Defining modular blockchains

[alexbeckett]

Introduction

The narrative for modular blockchains is starting to gain usage to tout the superiority of next-generation blockchain infrastructure. Since Celestia positions itself as modular, and there is no concrete definition of a modular blockchain or its monolithic counterpart, it makes sense that the concept should be defined in a way that allows for appropriate analysis and distinctions to be made.

What is a modular blockchain?

A modular blockchain is a blockchain that outsources at least one of three components to an external independent chain while also conceding its ability to temporarily handle that component/s – the three components being execution, consensus, and data availability. Alternatively, monolithic blockchains are blockchains that handle all three components of the modular stack.

Rollups are modular because they only handle execution, outsourcing both consensus and data availability to Ethereum, for example, while conceding the temporary ability to handle those components themselves. Data availability layers, like Celestia, and settlement layers that only provide execution and consensus are also modular because they don’t handle all three components themselves.

Modularity implies flexibility. Therefore, modular chains can outsource components to separate independent chains or handle them locally – can’t handle more than two out of three components, otherwise it implies monolithic limitations. Flexibility also allows modular chains to be created, exchanged, or replaced independently within a modular stack - a core element of the modular design principle.

The modular spectrum

Since modular blockchains vary by architecture and design, it is necessary to define further terms to encompass this spectrum that doesn’t fit the strict definition of “modular” or “monolithic” as given above.

Proto modular will refer to any monolithic blockchain that outsources at least one component of the stack to another modular blockchain. For example, Ethereum utilizes rollups for execution while also retaining the ability to handle that component itself - under this definition, Ethereum would be proto modular. However, a monolithic blockchain that outsources a component of the stack to another monolithic blockchain will remain monolithic. Essentially, a proto modular blockchain increases its capacity for resource usage through an independent modular blockchain.

Pseudo modular will refer to monolithic blockchains that don’t outsource any components to modular blockchains but do have a design that splits the single network into multiple parts. This would encompass blockchains that have shards or sub-networks, like Polkadot and Avalanche. Essentially, a pseudo modular blockchain increases its capacity for resource usage through the utilization of other monolithic chains that are bound and dependent on the single network.

Degree of modularity

Conclusion

The definitions provided attempt to give clarity on the distinctions first between modular and monolithic, then between the varying degrees of modular. As modular blockchains progress and their architectures change over time, it may be necessary to update and expand the list of terms used to define such blockchains.

[polynya]

Interesting post! I’d consider sharded networks with shared security like Polkadot or NEAR with off-protocol execution shards proto-modular too. This construction similar to Ethereum + rollups, which would also be proto-modular, though with much less flexibility on the execution layer (e.g. Polkadot and NEAR shards can only do fraud proofs). Of course, volitions settling on Ethereum are already fully modular. I agree that Avalanche fits in the pseudo modular bracket given its basically just multiple monolithic chains where security [split validator set] or decentralization [increase validator system requirements by validating more subnets] is being compromised but I believe all multi-chain ecosystems eventually end up with shared security, i.e. proto-modular.

An interesting case is ZK-L1s like Mina. I have called them semi-modular in the past, because they can have their validity proofs proven on a different settlement layer for a trustless-ish bridge (which is was the Mina <> Ethereum bridge will be - Mina’s full state verified on Ethereum. As long as Mina is not compromised, it behaves exactly like a ZKR on Ethereum. From there, it’s straightforward for Mina to become an Ethereum validium, for example. The option to go full rollup is also much easier than other L1s, as Mir > Polygon Zero is proving.)

Then there’s Cardano, which has given up on trustless execution, and has a model where most of the execution needs to happen on trusted (in some way) relayers. Probably just bad-modular here. Eventually they’ll move all of that to rollups & state channels, at which point it would resemble a proto-modular architecture too.

Further modularizing execution, which we can have with recursive rollups, and finally, validity proven settlement layers. That, to me, is the endgame for the modular vision.

[alexbeckett]

I’d agree with placing shared security sharded blockchains under proto-modular, particularly because at a high level they are similar in design to Ethereum + rollups. To redefine proto modular under this premise it would refer to any monolithic blockchain that outsources at least one component of the stack to another modular blockchain, or in the case that it is sharded with shared security provided by a common shard that relieves all other shards of at least one component.

To also redefine pseudo modular, it would refer to monolithic blockchains that split the single network into multiple parts where there is no sense of shared security, and each individual part is itself monolithic.

For Mina, once the bridge is live and functioning then I would consider it to be proto modular because it would then have the ability to utilize another proto modular blockchain (Ethereum) for consensus.

As for the modular endgame, the design is similar to the initial conception of Celestia and Evmos – as demonstrated by Mustafa here. The idea is that Evmos would provide an EVM-based settlement layer for rollups on Celestia. The flow would start with a transaction on the rollup providing execution. It would then be batched and sent to the rollup providing settlement and, in the case for a zkRollup a validity proof would also be verified. Finally, the settlement rollup would batch transaction data and send it to Celestia for DA guarantees and consensus over ordering. There’s the recursive rollup-on-rollup design.

[polynya]

Another reason I’d differentiate Mina from proto-modular is because its execution layer is essentially a canonical ZKR, so you get the benefit of having a very low cost of verification.

In the above diagram, I’d like to see the Evmos settlement rollup be validity proven, and hence have a trust-minimized bridge between settlement layers, like the Mina <> Ethereum bridge. (The delays with fraud proofs makes this harder to accomplish.) That said, I think it makes much more sense to just post proofs to Ethereum directly, commit data to Celestia, and skip the Evmos settlement rollup entirely. A settlement layer is only as good as its liquidity, security, and network effects, and for rollups at scale the costs are so negligible (a 1,000 TPS ZKR only spends $0.00003 per transaction in gas fees to Ethereum) that the more optimized settlement isn’t top priority. Of course, Ethereum has its optimizations in the roadmap too - statelessness, separation of DA with data sharding will address most of it. Of course, settlement rollups still make sense for low value, low activity rollups, and the hope is over the years/decades they can build network effects and liquidity to attract higher value rollups.

[alexbeckett]

Good point. It would probably be necessary to define another category for recursive zero knowledge blockchains – particularly as the design starts to proliferate.

The settlement rollup is going to be single round fraud provable. To be validity provable is a non-trivial task but would be made slightly easier knowing that a settlement rollup is optimized for rollup-only transactions and contracts, not a multitude of applications.

Settlement rollups would also provide execution rollups with a minimal overhead solution to bootstrap their network. While at scale costs may be negligible, currently there is still significant overhead in launching and operating a rollup, or any general blockchain. Over time, there is no reason that a range of validity provable settlement rollups couldn’t achieve very high levels of liquidity and network effects.

[polynya]

Speaking of recursive rollups… Fractal Scaling: From L2 to L3. It’s layers all the way down | by StarkWare | StarkWare | Dec, 2021 | Medium