a block is invalidated, the REP encumbered by the block producer in the ASC is slashed. This has the effect of motivating block producers to run the client software unedited/unhacked. | ||
Participation in policing is naturally incentivized by the Validation Pool, which rewards participants with half of all newly minted REP. This also stabilizes the owership of REP as new REP does not all go to the lucky producer who was selected to produce the new block. | Participation in policing is naturally incentivized by the Validation Pool, which rewards participants with half of all newly minted REP. This also stabilizes the owership of REP as new REP does not all go to the lucky producer who was selected to produce the new block. | ||
==== Legislative ==== | ==== Legislative ==== | ||
The above description of the algorithm does not include the important aspect of [[legislative governance]]. Continual improvements to the algorithm and software are incentivized since the WSC will include references to fund governance proposals that update the WSC, itself, and any other hard protocols that are improved. As detailed in the legislative governance page, such proposals involve slow and conscious deliberation and voting between members, unlike the automated executive governance of the client software for active block production. Nascent legislative governance is illustrated in the coding section, below. | The above description of the algorithm does not include the important aspect of [[legislative governance]]. Continual improvements to the algorithm and software are incentivized since the WSC will include references to fund governance proposals that update the WSC, itself, and any other hard protocols that are improved. As detailed in the legislative governance page, such proposals involve slow and conscious deliberation and voting between members, unlike the automated executive governance of the client software for active block production. Nascent legislative governance is illustrated in the [[Block producer DAO#reference mechanism|coding section]], below. | ||
==== Judicial ==== | ==== Judicial ==== | ||
[[Judicial governance]] under DGF makes it possible to retroactively slash REP that was used to violate the values of the Block | [[Judicial governance]] under DGF makes it possible to retroactively slash REP that was used to violate the values of the Block producer DAO. This allows the group to implement governance to prevent spirit-of-the-law-violations, such as censorship. Censorship in block production means, for example that a block producer chooses not to include a transaction in a block. Reasons for censorship include arbitrage opportunities such as leaving out votes in the validation pool so those producers don't share in the rewards for policing, manipulating the pseudorandom selection algorithm by leaving out ASCs, and leaving out customers' transactions to arbitrage markets. | ||
The mechanism for block producer selection detailed above ties all the REP generated by participating in the DAO together. That makes REP slashing increasingly effective as time goes on for disincentivizing Byzantine behavior. That makes the system increasingly secure at time goes on. | |||
=== Scaling === | === Scaling === | ||
Line 38: | Line 42: | ||
== Variations == | == Variations == | ||
There are many improvements possible on every aspect of the basic outline given above. DGF provides an evolutionary structure | There are many improvements possible on every aspect of the basic outline given above. DGF provides an evolutionary structure which incentivizes the continual improvement of the system. This is necessary for any PoS consensus mechanism, given the inevitable threat of [[Transcendent values#Game theory argument|gaming]], any cooperative system leads to an arms race of attack and defense as the system becomes more successful and valuable. A system devoted to creating valuable currency, globally open to anonymous members, absolutely demands a properly balanced incentive mechanism against gaming. | ||
In this section we give an introduction to areas that will be continually improved. | In this section we give an introduction to areas that will be continually improved. | ||
Line 50: | Line 54: | ||
Improvements to the automated selection process can be made to select REP in a more equitable manner. The lottery process of selecting members according to power, statistically will maintain power ratios if all members are continually encumbering all their REP in ASCs. Larger wallets of REP will always remain larger if the owner continually runs the block production software, and the effect is linear (stochastic process argument). However, we can incentivize homogeneity between wallets if we implement a more complicated selection process of remembering which REP was selected: REP that has been selected for block production is not selected again until the entire list of active ASCs are selected. This encourages sock puppets, but that effect is mollified by tying together REP as in the previous subsection. | Improvements to the automated selection process can be made to select REP in a more equitable manner. The lottery process of selecting members according to power, statistically will maintain power ratios if all members are continually encumbering all their REP in ASCs. Larger wallets of REP will always remain larger if the owner continually runs the block production software, and the effect is linear (stochastic process argument). However, we can incentivize homogeneity between wallets if we implement a more complicated selection process of remembering which REP was selected: REP that has been selected for block production is not selected again until the entire list of active ASCs are selected. This encourages sock puppets, but that effect is mollified by tying together REP as in the previous subsection. | ||
