"Modularism, not maximalism" Interview with Nick White
Feb 28, 2023 - 6 min read
An interview with Nick White, COO of Celestia Labs, about the modular blockchain paradigm, solving the data availability problem, plans to bootstrap the Celestia ecosystem, and the shape of the future blockchain landscape.
Dominic Weibel (DW): Let’s start with a brief introduction of yourself, your blockchain background and the road that led up to Celestia.
Nick White (NW): It’s my pleasure to be here and I am excited to talk about modular blockchains and Celestia’s role in the broader movement. I’m Nick White, COO of Celestia Labs and I first learned about blockchains in 2014. When I learned about Bitcoin, it didn’t really stick. And then I learned about the theory in 2016 which kind of peaked my interest. Throughout 2017, I started going down the rabbit hole, educating myself, first through books, then blogs and Twitter, and finally research papers. By then I realized that the biggest problem facing blockchains was scalability, because until we had solved that problem, blockchains wouldn’t be able to reach their full potential and they wouldn’t have the impact that they could because it would be too expensive for people to use around the world.
In 2018, I set out to solve that problem. I co-founded a project called Harmony and we took an approach of using Proof-of-Stake and sharding, which were very new technologies back at that time to solve the scaling problem for blockchains. That went really well.
But then in 2020, I was reading this white paper called Lazy Ledger, and it just immediately struck me as a genius idea of essentially modularizing the blockchain stack. It separates execution from consensus and data availability (DA), which is typically all handled within one monolithic protocol, into separate layers. As a result, you get flexibility, scalability, and cross-chain communication. It was clear to me that this was the future of blockchain infrastructure and so I very quickly got in touch with the founders and ended up joining the project. Closely after that Lazy Ledger rebranded to Celestia.
(DW) At a very high level, what is Celestia? Please talk us through the differences and nuances between modular vs. monolithic architecture and its underlying first principles.
(NW) Celestia is trying to build next generation blockchain infrastructure that solves the problems that have held back blockchains from being as useful and as ubiquitous as they could be. There’s a lot of different problems.
First and foremost, it is scalability, which is that most blockchains up until modular blockchains have had a finite capacity. They’re like a laptop or a mainframe computer that has a certain amount of memory. It has a certain amount of capacity to run applications, and once you exceed that, it breaks down. We wanted to build a system that can actually expand the capacity as needed, as more people want to use it. That’s one of the goals that modular blockchains aim to achieve.
Another issue is that monolithic blockchains come preloaded with an operating system. Ethereum has the EVM (Ethereum Virtual Machine), Solana has the Solana VM. And those sort of operating systems dictate what applications you can run. It’s similar to having a Windows machine, a Macintosh machine or Android or iOS. The operating system that the blockchain runs, limits you on what kind of applications you can build. A modular blockchain doesn’t even come with an operating system installed on it. You as the developer decide what you want to run. It adds this whole other dimension of flexibility for developers that they don’t have in a monolithic framework.
Finally, modular blockchains provide a somewhat Holy Grail, which is shared security, where multiple blockchains share the same security framework by pooling their resources into one place and communicate with each other in a secure way. Whereas in the wild when blockchains connect with each other they end up having all kinds of vulnerabilities in crosschain communication. Therefore, modular chains solve those three problems, and I could explain a little bit more about what exactly is going on in terms of splitting up the different layers and why we call it modular.
Data Availability (DA)
Data availability refers to a guarantee of nodes being able to download all transaction data from block proposers that previously published the entire data from the block header (includes metadata) to the block body (includes processed transactions) of their respective blocks. This allows nodes to verify blocks by re-executing transactions. Since all full nodes host this data, it is limited and expensive.
(DW) We get to that “splitting up” and the building blocks now. To dive in a bit deeper, it would be great if you could guide us through the fundamental definitions in the modular stack, and the building blocks that are consensus, data availability (DA), execution and settlement.
(NW) It could be helpful to use an analogy here. If you think of a blockchain like being a soccer game, consensus would be making sure that everyone sees the same order of events like each player when they kick or pass the ball. It happens in the same order among all people watching in the stadium. If we don’t have an agreement upon what things happen and in what order, we’re not actually watching the same game, right? So, consensus is all about ordering events.
DA is basically that you want to make sure that everyone is able to view the game. If it turns out that someone covers your eyes, and you miss part of the play then you won’t be able to know what actually happened, and then what? Someone could have covered your eyes while committing a foul, but you didn’t see it, and everyone pretends nothing happened and they get away with breaking the rules. So, DA is basically making sure that everyone is actually able to see the game.
