Reflections on Ethereum in 2023: The road ahead is long and we must not slack off

The following are the views and thoughts of the author of this article, Danny Ryan, which inevitably contain short-sighted and subjective factors.

After the merger, I spent quite a bit of time reflecting on the past five years as we look to the future. I had intended to write some retrospective posts in 2022, but the timing was never right — everything was already building up, and the core goal at the time was crystal clear: ending Proof of Work (PoW).

Now, the time has finally come, and after completing what I consider to be the most important upgrade in blockchain history, we have successfully moved toward withdrawals and initial scaling upgrades, and more major goals and upgrades are within reach. While the “Ethereum Killers” are stuck in a bear market, and Bitcoin’s culture and technology are struggling in a boring (but orderly?!) wasteland, we continue to build, improve, philosophize, and make an impact, and Ethereum has never been stronger.

The technical roadmap ahead of us is long and complex, but I am confident that we can execute on the timeline. We have proven that we have a high level of delivery capabilities, and that the Ethereum ecosystem handles things the right way (often the hard way) has made and will continue to make Ethereum more resilient in all aspects. Despite the large amount of fraud in the industry, Ethereum still stands firm; in the abyss of the bear market, Ethereum continues to grow and innovate; in the turmoil of the world, Ethereum finds ways to make an important impact.

While I will provide some reflections on the upcoming technical challenges, these reflections are more on some existential issues: Layer1 ecosystem health, capture, cartels, L2 alignment with Ethereum, and more.

technology

Ethereum’s 2022 performance has been phenomenal by any standard, and no one could have predicted that The Merge would go so smoothly. Even with the delay of several years, this is a watershed moment for the Ethereum protocol, the Ethereum community, and the crypto world. Ethereum has proven that it can not only achieve major technical goals, but also do so in a largely decentralized manner. The Merge is a unique achievement - no other project can match its combination of technical sophistication, execution capabilities, and global decentralization.

The merger has proven to me and to the world that a project and structure like this not only works, but that when it works, it is stronger and more resilient.

The switch to Proof of Stake (PoS) took longer than any of us expected. When I first started working at EF, Aya Miyaguchi asked us in the spring of 2018 when we would be able to complete the transition to PoS. I remember my answer was something like 8 months. For many reasons, I underestimated this time to a ridiculous degree. But finally, we have a multi-layered, rich multi-client, resilient ecosystem and protocol supporting a level of decentralized staking that no other blockchain has ever attempted. Yes, there are areas for improvement, and there are protocols and companies competing for control of the entire staking network, but in any case, the Ethereum PoS mechanism is working well.

The road ahead is long and difficult

While the merger brings a sigh of relief as a key sustainability and security milestone, the road ahead is still long, especially when considering where we are in 2023, with special interest groups becoming increasingly sophisticated, regulatory action looking inevitable, and applications and layer 2 protocols craving scalability and a foundational framework that will support development for decades to come.

Simply looking at the Ethereum roadmap Vitalik posted, you can’t help but feel the confusing weight of the long and complex years (5? 10?) ahead before the protocol is truly “finished.” We cannot become complacent with the merged delivery, but should use this momentum, and the expertise, tools, and processes to solve the many challenges that remain.

But what does "to be done" mean? It means to achieve a sufficient final state in all aspects. What counts as sufficient functionality, security, decentralization... It is crucial to have a clearer understanding of what to optimize and the minimum thresholds, both philosophically and technically. Having a clear understanding of these will help plan priorities in the future and find the right trade-off between complexity and functionality. This also helps to ossify the protocol, which we will talk about later.

Current technical issues

Although we still have quite a bit to do, many of these issues no longer bother me. Nevertheless, there are still some active areas of R&D that require attention in the short and medium term. Here are some of the technical points I have selected.

