status

A Journey Through Decentralized Protocol Design

Published 27.3.2024

The introduction of the Edinburgh Decentralization Index (EDI) ignited discussions about the appropriate methods for accurately quantifying decentralization. Justin Bons was among the early critics of the EDI, arguing that it overly favored Cardano in its results. It’s worth considering the essential features a protocol should possess to enhance its decentralization, and whether Justin’s critique holds merit.

Features Of A Decentralized Protocol

Decentralization revolves around the distribution of decision-making authority. This can pertain to aspects such as block production, governance, or client diversification. As it stands, the EDI primarily provides information on block production, which will be the focus of the article.

The ultimate aim of decentralization is to prevent a single entity or a consortium of influential entities from monopolizing decision-making processes within the network. In an ideal scenario, each participant would possess equal voting rights, akin to the ‘one person, one vote’ principle observed in political elections. However, within the realm of blockchain, decision-making power is tethered to possessing an expensive resource. Consequently, entities with more resources wield greater decision-making power, embodying the principle of ‘more resources, more power’.

A decentralized protocol, unlike a democracy, for instance, possesses distinct attributes. Let’s contemplate the characteristics that contribute to heightened decentralization:

  • Inclusivity
  • Fairness
  • Reliable Block Production
  • Resistance to Cartels and Whales

Inclusivity implies that the broadest possible array of entities, including those with limited resources, can partake in network consensus.

This encompasses both initial and operational costs. It could involve purchasing ASIC miners and regular electricity payments in the case of Bitcoin or acquiring ADA coins in the case of Cardano.

If a participant wishes to produce blocks, they must operate a full node and maintain a continuous internet and electricity connection (fail-over solution).

Inclusivity also entails simplicity. If participants are subjected to high technical demands, it could hinder the growth of decentralization. Ideally, participation in decentralization should require minimal technical skills. Thus, a light wallet should suffice for the participant.

Fairness signifies that the network’s and the participants’ interests align well. Only those who act in the network’s and other participants’ & users’ best interest are entitled to rewards. Everyone should receive a fair reward for quality work for the network. The protocol alone should direct the distribution of rewards, i.e., without third-party involvement.

Block production must be reliable. In a given period, the maximum possible number of blocks should be produced with a minimum of exceptions. This could occur, for instance, if someone operates a node in an area prone to frequent power outages, or where the internet is unreliable.

Block production can be a lucrative business in the event of a successful project. Interest will surge. Individual participants will compete with each other and may strive to oust competitors. Large players benefit from economies of scale. For instance, large PoW miners can secure discounts from ASIC hardware manufacturers or energy suppliers. For the sake of decentralization, the size advantage should be as minimal as possible.

Mechanisms must exist to prevent the formation of cartels (a significant issue with the EOS project) and the emergence of whales (more than 50% of blocks in Bitcoin are mined by 2 pools).

However, in practice, it is not feasible to achieve an ideal solution, as wealth is unevenly distributed in society. Therefore, it is possible to purchase decision-making power in a decentralized network. The only achievable goal is the best possible solution under unfavorable conditions.

The team must consider not only the characteristics that encourage decentralization but also others. For example, scalability, interoperability, security, sustainability, and so on. It is difficult to balance all aspects correctly. The result is always a balance between preferred properties.

Source Pooling Is A Must

I posit that the pinnacle of decentralization can only be attained when anyone, even those with minimal resources, can participate and stand a chance to reap rewards.

The entry cost to participate in Bitcoin mining is steep, with ASIC miners costing several thousand USD. Furthermore, energy prices vary across countries, making consensus participation exclusive. This leads to a relatively low level of network decentralization, with larger miners progressively gaining strength.

Operating an Ethereum validator necessitates the locking up of 32 ETH, establishing an exclusive input boundary defined by the protocol. Those with less than 32 ETH who wish to participate in the consensus must resort to solutions that pool together the ETH of smaller holders and run a validator on their behalf, such as CEXs and services akin to Lido.