=== Block | === Block producer selection === | ||
At Step 2 in the above process, the block producer is pseudo-randomly selected from the active Availability Smart Contracts. If this is not done carefully, a Byzantine actor who is the latest active block producer can anticipate the pseudo-randomness and manipulate the algorithm so that he is selected again, in perpetuity, which would destroy the decentralization of the blockchain. | At Step 2 in the above process, the block producer is pseudo-randomly selected from the active Availability Smart Contracts. If this is not done carefully, a Byzantine actor who is the latest active block producer can anticipate the pseudo-randomness and manipulate the algorithm so that he is selected again, in perpetuity, which would destroy the decentralization of the blockchain. | ||
Line 57: | Line 61: | ||
This creates complications for reasoning about what to do when the network is fragmented and some members fail to participate, or when a block producer censors commits or reveals. However, such issues can be policed and attacks disincentivized with [[Block producer DAO#Judicial|judicial governance]]. | This creates complications for reasoning about what to do when the network is fragmented and some members fail to participate, or when a block producer censors commits or reveals. However, such issues can be policed and attacks disincentivized with [[Block producer DAO#Judicial|judicial governance]]. | ||
== Code == | == Code == | ||
The following two stages illustrate the basic algorithm for creating a | The following two stages illustrate the basic algorithm for creating a BpDAO. | ||
The first stage, given by the first sequence diagram and pseudocode illustrates basic [[DAO Governance Framework#DGF workflow|DGF flow]] applied to block production. This is | === Workflow === | ||
The first stage, given by the first sequence diagram and pseudocode illustrates basic [[DAO Governance Framework#DGF workflow|DGF flow]] applied to block production. This is PoR, i.e., Proof of REP. [[File:BPSeqDiag.png|center|thumb|890x890px]] | |||
<syntaxhighlight lang="text" style="font-size:100%" line="1"> | <syntaxhighlight lang="text" style="font-size:100%" line="1"> | ||
Line 136: | Line 143: | ||
} | } | ||
} | } | ||
</syntaxhighlight>The next sequence diagram illustrates the reference mechanism, which provides a means of bootstrapping governance by making proposals which cite older proposals. This gives an incentive mechanism for a community to organically evolve | </syntaxhighlight> | ||
=== Reference mechanism === | |||
The next sequence diagram illustrates the reference mechanism, which provides a means of bootstrapping governance by making proposals which cite older proposals. This gives an incentive mechanism for a community to organically improve the block producer protocols. | |||
[[File:RefMechanismSeqDiag2.png|center|thumb|848x848px]] | |||
For example, to improve the consensus mechanism WSC, the idea is to evolve the conversation in 4 phases. | |||
In Phase 1, people chat about improvements they would like to see. | |||
In Phase 2, a collection of improvement proposals are offered. Improvement proposals cite older proposals that they build on (detailed in the second sequence diagram, above). Improvement proposals are continually made, in perpetuity. | |||
In Phase 3, the latest favored proposal will become the active WSC, ''de facto'' because it is used by the group. The active WSC primarily rewards the worker whose ASC is selected with the newly minted REP. Secondarily the WSC sends part of this REP reward to the author of the WSC. The sub-DAG of proposals connected by references to the currently favored proposal will all be rewarded with REP and subsequently $. Proposals not connected with references are not rewarded. | |||
In Phase 4, a new type of REP is invented, called gREP, to promote active, explicit, on-chain governance of the Block producer DAO. The active WSC will be changed to include a line sharing a percentage of the monetary rewards for block production as gREP salary. Owners of gREP use it for legislative governance, i.e., to make proposals (the work of the governance wing of the DAO) and to vote on which proposals are canonized by binding validation pools using gREP. The amount set aside for governance is a continual negotiation between block producers (wREP holders), and governors (gREP holders), who will typically overlap to some degree. That amount is the dev budget. | |||
In summary, the means by which a community creates a DAO using DGF is to engender two systems: DGF flow (for executive governance) and the reference mechanism (for legislative governance). | |||
== Applications == | == Applications == | ||
Latest revision as of 14:35, 26 February 2024
DGF is used to make Block producer DAOs (BpDAOs) which govern block production in blockchains. These BpDAOs consist of members who earn REP tokens for producing canonical blocks and governing the development of the system. The REP minting mechanism of DGF incentivizes faithful production of blocks with a consensus mechanism called Proof of Reputation (PoR). DAO governance guides and rewards improvements to the platform.
Overview[edit | edit source]
A fork of Ethereum (or any other blockchain) with the original consensus mechanism initiates the new blockchain. The difference between an original blockchain consensus mechanism and a DGF variation is in governance. This section details executive governance.
The fees are in the native currency token, denoted $. Initially $ is ETH in the case of an Ethereum fork. Then REP is minted when a block is produced. Distribution of REP and $ fees follow DGF flow:
- A block producer is a holder of REP. A producer signals their availability to produce blocks by encumbering their REP in the availability smart contract (ASC).