Execution is basically enforcing the rules of the game. When someone commits a foul, you know there’s a referee that can blow the whistle and confirm this person committed fraud that needs to be punished, or we restart the game, or we do a penalty kick.
And finally, settlement doesn’t quite fit in this analogy, but it is a way of resolving disputes. Let’s say there are two referees who disagree on the call. Then settlement provides a place where those two referees can duke it out and you can decide which one is making the correct call.
In a monolithic blockchain all these functions happen at the same time. The beauty of modular blockchains is that you can split those out into separate functions, and then those layers that specialize, like Celestia does in consensus and DA, get really optimized for that one purpose. For instance, you can build very specific execution layers that plug into it, and so you end up having a very specialized, scalable system.
To stimulate that soccer analogy, a monolithic blockchain is comparable to a single stadium that plays only one single game, such as soccer. It cannot expand the size of the stadium. If more people want to participate, it cannot contain more people because it is fixed in size. It can also only serve to play soccer as it only knows the rules to soccer. A modular blockchain is not a fixed stadium, it is more like an ever-expanding field. You can play soccer on that field, tennis, or basketball. You can play any kind of sports you want, and there isn’t any fixed capacity. As more people want to play, you can increase the size of the stadium. So, a modular blockchain is equivalent to the Olympic Games with tons of different events, enabling a huge number of players and diversity while the monolithic blockchain is just one single, limited stadium serving one game.
(DW) Very digestible analogy. Taking these core functionalities or modular building blocks, what kind of modular stack combinations make sense to you? Is the Holy Grail having consensus and DA together on the base layer like Celestia is aiming for? Or do you also envision modular blockchains that have other, more exotic combinations?
(NW) You can get very exotic as you can have each layer running separately. There are, however, some benefits of coupling consensus and DA specifically, such as interoperability. And in general, if you are running a rollup, which is an execution layer, you’re already relying on another external layer for DA, adding another sort of dependency. That also counts for consensus that can make the architecture slightly convoluted and overly complex. Whereas if you combine consensus and DA together, it makes the overall stack more efficient and reliable. Since part of interoperability is that you can rely on the ordering of the other chain you want to talk to, that it has the same ordering as your chain and the same DA, you also get better interoperability guarantees. It’s a nuanced concept. However, I fully expect there to be more exotic combinations. For example, one thing that we’ve built is the Quantum Gravity Bridge, which allows someone to build a rollup that uses Ethereum for consensus and settlement and Celestia for DA. The advantage in that variant is that Celestia is very cheap as there’s a lot of block space, but Ethereum has lots of assets and users. By settling to Ethereum you get access to that ecosystem by splitting the consensus and DA parts and coupling settlement and consensus together. It can get complicated, but there’s a limited number of different configurations at the end of the day.
(DW) I personally find it especially hard wrapping my head around the degree of freedom you get with a modular stack when it comes to the execution and settlement environment. Only looking at the rollup possibilities, offering sovereign rollups, settlement rollups and smart contract rollups, and the range of potential combinations is mind bending. It will be very exciting to see how this evolves and how the design space is utilized in the future.
Let’s focus on some properties that Celestia will offer once launched. Celestia is sometimes referred to being a hybrid approach that offers the best of both worlds between Cosmos and Ethereum by having sovereignty, shared security, that you already mentioned, and peer to peer bridging. Why are these aspects so important and where do you consider the trade-offs of going modular and having no settlement layer?
(NW) I really like that you brought up this topic of Ethereum and Cosmos because when I think of the evolution of blockchain architectures, Ethereum was a massive leap forward because it created this shared smart contract platform where developers could write a new application and deploy it very easily with- out having to create a new blockchain. And out-of-the-box it would have the shared security of the Ethereum network behind it. And that’s beautiful because it becomes very easy to build new applications and those applications can connect to each other, interoperate and be composable.
But then the downside of that is that everyone’s building on the same finite computer, right? And as more applications get built and more users come online, those applications are fighting for resources. And eventually the machine gets overloaded, and you end up having congestion and the fees spike. And so Ethereum has this problem of scalability, but also has a problem with sovereignty, a more nuanced concept. Blockchains are coordination mechanisms, and if you are a developer writing applications, you want to be able to control that application at the social level. If something bad happens, you want to be able to reallocate resources, undo a hack or just upgrade. Let’s say I want to make a change to the EVM. I want to be able to upgrade the EVM to better support my application. Unfortunately, that application is locked within the broader social consensus and sovereignty of Ethereum, so you don’t actually have application-level sovereignty. Ethereum was a massive step forward but had those two drawbacks.