Both Bitcoin and Ethereum involve some form of resource pooling. Multiple participants must band together to stand a chance at receiving a reward. A lone miner may attempt to mine BTC, but the odds of reward are slim. However, if 1000 miners unite, their chances of reward increase, including more frequent rewards.

Bitcoin has pools, while Ethereum has services similar to Lido.

The Bitcoin protocol remains oblivious to the existence of pools. It simply rewards the miner of the block, without any knowledge of the entity aggregating the hash rate and distributing work among multiple miners.

The situation mirrors that of Ethereum. There are numerous validators, and the protocol is indifferent to who operates them and who possesses the ETH keys. Lido represents a weak point in decentralization, as over 30% of all staked ETH resides in a single smart contract.

In every existing ecosystem, a similar scenario will inevitably emerge sooner or later. Participants who have bought a resource will often delegate it to the entity producing the blocks. Therefore, when quantifying decentralization, it’s imperative to take this into account.

The IOG team recognized that resource pooling is a prevalent and likely inevitable occurrence. Moreover, for the highest level of decentralization to be realized, it’s desirable for all ADA coin holders, irrespective of the quantity, to have the opportunity to participate.

As such, Cardano has incorporated the concept of pools directly into its protocol. This enables Cardano to have native liquid staking, eliminating the need for third parties like Ethereum.

Without some form of resource pooling, all minor holders would be barred from participating in the consensus, which is not a desirable outcome.

There are several other pertinent reasons why resource pooling is beneficial.

One such advantage is that block production is guaranteed by IT professionals who can effortlessly create a functional fail-over solution and comprehend security. If block-producing nodes were operated by non-experts, PoS coins would frequently be stolen. In the Cardano ecosystem, this occurred a few times at the onset of staking. However, such incidents are now a thing of the past.

If every resource holder (for instance, someone who only owns 100 ADA) could operate their block-producing node, the quality of the production of blocks would deteriorate. Numerous blocks would be missed, perhaps because the automatic update of the operating system would initiate at the moment when the node was scheduled to produce a new block.

Another consideration is that as the number of nodes that must reach consensus increases for certain protocols, so does the complexity. For instance, with Ethereum, a high number of validators is not desirable. This is why the 32 ETH limit is necessary, and reducing it is not the solution. If the limit were lowered, the number of validators would rise, thereby increasing the complexity of mutual communication, which is undesirable.

Delegating a resource to a block producer (pool) is a favorable occurrence, but only when the delegators maintain control over the resource and have the flexibility to alter their decision at will. From a security standpoint, it's detrimental if delegators are compelled to relinquish control over the resource. For instance, transferring ETH to an exchange to participate in staking is an unfavorable practice. There's a risk that the ETH could be stolen by an insider at the exchange or by a hacker.

Delegation can be categorized into two types: direct and indirect. Direct delegation, as exemplified by the delegation of ADA coins to a selected Cardano pool, allows the coins to remain in the owner’s wallet. The owner retains the flexibility to modify the delegation at any point. On the other hand, indirect delegation involves the owner forfeiting direct control over the resource, rendering them unable to alter the delegation during a critical moment, such as a network attack. Instances of this include cloud mining in the context of Bitcoin, or CEX staking ETH in the case of Ethereum.

How Should We Quantify Decentralization Well?

In assessing decentralization, it’s crucial to take into account the number of entities that produce blocks, that is, those that operate a full node. Additionally, the resources utilized by block producers, whether they are exclusively owned or delegated, should also be considered.

Key factors to consider when quantifying decentralization include:

  • The number of block producers (pools)
  • The count of resource holders (those participating in the consensus)

For instance, if a Centralized Exchange (CEX) stakes ETH on behalf of 100K users, from the Ethereum protocol’s perspective, the CEX is the owner of the ETH, not the individuals who purchased the ETH. In this scenario, the CEX becomes a single point of failure and the entity accountable for block production.