PoS Evolution

Between proposer/builder separation (PBS), LMD-GHOST confirmation rules, single slot finality (SSF), single secret leader election (SSLE), proof of custody, proof of execution (PoE), and early RANDAO reveal slashing, there is still a lot of research, specification, and construction work to be done to continue to evolve Ethereum Proof of Stake towards its final, most secure and sustainable form.

Most of the above mechanisms are related to these goals: reducing or eliminating gameability, cartel pain points, tragedy of the commons, honesty assumptions, and strengthening the protocol even in special and highly adversarial situations. Although the network can eventually fall back to social recovery schemes, it will take a lot of work to ensure that the protocol can independently cope with the widest range of attack scenarios.

DAS in practice

Data availability sampling (DAS) is an elegant mathematical solution to the data availability problem, but when this architecture meets actual p2p networks, it is not so simple to instantiate it in an effective and robust way. In 2022, many teams have joined the ranks of solving this problem (both in academia and in real engineering), but a lot of work is needed to achieve an effective solution.

The problem is not difficult, but the design space is huge, and it is difficult to find a suitable balance between complexity and robustness. If we are willing to compromise decentralization and robustness, DAS can reuse existing network components, in which case DAS can work well. But considering that attacks on this network architecture may completely destroy the activity of the chain, it is not enough for the system to be designed to be effective, it needs to be as tough as the rest of the protocol.

EIP-4844 provides a basis for scaling data, and gives us a parameter (data gas limit) to adjust so that the DAS can handle more load. I recommend being very conservative in raising this parameter - e.g. setting the data gas limit well below the theoretical limit; running the DAS and full 4844 data downloads simultaneously for a while; maintaining high network redundancy, etc.

It's debatable what order DAS should be rolled out in compared to other competing upgrades - stateless being another obvious major contender. My gut feeling is to give DAS more time to mature and focus on non-scaling upgrades (stateless, security improvements, PBS, etc) for 12+ months after 4844, while gradually increasing the gas limit for data based on the initialization parameters after mainnet observation and integration of potentially more advanced gossip technologies like episub.

MEV, opening Pandora's box

One of the core themes of the past 24 months has become Maximum Extractable Value (MEV) - not only has an impact on the application layer, but also on the consensus layer and security of Layer 1. It's like opening Pandora's box, MEV has turned a lot of things we thought and hoped were simple into a more complex combination of incentives, attacks, and more. MEV has a centralizing effect on the validator set, making the rewards for monopolies much larger than before, directly incentivizing reorganization; and complicates the simple "user-validator model" by bringing in a large number of new PvP (player vs. player) players.

MEV has a significant impact on the "completion" of the Ethereum roadmap. The protocol cannot turn a blind eye to the MEV problem, otherwise the incentives it brings may cause network instability or even destroy the network. There are some solutions now, such as proposer/builder separation (PBS), MEV destruction, single slot finality (SSF), and anti-censorship in the context of PBS.

I am torn between advocating for rolling out MEV mitigations as quickly as possible (the threat is real) and taking more time to understand the problem and solutions and allow research to evolve. I worry that even though we have made leaps and bounds in our understanding of the scope of this problem, research is still in its infancy and what we are doing today will not be enough for Ethereum’s long-term needs.

While we take the time to research, the incentives inherent in the MEV space will lead to the building of solutions, networks, and products that tend to gain rapid adoption and have a broad impact. If we don't act quickly, third-party solutions and behaviors will become, define, or even erode key protocol specifications.

Complexity

Complexity is evil.

Ethereum is complex and will only get more complex, and I worry that this dynamic system is or will become too complex to use, understand, and secure.

Modularity is the best way to handle system complexity. We separate execution and consensus, and by using libp2p for the (consensus) networking layer and robust, audited/validated libraries for the cryptographic layer, a lot of the complexity is masked. This allows each part to be revised separately, and each relevant individual and team to specialize, and then aggregate them into a more complex whole.

Furthermore, we must not extend the core protocol beyond minimal sufficiency in any area. Moving beyond sufficiency would introduce unnecessary complexity to the protocol.

