DBL = Disaggregated Blockchain Layers
Not only is Celestia a DBL, but Danksharding is also a DBL, I believe that we can say ‘DBL has become a mainstream blockchain paradigm’
Disaggregated Blockchain Layers (thanks for @epolynya, DBL makes more sense than ‘modular blockchain’) means separating execution, consensus, and data availability into multiple specialized layers. (Current mainstream blockchains do these in one layer)
If we don't want to trade off on the impossible triangle and still want to scale current mainstream blockchains (called monolithic blockchain), we need better hardware support. Unfortunately, hardware (CPU, storage, bandwidth) can’t (脚标1).
DBL, a new paradigm, can scale without/with less trade-off. The logic is the operator of each individual layer has less load to deal with when compares with the operator of monolithic blockchain, then more hardware can be a node, a more efficiency node.
Today, with the likes of zkVM, uVM, we have already witnessed L2 on Ethereum can break free from EVM!
While Celestia is a little wilder: Nodes are free to hard/soft fork by upgrading their software and choosing to make sense of the underlying data in a different way.
Ethereums’ consensus concept has been further refined by Celestia: Txs Sequence Consensus Then we need Global State Consensus to determine the network state after executing valid transactions.
1. ETH 1.0
The representative of current mainstream blockchains (same for Solana, BSC, Tendermint). Just do everything on Ethereum. While this gives you maximum security and simplicity, it comes with limited scalability and high costs (current cost of data availability are almost beyond processing capacity).
2. Current Rollup
The original breakthrough in modular scaling (Nervos is earlier than this, but not widely accepted until now). Create a separate execution layer that anchors back to Ethereum for Consensus.
The benefit of doing this is reducing the cost of storing transactions by spreading fixed costs and some compression techniques.
A modification to rollups that moves DA off Ethereum.
This increases scalability (Blocks don’t need to include DA anymore), at the expense of security. There are different flavors (Data Availability Committees, Guardians, etc), with different trade-offs, but all lower security than a rollup.
As the latest design of Ethereum in ETH 2.0 era, Danksharding goes further & better in ‘DBL’ direction than current rollup & validum: split DA as an individual layer (more efficient than rollup) but still on chain (higher security than validum).
The benefit of doing this is maintaining validator reliability in the case of large blocks by reducing the workload of validators (verify only a small portion of DA instead of all), which leads allow a higher transaction per seconds with even lower cost to validate all transaction.
1. ‘Sovereign Chain’
The originally proposed model with Celestia is with a execution layer and final network state on top. In this design, the chain operator publishes transactions (can also publish roll-up) to Celestia. And Celestia stores these transactions with a consensus on the right sequence.
Considering that the global state of the network is the basis for external interaction, and that the global state in this model is the responsibility of the chain itself, we call this model ‘Sovereign Chain’
You should notice that the global state is not secure by Celestia, and when the chain operator does evil, the network is compromised and Celestia can’t do anything. E.g. chain operators only include parts of the transactions published to Celestia. Although those ‘Sovereign Chain’ can implement fault-proof & validity proof to avoid chain operator’s maliciously, but good design still relies heavily on good executors to realize its value, and that's where the value of Ethereum as the most decentralized network comes in.
I really love this design when compared with other Celestia-Centric designs: Nodes are free to upgrade their software and choose to make sense of the underlying data in a different way. For example, if a rollup community is contentiously debating over a change of block size or finality determination, opposing camps can update their software to follow different validity rules. But still, get the right & same result!
More importantly, with this design, the chain’s community has sovereignty and doesn’t derive its authority from any other state machine (as a comparison, the community behind smart contract or rollup on Ethereum is bound by the social consensus of the Ethereum community)
2. ‘Execution Rollup’ with common state consensus layer
There is a common global state consensus layer on top of Celestia, that developers can build their execution rollups that are anchored to as part of a single trust-minimized ‘cluster’.
It’s a little bit like Rollup & Ethereum paradigm: Rollup (Execution Rollup) anchors back to Ethereum (Cevmos/dYmension) for Txs Sequence& Global State Consensus (Global State Consensus, while Txs Sequence consensus left to Celestia) .
This design further reduces the cost of running a chain: no need to prepare node for produce global state when compared with ‘Sovereign Chain’, no need to pay high cost of DA when compared with Ethereum Rollup.
Another important feature is this design enables trust-minimized bridging between Rollups on the same global state consensus layer, and it will be brand new cross-rollup paradigm from any case we can see today, so remains to be seen if it can gain adoption.
The latest design is to use Celestia for Data Availability and Ethereum for Txs Sequence Consensus and Global State Consensus. This is the safest DA option when compared with Validium and Data Availability Committees (DAC), but much cheaper and fixed regardless the number of transactions (only verify signatures from DA nodes instead of specific transaction data) when compared with Rollup.
You can find more differences between DA options here
You might also notice that Danksharding (same for other Ethereum-centric design) let Ethereum beacon chain(or current Ethereum) to handle the Global State Consensus, while Celestia-centric design won’t cover the same part at all. The following could be some reasons:
As a DBL pioneer even older than Ethereum Rollup, Nervos Network published this ‘The Nervos Network has a layered architecture, with a single Layer 1 blockchain, the Nervos Common Knowledge Base (CKB), and many Layer 2 systems. Layer 1 is focused on providing security and decentralization, while developers can construct layer 2 systems (focused on functionality and performance) that are securely rooted in and interoperable with the Layer 1 CKB.’ in 2018.
Although the architecture is similar to current rollup design, but Nervos CKB adopted with UTXO model instead of account based model, which bring many difficulties to deliver a developer friendly smart contract framework. Fortunately, EVM compatible layer 2 is now available.
Looking forward to seeing more different possibilities of UTXO which can’t be achieved from account-based: Fuel has already shown out some possibilities.
P.S. DA here is a little bit different: Arweave not only stores transactions but also stores the code of every application built on it.
This design leverages Arweave as a DA layer, which leads to a precious feature: DA on a permanent hard drive = permanent DA!
But Arweave, as a storage network, can only ensure the storage and reading of data, but not to filter and sort specific data. And this means the blockchain needs to do Txs Sequence Consensus, Global State Consensus and Execution all by itself, which means the security of the network heavily relies on chain operators and Arweave can do little on it.
Arweave foundation offers SmartWeave, a smart contract framework for developers to build applications in smart contract level.
Ever.Finance is a more usable and DDOS-resistant smart contract platform when compared with SmartWeave.
Developers can also build some new SmartWeave similar chains with new optimization and different trade-off.
Polygon named it Avail, a Celestia-similar layer on Polygon, while Polygon performs as the Global State Consensus layer. Then developers can build their own ‘execution rollup’ on it.
Those innovations have mostly happened in the execution layer, which means a great execution layer containing these innovations is regrettable in DBL paradigm.
Considering many innovations include its own special design on Global State Consensus, Celestia-centric mighe be easier to adopt innovation than Danksharding.