Lido is something very similar to CEX, yet more decentralized.

It’s crucial to differentiate between block producers and delegators (stakers) for several reasons.

Block producers are the entities that actively participate in the creation of new blocks in the blockchain. They validate transactions, create new blocks, and add them to the blockchain. Their role is critical in maintaining the integrity and functionality of the network.

On the other hand, delegators, or stakers, are participants who delegate their resources (like coins or hash rate) to these block producers. While they may not actively participate in block production, their role is equally important. They support the network by providing their resources and in return, they share in the rewards generated by block production.

The distinction between these two roles is essential for understanding the dynamics of blockchain networks. It helps clarify the distribution of responsibilities and rewards within the network, and it’s a key factor in assessing the level of decentralization in the network.

The more distributed the block production and delegation, the more decentralized the network is.

In the context of Bitcoin, miners delegate their computational power (hash rate) to mining pools, which act as block producers. This delegation allows miners to have a more predictable income, and it enables the network to have a more stable block production. However, it’s important to note that this also introduces a level of centralization, as the control over block production is concentrated in a few large mining pools. This is one of the many trade-offs in the design of decentralized networks.

The concept of staking in Cardano bears a strong resemblance to the mining process in Bitcoin. However, a key distinction lies in the aspect of inclusivity. This characteristic in Cardano fosters a greater degree of decentralization. Cardano boasts a larger number of pools and, in all likelihood, a higher count of stakers compared to the number of miners in the Bitcoin network.

In all other networks that the EDI tracks, as well as those not yet included in the EDI, we observe comparable dynamics. I am not aware of any network where every resource holder, irrespective of the quantity, has the chance to produce blocks and actually does so.

Dynamics Between Block Producers And Delegators

Let’s delve into the dynamics between block producers and delegators, using Cardano as an illustrative example. In principle, it is very similar to Bitcoin, but with several significant differences.

In the Cardano network, a pool is a node that produces blocks. The total stake of the pool, which is the sum of the ADA coins held by the pool operator and all stakers (delegators), determines the number of blocks that the pool mints within a given period.

You could liken each pool operator to an independent entity akin to Lido.

A pool is a single node that is operated (owned) by pool operators. All stakers bolster this pool with their ADA coins, and the pool mints blocks on behalf of all stakers.

The pool operator has an incentive to act fairly, as their reward is contingent on the total stake of the pool. Therefore, the operators rely on the ADA coins delegated to them.

ADA holders can delegate coins to their chosen pool, but the ADA coins never leave their wallets. They can even spend ADA coins at any time, as Cardano does not have a locking period (a no slashing).

This implies that if the pool operator begins to censor transactions or fails to mint blocks, stakers can promptly exit the pool and delegate elsewhere (the effect of redelegation takes a few days).

ADA holders, who control the production of blocks, can be viewed as granting a mandate to pool operators to serve the network.

Cardano has safeguards in place to prevent the formation of large pools. The size of the pool is restricted by what is known as the saturation point. Once a pool surpasses the saturation point, the rewards for everyone are diminished. Stakers, therefore, have an economic incentive to delegate elsewhere.

The network includes multi-pool operators, i.e., entities that operate multiple pools. These are typically well-known figures in the ecosystem who have earned the trust of stakers. However, from a decentralization perspective, they are considered whales.

The saturation point is more of a psychological barrier, as it becomes apparent that a given entity operates multiple pools.

Of course, a single entity can operate multiple pools under different tickers, and this fact may not be observable from the outside.

In the Cardano ecosystem, pools compete with each other under similar conditions (fees). It’s relatively straightforward for stakers to delegate to another pool. Their choice can be based not only on fees, but also on the additional contributions operators make to the community or other preferred factors.