Cosmos then came online with a different vision, which is sort of the Internet of blockchains. Being able to build sovereign blockchains according to the Cosmos vision was very important because people want to be able to customize their blockchain for their application. They want to be able to fork it, upgrade it and do what they want. The second pillar of the Cosmos vision was the relevance of scalability. By having every application run on its own blockchain, it is comparable to every application running on its own computer. Each computer has way more capacity than if you are trying to stuff all those applications on a single computer. But what they gave up in their design was that each application must be deployed on a new blockchain. Thus, you lose the ease of deployment that you have with Ethereum, where you just write a smart contract, press deploy, and it is done. What they also lost was this shared security component as all the applications that are built upon the Ethereum security layer do not have to bootstrap a new secure consensus network from scratch.
Those two networks had those two problems, and basically Celestia is like the marriage of those two visions where we can get all four of the above. Because Celestia is like Ethereum in that it provides this shared security layer. It makes it very easy to deploy your own blockchain and is similar to a consensus network that you can deploy your application logic to. Yet, it also gives you sovereignty because your execution layer only plugs into Celestia. Celestia has no governance control nor dictates what you do with your execution layer. Moreover, it does not suffer from scalability problems. As it uses data availability sampling (DAS), you can expand its capacity with the number of nodes in the network. So unlike Ethereum, it won’t run into congestion problems in the future and that’s why it is basically the best of both worlds. It takes the best of Ethereum and the best of Cosmos and brings them together into one ecosystem. And we call that the Internet of modular blockchains.
In the future Ethereum will pursue a modular blockchain development path. However, they started out as a monolithic blockchain with a EVM, a fully monolithic state machine.
The problem is this adds friction and is not fully modular. The benefit however is that Ethereum comes with a built-in settlement layer, meaning a place where rollups can post their proofs in order to bridge and resolve disputes, a useful tooling when building rollups.
Now, Celestia has explicitly decided not to do that, because we want to be as modular as possible. Therefore, if we enshrine a settlement layer, first of all, it’s not credibly neutral. We end up starting to try to compete with other settlement layers that might want to be built on top of Celestia. And second of all, we also compromise the modularity and create more overhead on people that are using the Celestia network.
Rather, what we want to encourage is that people build settlement layers on top of Celestia, and there are already several teams that are doing that. It does not make sense to choose one single settlement layer because there’s going to be lots of different ways of optimizing the settlement layer for different kinds of rollups or different kinds of use cases. It also seems premature to say that the EVM is the destined execution or settlement environment. So luckily, we are not locked into any of those decisions now. The downside is that there is not a native way to bridge the Celestia token up to the second layer. There are trusted ways to do so, but there is no trust minimized way to do that which you can do in the Ethereum ecosystem. However, I think it is a problem that we’re aware of and working on. I think there’s going to be good solutions that emerge, and we already have the beginnings of some, but it is a downside.
Data Availability Sampling (DAS)
Data availability sampling refers to a cryptographic method for verifying DA without downloading the entire block data. Light nodes utilize DAS by doing numerous rounds of random sampling small subsets of block data. The node’s confidence in available data grows with each round of data sampling until a predetermined threshold that indicates DA is reached.
DAS enables cheap hardware (such as light nodes) to take over more important tasks within network security and throughput, that were previously solely reserved for full nodes.
(DW) Very interesting. Adding as a side note as it matches context: I recently saw a talk from the founder of EigenLayer and he also came up with this aspect of having Ethereum modularizing its decentralized trust pool for example. And I guess that’s one problem or one challenge that Celestia is facing, starting a new protocol or chain that must bootstrap the network in the first place.
The modular blockchain paradigm is recently all over the place, and it seems to be the hot topic of ’23. As modular blockchains are getting traction, more projects enter the space. Could you shine some light on how Celestia differs to modular projects with similar approaches such as Polygon’s Avail, EigenDA or even rollup provided DA such as StarkEX, zkPorter or Arbitrum Nova and their design considerations?
(NW) Celestia was the first modular blockchain network and sort of the first data availability layer to be built. We were first to testnet, and I think likely the first to launch to maintain later this year. But there’s been a lot of other projects that have come online. First of all, Ethereum has adopted a modular sort of blockchain development roadmap, which is very exciting. And then other projects like Polygon have come online. And as you mentioned, zkPorter and even Arbitrum Nova, StarkWare and EigenDA. A lot of people are building their own DA solutions.