L1 Coordination Games

We have seen great growth in the coordination and organization efforts around the delivery of Ethereum L1. Client teams have matured, devops are supported with testnets, testing resources have increased, and a lot of the complexity has been split between the two layers.

L1 Development

Over the past 5 years, as we’ve made leaps and bounds in providing complex upgrade capabilities to the protocol, the number of teams and individuals involved has increased 5x — increasing both the resilience of the software and the sharing of ideas. No single team or organization has a decisive influence on the development and health of the protocol.

The diversity of expertise in the consensus client ecosystem, and now the diversity of client usage, is a crowning achievement. The execution client ecosystem has more asymmetry to begin with, and so will take more time to reach the same level of expertise and reliability, but I believe we will eventually transition from 80%+ dominance on geth mainnet usage in 2023. Realistically, the merger is an inflection point here.

excitation

At the height of the bull run, I and others were very concerned about incentives on Ethereum L1, compared to Ethereum applications and L2 layers and L1 competitors. In all of these alternatives, there are token incentives that have very high upside potential because there are new tokens with high liquidity.

During the bull run, valuations on liquidity products were insane. So we were facing real personnel changes at the time, and a more subtle and growing dissatisfaction with the huge disparity in rewards between contributors to key infrastructure and those building on top of it. I’m not sure if I mentioned it at the time, but in retrospect, I have a sour feeling about the staggering wealth accumulated by other layers built on the L1 protocol layer. As for other L1 public chains (alt-L1), some people claimed that misaligned incentives are fundamental in blockchain and that alt-L1s will always consume and squeeze established L1s. These statements also felt naive at the time, but the impact and negative sentiment are still increasing.

While I’m not sure Ethereum L1 work will have the asymmetric upside that startups may have, there are a lot of positive improvements in incentives for L1 development right now. The Protocol Guild’s year-long pilot raised over $12M for over 120 members from over 20 teams. Retroactive public goods fundraising experiments may prove to provide at least partial support for the software and security that applications and L2s rely on. There are other fundraising mechanisms like gitcoin and clr.fund that continue to innovate to support public goods for the ecosystem. In addition, I see a lot of promise in other fundraising experiments in the social layer, such as share contracts with optional contributions to staker transaction fees/MEV (e.g. 1% of staked transaction fees go to the client you use).

capture

There are now a lot of well-resourced and sophisticated entities that are interested not only in studying Ethereum’s success, but also in manipulating the Ethereum protocol. Today, such entities generally seem to have no malicious intent, but they do have a lot of influence on how and at what speed the protocol develops. As a result, L2, VC, exchanges, and dapps have entered the governance and development process of L1. We are in a unique period in which special interest groups such as these have formed on the Ethereum network and joined the conversation at the negotiating table. But many areas of the Ethereum protocol are almost not fully developed and ossified by any standard. This makes the risks in the next few years particularly high.

A few years ago, these interested and mature, well-resourced entities simply didn’t exist, or wouldn’t even consider “helping” and bothering to participate. It was mainly the Ethereum Foundation, client teams, and various independent people who were involved in advancing the core protocol. So, if a few years later (3? 5? 10?) various entities like this show up and help, the protocol should be more rigid than it is now, and there will be less risk for interested parties to help. In fact, it will be necessary for entities that rely on the health and stability of the Ethereum network to own key infrastructure in the long run. However, for now, ownership of infrastructure means having a say in a still nascent protocol. Eventually, the malleable components of the protocol will become smaller and less important: network connection protocols, light client communication, content delivery networks, and other elements around the core. At this unique point, although enough interest groups exist, the protocol can still be controlled by handling upgrades that are less rigid, messy, and human-involved.