What Justin Bons may not realize when criticizing EDI

Justin: ‘The biggest problem with the EDI is that they count all pools as single entities’

Justin: ‘This is clearly a flawed approach to measuring decentralization’

In the realm of blockchain, pools serve as block producers and should be considered as individual entities. Delegators, or stakers, do not produce blocks themselves but rather supervise the process of block production.

From a security standpoint, such as in the context of Distributed Denial of Service (DDoS) attacks, it’s beneficial to have a large number of block-producing nodes within the network. This makes the network less vulnerable to attacks. This is why the EDI team researches Shannon Entropy. The more unpredictable the next block-producing node is, the more secure the network becomes.

A DDoS attack on a light wallet, i.e., a delegator, is not feasible. While it’s possible to launch a DDoS attack on a miner, it would have a minimal impact on the production of blocks in the Bitcoin network. On the other hand, a DDoS attack on major mining pools like Foundry USA or AntPool could instantly disrupt the quality of block production.

If only the resource holders and not the block producers were counted, EDI would not tell anything about who is producing the blocks.

Justin: ‘A valid metric; it wouldn't be so bad if they didn't count all pools as single entities. The alternative solution here is to identify individual pool participants through chain analysis & other means.’

A whale with 10M ADA can run its pool and will produce fewer blocks than a pool with 5K stakers that has a total stake of 50M ADA.

Consider who has direct control over the selection of transactions for a block or the censorship of transactions. It’s solely the pool operator, not the delegators.

The decision-making process of a single entity differs from that of a group of 5K independent entities collaborating to produce blocks.

A single entity can autonomously decide to start censoring transactions. However, a group of 5K entities would likely disagree on this. If an individual within this group wanted to censor transactions, they would need to exit the group.

Let’s view this from another angle.

From the perspective of potentially initiating an attack on the network (for instance, a double-spend attack), the entity controlling the node is crucial. Delegators do not have direct control over the node. A pool operator can initiate such an attack regardless of delegators. He can hope that the attack will be successful before the delegators notice.

It’s untenable to claim that a pool is not a valid entity or that it shouldn’t be counted as a single entity. If Justin asserts this, it’s as if he’s claiming that one of the delegates has control over whether the node will censor transactions. However, objectively, no delegator possesses this power.

I concur with Justin’s assertion that the Edinburgh Decentralization Index (EDI) lacks data on the quantity of resource holders.

Within the Bitcoin network, two pools, Foundry USA and AntPool, can launch an attack. However, miners have the option to delegate their resources elsewhere, assuming they detect the attack promptly.

Large miners can collude, establish their pool, and orchestrate an attack through it. Hence, it’s crucial to maintain an overview of the resource holders. There are whales in every PoS network and it's good to know about them.

Delegators, or a subset of them, can consent to the attack and thereby collaborate with the block producer. This raises the question: should the Nakamoto coefficient be exclusively applicable to block producers, or should a similar coefficient be established for delegators? In my view, both approaches are valid.

Conclusion

In my opinion, Justin’s criticism appears to stem from a misunderstanding of crucial contexts related to decentralization (different types of attacks, complexity of communication between nodes, inclusivity, etc.) or a lack of comprehension of the difference between block producers and delegators. He likely believes that it’s pointless to focus on block producers and that only resource distribution is important.

It’s essential to recognize the differences in the roles of these actors. Even a group of 5K delegators won’t produce a single block unless one of them operates a full node that inserts transactions into this block and disperses it across the network. Each of these delegators cannot have their block-producing node, as they would have to agree (consensus) on which of them will produce a block at a given moment.

It can be objectively stated that the position of the block producer is stronger than the position of the delegator.

The Edinburgh Decentralization Index (EDI) is an open-source initiative, open to contributions from all. If there’s dissatisfaction with Cardano being ranked as the most decentralized based on the results, it’s crucial to identify objective inaccuracies in the methodology. Justin contends that considering pools as individual entities is erroneous. However, in my perspective, this argument lacks validity.

Featured:

Related articles

Did you enjoy this article? Other great articles by the same author