- Previous blocks pseudorandomly select a producer from the active ASCs.
- The selected producer follows the current protocol for block production which is encoded in the current work smart contract (WSC).
- When the block is finished it is posted to the Forum with the $ fees collected from the block, opening a Validation Pool (VPSC):
- The fees include any transaction fees and any newly minted $ from that block. These sources of $ are two types of production taxes to the system.
- The VPSC mints new REP in proportion to the $ fees and encumbers the REP appropriately--default is 50% for the producer, 50% against the producer.
- The Bench of experts evaluate the block and stake their REP on the outcome.
- VPSC resolves the vote, distributing the REP to the winners in proportion to their stakes.
- VPSC distributes the $ fees to all members of the DAO in proportion to their REP holdings (REP salary).
- Repeat
Comments[edit | edit source]
Governance[edit | edit source]
Executive[edit | edit source]
The entire high-level algorithm described above is automated as part of executive governance. Each vote in each validation pool is an explicit coordination action between the set of all nodes maintaining the blockchain. However, these votes are automated.
If the node's client software recognizes a block as being properly constructed (lines 53/54), then the client votes to validate. If not, the client votes to reject. If a block is invalidated, the REP encumbered by the block producer in the ASC is slashed. This has the effect of motivating block producers to run the client software unedited/unhacked.
Participation in policing is naturally incentivized by the Validation Pool, which rewards participants with half of all newly minted REP. This also stabilizes the owership of REP as new REP does not all go to the lucky producer who was selected to produce the new block.
Legislative[edit | edit source]
The above description of the algorithm does not include the important aspect of legislative governance. Continual improvements to the algorithm and software are incentivized since the WSC will include references to fund governance proposals that update the WSC, itself, and any other hard protocols that are improved. As detailed in the legislative governance page, such proposals involve slow and conscious deliberation and voting between members, unlike the automated executive governance of the client software for active block production. Nascent legislative governance is illustrated in the coding section, below.
Judicial[edit | edit source]
Judicial governance under DGF makes it possible to retroactively slash REP that was used to violate the values of the Block producer DAO. This allows the group to implement governance to prevent spirit-of-the-law-violations, such as censorship. Censorship in block production means, for example that a block producer chooses not to include a transaction in a block. Reasons for censorship include arbitrage opportunities such as leaving out votes in the validation pool so those producers don't share in the rewards for policing, manipulating the pseudorandom selection algorithm by leaving out ASCs, and leaving out customers' transactions to arbitrage markets.
The mechanism for block producer selection detailed above ties all the REP generated by participating in the DAO together. That makes REP slashing increasingly effective as time goes on for disincentivizing Byzantine behavior. That makes the system increasingly secure at time goes on.
Scaling[edit | edit source]
Governance under DGF allows nimble reweighting of parameters in the WSC to adjust to network performance. For example, when the number of block producers grows to make validation pool coordination cumbersome, random selections of subpopulations of the block producers can be made each round to optimize for security, latency, throughput, and bandwidth.
Variations[edit | edit source]
There are many improvements possible on every aspect of the basic outline given above. DGF provides an evolutionary structure which incentivizes the continual improvement of the system. This is necessary for any PoS consensus mechanism, given the inevitable threat of gaming, any cooperative system leads to an arms race of attack and defense as the system becomes more successful and valuable. A system devoted to creating valuable currency, globally open to anonymous members, absolutely demands a properly balanced incentive mechanism against gaming.
In this section we give an introduction to areas that will be continually improved.
Availability Smart Contract[edit | edit source]
Pseudonymous REP tied together without KYC[edit | edit source]
One important aspect of this implementation of DGF is that a block producer's REP can be tied together without exposing the owner of the wallet holding the REP. The REP used to run the automated block production client software grows automatically as they participate in block production and policing. The REP encumbered in the ASC can be tied to the REP gained by minting blocks when their ASC is selected. If it is ever necessary to slash a block producer's REP, then all of the REP they've used can be slashed without revealing the block producer's identity.
More stable selection[edit | edit source]
Improvements to the automated selection process can be made to select REP in a more equitable manner. The lottery process of selecting members according to power, statistically will maintain power ratios if all members are continually encumbering all their REP in ASCs. Larger wallets of REP will always remain larger if the owner continually runs the block production software, and the effect is linear (stochastic process argument). However, we can incentivize homogeneity between wallets if we implement a more complicated selection process of remembering which REP was selected: REP that has been selected for block production is not selected again until the entire list of active ASCs are selected. This encourages sock puppets, but that effect is mollified by tying together REP as in the previous subsection.