Given the increasing interests of all parties, there is a desire to find a balance that allows non-client developers/non-researchers to have a say in governance. Trying to be a gatekeeper for everyone else is a difficult line to walk, and in fact, there is never a real door - as long as you show up and work in good faith, we will welcome you. However, those who show up have increasingly complex backgrounds and motivations. As conflicts of interest increase, this even extends to core maintainers - investors in applications/L2s, representatives of DAOs, advisors to venture capital firms, loyalty to enterprises, and various product offerings; not to mention various geopolitical relationships that begin to get involved as the world pays more and more attention to cryptocurrency.

I firmly believe that formalization of anything beyond "developers write code, users run software" is a long-term danger to Ethereum. I've seen community calls for structures similar to political parties or lobbying interest groups. The more structure given to the process, the easier it is to abuse (capture), and the less likely it is to be removed (which would hamper the process of protocol ossification).

Consensus is hard, but Ethereum performs remarkably well with the current minimal and chaotic governance process. It's not perfect. It can't and won't always choose the optimal path, but we're not just competing for optimality of features and delivery. We're competing for a protocol that is adequate, robust, and prevents capture.

I suggest that we spend the next 5 years adding as little additional governance structure as possible, and focus more on soft things like philosophical alignment, a strong culture of only doing mission-critical upgrades, and fostering two-way communication between the developer and extension communities in organic, less formal forums. That said, the governance process is a beast that can be managed but not fully controlled. Individuals can influence outcomes here through soft guidance and philosophical thinking, and can speak up if things are really going off the track, but Ethereum governance is an independent organism that is not currently controlled by any one party.

Clients - Buy or Build?

Even though the protocol development and upgrade process is tied to “developers write code, users run software” and new formal structures have not yet been added, developing and maintaining a client is still the entry point for anyone to enter the conversation and process.

As we saw in 2022, this can be done by "buying" (Arbitrum acquiring Prysm) or "building" (Paradigm building reth), both of which have their own difficulty, acceptance, and sustainability. For example, buying a client is direct but seems to sacrifice acceptance, and the sustainability of the impact is very fragile - clients are built by multiple individuals, and an acquired company may drastically change priorities and perspectives, causing individuals to leave or even defect. Building a client is more difficult, but more acceptable and sustainable. In 2023, both experiments are worth watching.

Ethereum development and governance relies on a security-first, consistent development culture. My assessment is that with the current layout of client teams, it is difficult for one or two unconventional client teams or individuals to change the protocol path or feature set in a way that we call "capture" because the entire team has the ability to evaluate the merits of proposals and can prioritize and lead the entire network forward with a general philosophical perspective. However, my concern is whether new members and individuals in other client teams will deviate from "consistency" in the next few years, or some existing client teams that are consistent with Ethereum will leave, or more client teams will be acquired or created by less consistent entities in the next few years.

The multi-client ethos and ecosystem is one of Ethereum’s superpowers, resulting in a more resilient, accepted, and secure network through diversity of software and perspectives. This is one way Ethereum has been able to gather so many advanced ideas, and I believe this is one of the main feedback loops that makes Ethereum so unstoppable. I do not recommend trying to set up barriers to these types of informal client teams. Instead, I recommend keeping the door open and welcoming more client teams that align with Ethereum’s cultural philosophies to join in the coming years. While we’re at it, we should also increase the geographic and jurisdictional diversity of Ethereum client teams. Right now, client teams are concentrated in certain regions around the world. Promoting the global distribution of client teams will make Ethereum more resilient while further strengthening it as a gathering system for advanced ideas.

Rigidity of protocol

Protocol ossification, which sometimes seems to me increasingly critical to the long-term health and success of Ethereum, is less and less evident to me, especially in the short term (5-10 years?). We have more than a lot to do before Ethereum is "done", but on the other hand, this "protocol eventually ossifies" culture doesn't seem to be mainstream in the Ethereum community, and at best has support among client teams.