Compared to Ethereum, Celestia has no enshrined settlement, so we are just going pure DA, sort of maximally modular. And we think that is the right long-term architecture and design decision. With Polygon Avail, they’re choosing an architecture that’s based on KZG commitments, which are lot more computationally expensive, slower and have a lot of drawbacks that we think are, at least for now, not the right sort of commitment scheme to use. And the other problem I would say with Avail is that they’re not credibly neutral because they are also building their own rollups and their own ecosystem. And so Celestia is, on purpose, not delving into the rollup development side of things. Imagine if Ethereum was building their own Uniswap or their own applications and thereby competing with the developers on their chain. We think that is a) not credibly neutral and b) likely won’t attract the right kinds of developers.
And so, I think that’s one issue that Avail has, and I would say something similar in regard to the other rollups like Arbitrum or StarkWare that are building their own DA solutions. A lot of them are a) just not actually the fully fledged data availability sampling solutions and b) they’re isolating the ecosystems as they are building a rollup and their own DA solution. It’s like a monolithic ecosystem that behaves like a vertically integrated sort of siloed thing. Yet, the benefits of this modular blockchain stack are that you do not have to build everything on your own and end up sharing security and this ecosystem, having interoperability with different rollups. If Arbitrum, StarkEX, zkSync and all the Polygon rollups are running on their own DA layer, it defeats the original purpose of why we’re building the modular paradigm in the first place and that’s why we say modularism, not maximalism. We want to have this open ecosystem where everyone’s building in a shared way. A place where we collaborate rather than trying to carve out our own siloed piece of the pie and not share with other people.
DW: That’s a compelling perspective, especially considering the roadmap of Ethereum that is also heading into the modular direction. Hence, as Celestia will be maximally flexible, modular stacks that leverage Celestia’s DA can plug into Ethereum like Celestiums are aiming for.
As we move on, let’s briefly focus on the consensus engine of Celestia. With the DA layer of Celestia running on a PoS blockchain built with Cosmos SDK, are there any exciting new technical quirks when it comes to Celestia’s Core consensus mechanism? And will Celestia be plugged into Cosmos Hub?
NW: We are built on the Cosmos stack, so we use Tendermint and the Cosmos SDK, but we also have modified a lot of those things to add in the capabilities for data availability sampling and to support the core use case which is being a DA layer. Therefore, we have made lots of different changes, but at our core, a lot of it is based on Tendermint and the Cosmos SDK. The good thing about that is that we do get IBC support out-of-the-box. Also, we get support for all different kinds of wal- lets and block explorers out-of-the-box, which is advantageous.
One of the challenges for doing data availability sampling, is that there is a lot of networking complexity because most blockchains do not really have this topography where all these light nodes are requesting random samples of data. And so that was among the first modifications made, building a component of our node software that handles all that complexity, both in networking and the different kinds of requests that you get. It’s just a totally different functionality that the blockchain needs to have when you want to support something like data availability sampling and to be able to reconstruct blocks. So that’s one of the big things. The other big thing is within Tendermint.
The way that the blocks are encoded to support data availability sampling is novel. We had to encode the blocks using this thing called 2D Reed-Solomon erasure coding and so we had to really modify Tendermint to be able to encode that block box in a new way. We had to create new kinds of transaction formats that are all about paying to include just a blob of data rather than paying for a typical transaction and pack the block in an effective way. Because we want to have very large blocks in the future, we had to think about how to gossip the block data and the mempool in an efficient way, so that we can build blocks very quickly while essentially minimizing bandwidth overhead. This is a high-level summary of some of the bigger changes that we’ve made and some engineering innovations that we have done under the hood.
DW: Speaking about development, I just recently got an e-mail for the upcoming incentivized testnet and I wondered if you are happy with the current progress especially within the Mamaki testnet and what the biggest obstacles have been? As we inch closer to mainnet launch in 2023, Celestia will start its network from scratch. How is Celestia tackling this problem of new protocols having to bootstrap its decentralized trust network? Are there any mechanisms and incentives planned to catalyze network distribution and get more validators on board?
(NW) First of all, we’re very happy with the progress. To build a live network that supports data availability sampling is a major engineering achievement, not just for Celestia, but for the blockchain industry overall. We’re doing pioneering work, and that is not easy. We are on track for the mainnet launch this year, which is exciting.