Block producer selection[edit | edit source]
At Step 2 in the above process, the block producer is pseudo-randomly selected from the active Availability Smart Contracts. If this is not done carefully, a Byzantine actor who is the latest active block producer can anticipate the pseudo-randomness and manipulate the algorithm so that he is selected again, in perpetuity, which would destroy the decentralization of the blockchain.
To guarantee the pseudo-random selection mechanism is resistant to such capture, there are several algorithms available. Here we describe a basic Commit-Reveal scheme, such as the Randao algorithm. Basically, selection of the block producer is done in two stages. First, every block producer sends an encrypted salt message for the pseudorandom algorithm to the current block producer to include in the current block. Then after the block is published, the block producers send in their keys to unencrypt their salt messages in the next block. The union of the salts is added to the pseudorandom number generator to determine the next block producer.
This creates complications for reasoning about what to do when the network is fragmented and some members fail to participate, or when a block producer censors commits or reveals. However, such issues can be policed and attacks disincentivized with judicial governance.
Code[edit | edit source]
The following two stages illustrate the basic algorithm for creating a BpDAO.
Workflow[edit | edit source]
The first stage, given by the first sequence diagram and pseudocode illustrates basic DGF flow applied to block production. This is PoR, i.e., Proof of REP.
Block Production Algorithm Pseudo-code
Inputs: dynamic set of desired transactions from customers
Output: a blockchain of settled transactions
__________
Definitions
REP := reputation token
$ := currency token
wallet := program for holding tokens for an individual participant
Tx := transaction sending a currency token from one wallet to another
customer := has
* unique identifying address
* wallet for holding $ and sending Txs
block producer := has
* unique identifying address
* wallet for receiving $ and REP
* ability to give inputs to ASC, WSC, & VPSC
ASC := Availability Smart Contract
WSC := Work Smart Contract
VPSC := Validation Pool Smart Contract
Forum := database holding history of blocks & Validation Pool results
gossip network := set of all active Tx held by block producers
Validation Pool := dynamically changing set of block producers who are currently policing a proposed block
__________
Algorithm
For i = 1 to infinity
Produce block i := {
WSC function:= {
* listens to active ASCs
* pseudorandomly selects 1 block producer based on weights of encumbered REP in ASCs
* announces current block producer
* takes current block producer's REP from ASC
* waits for a set of Txs from current block producer
* produces a block
* publishes proposed block to Forum
* sends 4 things to the VPSC:
1. $ from Txs
2. new $ minted in block
3. block producer's REP from ASC
4. address of proposed block
}
VPSC function: := {
* mints REP tokens in proportion to $ tokens
* announces new validation pool and proposed block address to block producers
* collects votes from block producers
* tallies votes to determine distribution of new REP tokens
* distributes new REP
* distributes REP salary
* publishes vote results to Forum
[Forum signals to Tx status to customer]
* signals WSC to produce next block
}
}
Reference mechanism[edit | edit source]
The next sequence diagram illustrates the reference mechanism, which provides a means of bootstrapping governance by making proposals which cite older proposals. This gives an incentive mechanism for a community to organically improve the block producer protocols.
For example, to improve the consensus mechanism WSC, the idea is to evolve the conversation in 4 phases.
In Phase 1, people chat about improvements they would like to see.
In Phase 2, a collection of improvement proposals are offered. Improvement proposals cite older proposals that they build on (detailed in the second sequence diagram, above). Improvement proposals are continually made, in perpetuity.
In Phase 3, the latest favored proposal will become the active WSC, de facto because it is used by the group. The active WSC primarily rewards the worker whose ASC is selected with the newly minted REP. Secondarily the WSC sends part of this REP reward to the author of the WSC. The sub-DAG of proposals connected by references to the currently favored proposal will all be rewarded with REP and subsequently $. Proposals not connected with references are not rewarded.
In Phase 4, a new type of REP is invented, called gREP, to promote active, explicit, on-chain governance of the Block producer DAO. The active WSC will be changed to include a line sharing a percentage of the monetary rewards for block production as gREP salary. Owners of gREP use it for legislative governance, i.e., to make proposals (the work of the governance wing of the DAO) and to vote on which proposals are canonized by binding validation pools using gREP. The amount set aside for governance is a continual negotiation between block producers (wREP holders), and governors (gREP holders), who will typically overlap to some degree. That amount is the dev budget.
In summary, the means by which a community creates a DAO using DGF is to engender two systems: DGF flow (for executive governance) and the reference mechanism (for legislative governance).