Bitcoin has tainted the concept of rigidity. Rigid Bitcoin is seen by the Ethereum community as a useless protocol, championed by an intellectual cesspool of religious zealotry that cannot fathom any argument that does not praise Bitcoin’s perfection or instruct an alternative design. This raises the question — is this due to the rigid object itself (in this case Bitcoin itself) or is the act of rigidity itself having this negative effect? ​​My hunch is that the religious anti-intellectualism is due to Bitcoin’s inadequacy. The Bitcoin protocol is insufficient to satisfy their religious beliefs, and so the motivations of its supporters have been twisted to support rigidity.

I believe that a protocol that meets its mission (in our case, secure and scalable enough, with sufficient capacity) can have a coherent argument for its eventual ossification. Therefore, its community does not need to defend it like a religion. And even a rigid Ethereum protocol can support fluidity and scalability between different layers, which allows the community to remain intellectually engaged through exploration and optimization of more layers. However, some community members worry that without a continuously evolving protocol, Ethereum will lose its fascinating advanced ideas and many other things that make it so special and vibrant. And I think that as long as support for layer 2 is sufficient, we can also benefit from it - rigidify the secure base layer while supporting a vibrant ecosystem for intellectual exploration and optimization at higher layers, at least that's the dream.

A fool's mission?

But is rigidity a mission for fools? Can it be done in the context of Ethereum? And what does rigidity actually mean in a useful way? Is rigidity absolute? Or can we be “rigid” to get the most value? And finally, is the real intention of rigidity enough to protect us — can the ever-looming threat of rigidity keep progress on the critical path, ensure our focus is better (more aligned with Ethereum?), and strengthen the immune system that protects the network from special interests and capture?

I do start to worry that achieving ossification in the near term is unrealistic given the flurry of upgrades and security improvements on the critical path — sharding, PoS evolution (SSF, PBS, etc.), security improvements (SSLE, escrow gaming, proof of execution, etc.), statelessness, ZK-EVM, quantum-resistant cryptography, etc. Given the complexity of the tasks at hand and the speed of delivery, this is at least 5 years if not 10 years away, and I worry that cryptographic developments and emerging issues will require many more upgrades in the foreseeable future.

So if this is a perpetual process, at least the rate of change will slow down, right? A more useful definition of rigidity might be a trend rather than an end result — that is, it becomes increasingly difficult to modify the protocol, and requires more justification, effort, and time. The activation energy to implement changes will eventually exceed a critical threshold. In other words, it is not that the protocol will actually rigidify, but that the best thing we can do is to always be in dialogue with rigidity. If our spirit is to try to achieve rigidity or be in a more rigid state, rather than always expecting, desiring, or needing change, then the dialogue with rigidity will bias us towards change skepticism. In this way, enough skepticism may be enough to protect the protocol even in the process of (slow) evolution.

In some reflection this year, I’ve come to believe that this conversation with rigidity is both the only thing we have at our disposal and the only way to ensure that Ethereum doesn’t hit an unpredictable limit in a world where tomorrow is unpredictable. But I also believe that rigidists — those who prefer to be conservative, only modify certain components when absolutely necessary, and fear the impact of increasing complexity and change on other layers above the base layer — are currently too small in the core L1 process and the growing community. This camp and its spirit are not enough to become the necessary immune system for the protocol, and while there are many respected people in this camp speaking out to ensure some minimum thresholds, the opinions and virtues of a protocol that tends to be rigid must be clarified in the coming years to ensure that it does not lose balance at this critical moment of rapid change.

Consistency between L2 and L1

One of the great things about the L2-centric roadmap is that it allows L1 to hand over all design decisions (cross-domain communication, choice and optimization of VMs, parallelization, state management, etc.) to the “market”, rather than laboriously (and potentially suboptimally) selecting and implementing a particular solution. This conceptual market on top of Ethereum’s L1 layer can not only bring good solutions in a similar form, but also adapt to the needs of the world as they continue to grow and change over the next few decades. Need parallelization? Try a new L2, a better VM that supports the necessary formal verification emerges? L2 can adapt, or a new L2 will emerge. This anti-rigidity picture built on a relatively rigid base layer is the best of both worlds.

