For the first decade or so, the blockchain industry only had monolithic blockchains. Early experiments like plasma, multi-chain and sharding attempted to break this up, but it’s only recently with rollups, validiums and data availability chains that it’s become clear that the era of the monolithic blockchain is ending. Yet, we are still tied to the monolithic perspective, using terminologies like L1 and L2 which are limited and do not capture the expressiveness of this revolutionary new design space. Here’s a thought experiment from a few months ago with more descriptive nomenclature.
I believe we a shift in perspective is required if we’re to understand the modular blockchain or blockchain legos era — not sure which is the better meme yet! What do you think? Do you have a better one?
But first, what’s a monolithic blockchain? Oversimplifying, a blockchain has three basic tasks — execution, security, and data availability. For the longest time, a blockchain had to do all of these themselves, which led to crippling inefficiencies, reflected in the blockchain trilemma. Bitcoin and Ethereum chose to be highly secure and decentralized, trading off scalability; while other chains made different trade-offs.
In the modular blockchain era, we are no longer bound to these and can eliminate these inefficiencies and the blockchain trilemma by age-old trick of specialization. Now, instead of just having one monolithic blockchains, we have three different types of chains or layers. Let’s analyze the lay of the land:
This is what users interact with — it’s where all the transactions happen. To the end user, this layer will be indistinguishable from using a monolithic blockchain, and will be directly comparable.
Execution-exclusive layers laser focused on processing transactions as fast as possible, while “outsourcing” the challenging work of security and data availability to other projects.
Rollups are the premiere execution layers, but we also have validiums and volitions. Currently, Arbitrum One has a significant time-to-market advantage, with Optimistic Ethereum following closely. However, both A1 and OE are at an early stage, with basic calldata compression optimizations like signature aggregation missing.
StarkNet has been on public testnet for 3 months now, and is getting closer to a MVP. I believe the last big hurdles are wide compatibility with web3 wallets, account contracts etc. StarkNet’s predecessor — StarkEx — already implements calldata compression techniques, and signature aggregation a default feature of zkRs so transaction fees will be significantly lower than ORs now — e.g. the average dYdX trade is settled for <$0.20. Even if Arbitrum One is able to implement these optimizations in a timely manner, zkRs fundamentally can compress calldata farther than ORs. StarkWare is confident that StarkNet v1 will release on mainnet with EVM-compatibility through the Warp transpiler by the end of the year, though conservatively it’s very likely to happen by early 2022 latest. Another advantage of StarkNet is that it’ll actually be a volition, not a rollup, but we’re awaiting more details on that.
zkSync 2.0 is the another promising EVM-compatible zkR. Oh, it’s actually not a rollup either — it’s a volition like StarkNet. We have more details about zkSync 2.0’s architecture, though. Arbitrum One, as a rollup does all execution itself, but relies on Ethereum for both security and data availability. However, Ethereum is expensive as a data availability layer. So, what a volition does is offer the user the choice between data availability on Ethereum (rollup mode) and data availability on a different chain (validium mode). In the case of zkSync 2.0, they will have their own data availability chain called zkPorter. The rollup mode remains the most secure option, while zkPorter mode will offer very low fees (think ~$0.0X) while still being more secure than sidechains and alternate monolithic chains. You can already get a preview of this from Immutable X. I expect zkSync 2.0 to release a public testnet this month, with a mainnet release in early 2022 — but do note delays are always on the cards for cutting-edge tech.
There are other players, of course, and I expect the execution layer space to be highly competitive over the next couple of years. Eventually, I expect most projects to be volitions, with security on the most secure chain through validity proofs, and data availability options available to users. It truly gets the best of all worlds. Finally, I’ll note that monolithic blockchains’ execution layers are highly uncompetitive — including Ethereum’s — so I expect 90+% of all blockchain activity to happen on rollups, validiums or volitions in the next couple of years.
Previously, I called this “Consensus”, but I think “Security” is better to not confuse with execution and DA layers which may or may not also have their own consensus mechanisms.
Of the three, this is by far the hardest layer. At this time, there are only two solutions that are adequately secure and decentralized or even attempting to be— Bitcoin and Ethereum. Most other chains didn’t see the blockchain legos tsunami approaching and made crippling sacrifices to security and /or decentralization to achieve higher scalability.
So, what will it take to compete with Ethereum as a security layer? A wide token distribution that can only be achieved from 6 years of intense activity and high-inflation proof-of-work. A consensus mechanism which can handle a million validators without resorting to in-protocol delegations. A culture of users and developers running full nodes, and focusing on solutions like statelessness to make this sustainable long term. At this time, to me the only realistic competitor to Ethereum is if Bitcoin adds functionality to verify zk-SN(T)ARKs, and even that seems highly unlikely they will. The other option is some revolutionary new tech.
Ethereum also has the best roadmap for data availability long term — both in terms of technology with KZG commitments and data availability sampling — but also sheer brute force, leveraging its industry-leading security chain for deploy a large number of data shard chains.
But Ethereum’s data availability layer is probably ~18 months away. In the short term, validiums and volitions can leverage Ethereum’s security, while commiting transaction data (in compressed form) to separate data availability layers. We have data availability chains like Polygon Avail, Celestia and zkPorter; and committees like StarkEx’s DAC, who will pick up the slack, and have every chance of building network effects. It should be noted that some of these chains are also security chains, but as covered above, I don’t think they’ll be competitive with Ethereum on that front.
As an outside candidate— we could also have (ex?)monolithic chains like Tezos and NEAR offering sharded data availability before Ethereum. Even though those chains are significantly inferior to Ethereum for security and decentralization; they can act as data availability chains.
Finally, it’s not just about data availability chains. We can have innovative data availability layers that guarantee validity and availability without needing consensus mechanisms. I don’t think anyone has solved this yet in a decentralized manner (you could argue StarkEx DAC has solved this in a semi-centralized manner), but if they do, it can potentially be more efficient than data availability chains. Even if it’s not a hard guarantee, the cost savings may be worth the risk to some users.
We’re entering a bold new era of blockchain legos, that bring orders of magnitude greater efficiencies to the industry. I hope this post will lay out the competitive landscape in the future. Monolithic blockchains are pretty much obsolete, they need to pivot to focusing on execution, security or data availability — it’s impossible to compete if you’re still trying to do it all. Projects that have picked their areas of focus — as listed above — will be the big winners in the next couple of years and are worth following & supporting. I expect a mad scramble into this space — particularly on the execution front — over the coming months and years as the exponential increase in efficiency of the modular model compared to monolithic becomes obvious to everyone.