In terms of bootstrapping the network, we do have an incentivized testnet coming up later this quarter. That’s going to be one of the primary ways that we start out this validator community, get a lot of people on board running nodes and educate the community more broadly about all the different node types. This will form the seed for the network launch later this year. A lot of those same validators would be the core original validators of the network. Decentralization and bootstrapping this new consensus network are aspects that we take very seriously. Over time, we want to increase the number of nodes in the network.
However, one of the beautiful things about Celestia and the way it is designed is you do not have to trust the validators. You do not have to worry about them being honest or not, because if you are running a light node, which anyone will be able to do on their smartphone, you can self-verify that the validators are not cheating you or doing anything wrong. Therefore, you don’t have to trust based on assumptions derived from the validator set. You can just directly verify the network yourself. Thus, we are not as dependent on having this sort of decentralized network of validators as it is still critical but not security critical because the validators are very limited in what they can do. That underpins the original vision of blockchains as you do not have to delegate your trust to another source. The whole point is peer-to-peer by which each peer in the network can verify the chain directly themselves. If I start delegating my trust to another group, even a highly decentralized group of validators, it starts to look more like a consortium, more like web2 and not like web3. Therefore, things like Solana, at least in their current form, don’t really achieve what blockchains are meant to do from our perspective, because even if there’s a bunch of different nodes, no user has any ability to verify what is actually happening on the chain. You’re just trusting this group of validators who you do not know that they do the right thing.
DW: Looking forward to a future, where everybody can participate in consensus by running light nodes from private phones.
(NW) Moving into 2023, apart from the Celestia mainnet launch of course, what are you most excited about and how do you envision the crypto landscape to shape in the upcoming years? Will we move away from yet another monolithic Layer1 only offering incremental improvements to a modular blockchain world? What role will Ethereum play in that future
and what role will other (multi-)monolithic Layer1s play? Is the future both modular and multichain as we will see with Celestia that is plugged into Cosmos? Or will application specific Layer3s even render appchains and multichains obsolete?
NW: Answering the broader question, it’s no surprise that I’m very excited about modular blockchains because I think they are going to solve the core problems of blockchain infrastructure. To date, everyone has been trying to solve that problem. That’s kind of why there’s been so many Layer1s, right? While everybody thinks to have found the solution, we are trapped in this perpetual cycle of a new Layer1 that has some new cool fancy thing without any actual fundamental innovation. So modular blockchains is the first time that people have gone back to the drawing board and realize that there is a fundamentally better solution, a new foundation to build upon and we can innovate on top of this. In a sequential way, moving forward, rather than just a cycle where we are going to loop, which is where we were stuck with this monolithic blockchain world. So that’s obviously the foundation of what I’m most excited about now.
Within the modular blockchain stack, there’s a lot of things to talk about. For instance, one is zero knowledge (zk) technology. There’s been a huge amount of innovation in the zk rollup space and there are a lot of teams who are building the foundation to make zk rollups a lot more practical. What is special about zk rollups is that they have better interoperability and finality properties. They also have better verification that eliminate complexities of fraud proofs that are utilized by optimistic rollups.
I’m also excited about people building rollup SDKs. Celestia provides the base level infrastructure that you can plug into but there’s so many other pieces that need to come into play to make it easy for anyone to deploy a new rollup. Developer tooling like SDKs, is a big part of that. Another part of this are decentralized sequencer networks.
Even though a rollup can outsource its decentralization to a network like Celestia, you still must have someone to run the rollup execution layer nodes. It can be one node, or it can be a set of nodes, but ideally that could be a service that you can also tap into. Hence, there are several people working on making that itself into a service, essentially sequencing as a service. With the possible combinations of these things together, the rate of innovation will absolutely explode, and I think that might only be a couple of years away.
That leads me to another component, which is people building new types of execution environments. There are teams that build a new gaming execution environment. A lot of the people who are building games now are building on the EVM or the Solana VM without realizing the constraints that this imposes. I cannot share too much because they are still in stealth. But if you start over, you can create an execution environment that has pre-built-in a lot of the primitives that you would want if you are building gaming applications. They want to basically build the operating system for blockchain gaming. Overall, there is just a lot of innovation that still has not really been tapped into from that perspective.
It is amazing to see the Cambrian explosion of different ways that people are interpreting the modular blockchain thesis and expanding on it on all these different dimensions.
DW: Very exciting. There seems to be a huge momentum ahead as we are almost able to provide the infrastructure for global adoption. Thank you very much for your precious time, Nick. We wish the team and you the best of success with the mainnet launch in 2023, that so far promises to be explosive.