As for the health of this market, the Ethereum L2 ecosystem is doing relatively well, with a few different optimistic EVM routes running, a lot of different EVM-compatible/equivalent/various zk-rollups competing with each other, and some exotic, more experimental rollups on the fringe (Fuel, anyone?). Some think exploration is stuck in a quagmire - everything revolves around the EVM - and expect more dramatic changes away from L1 structures - Solana-rollup, hyper-parallel rollup, etc - but the market can adapt to the rise and fall of certain views for now. The EVM looks like a moat, and most people within the moat are building.

Since all eggs are now in the L2 basket, I worry about the consistency of L2 with L1, both in the short and long term. There are two main issues - (1) L2 is parasitic and will eventually fork off to become L1 itself, and (2) L2 becomes the standard bearer for Ethereum and where users interact, but does not uphold the values ​​of Ethereum - decentralization, censorship resistance, support for public goods, radical cooperation, etc.

The former is more of an existential question - is there really value in being anchored in the Ethereum security zone? This is essentially the thesis of the L2 roadmap - these scalable environments inherit Ethereum's security, native bridges are valuable to users, and therefore valuable to developers, businesses, and communities building and maintaining them. I agree with this thesis - bootstrapping sufficient cryptoeconomic security is hard, and in increasingly adversarial environments, most chains will inevitably fail to reach sufficient levels. Cryptoeconomic security is a finite resource that has the function of maintaining the economic needs of these systems. So while I do expect some L2s to "pull the rug out" from Ethereum and attempt to split off - some will succeed, some will fail - I don't think these will happen on a large scale, and a few L2s leaving does not break the cryptoeconomic security as a service thesis.

L2 will inevitably become the first touchpoint for most users. In most cases, they will interact within and bridge across L2s because these will be secure and inexpensive. As a result, the face of Ethereum becomes the L2s themselves. They may be secure, but are they decentralized, censorship-resistant, do they uphold the values ​​of Ethereum, and do they encourage the world to continue to reimagine itself? Currently, the answer to these questions is clearly not yes. VCs have their tendrils deep into L2s. Tokens are distributed everywhere to insiders, either very small shares or horribly large shares. And most governance models are oligarchic and can be upgraded at will without notice. Not to mention that most L2s sacrifice security in their security models to go to market in the hope of continuously iterating towards decentralization (e.g. no fraud proofs, single sequencer, unclear emergency exit mechanism, etc.).

There is an interesting balance here, L2 layers can and are expected to focus more on advertising and business development, competing with alt-L1s that are very aggressive in this regard. This allows Ethereum L1 to remain neutral, and the layers above it can experiment with many user acquisition and onboarding technologies. But L2 does not seem to maintain the brand, values ​​and soul of Ethereum by default.

Managing a healthy L2 ecosystem is paramount. This will require getting things right on many levels — researching and promoting secure structures, shining a light on L2 support efforts (showing what they are, not what they are portrayed to be), and addressing L2 governance risks, security tradeoffs, poor token distribution, value alignment, and other contingencies when possible. And we can’t just focus on the negative, but also celebrate the positive, secure, and aligned aspects. The Ethereum community today has tremendous power to define the development of the L2 movement for decades to come. We must ensure that L2 inherits not only the security of Ethereum, but also its acceptance.

Summarize

As I mentioned at the outset, this article is extremely short-sighted and subjective. There are many issues and even more successes that I either don't know about or I don't have the time to explore.

In short, Ethereum is stronger than ever. It’s been an amazing experience to see the community building core infrastructure, the community layering expansion, and participating in and watching the community on Ethereum. However, there are still big challenges; there are still huge risks.

Ethereum is running strong, let’s do our part to keep it going.