Discussion based on the subtraction theory

Two years ago, when L2 and zk were all the rage, we predicted that L1 would make a comeback.

Author: Kan Jia, PhD in Cryptography from the University of Liverpool, discoverer of the proof-of-work of economics, startup founder of storage public chain, serial entrepreneur

Two years ago, when L2 and zk were all the rage, we predicted that L1 would make a comeback. We were the first to propose the concept of the "Son of Heaven and the Feudal Lords", and we also witnessed some L2s trying to break free from ETH as the gas fee. Of course, this is all just gossip in the industry.

Recently, we haven't paid much attention to ETH, but the concept of "based rollup" has made us feel a new consensus on the direction of blockchain technology development. We see many researchers discussing what "based rollup" is? Is this a new form of L2? We can say with certainty that "based rollup" is a euphemistic denial of L2. We are pleased to see that the era of L1 is about to return. This is a new consensus reached by mainstream thinking on the development of blockchain technology.

First, let's review the timeline:

In the early days of blockchain, in addition to colored coin technology, there was a technology called counterparty that could issue other virtual currencies based on the Bitcoin mainnet.

In 23, the hottest thing was Glyphs, and you may still remember the Facet project that evolved from ETHS.

Also in 23, Monad became the leader in the track by parallel EVM, but in their documentation, they also proposed the concept of Delayed Execution.

On the 1st day of 24, we had a 180-degree turn in our understanding of Glyphs technology. From previously being bearish, we changed to believing that Glyphs technology is the future of blockchain technology, and in the same month we launched the Subtraction Theory to explain the inherent advantages of this technology from the consensus layer.
In March 24, "based rollup" was proposed on the Ethereum research forum.

In our view, "based rollup" is a denial of L2, rather than a new L2 technology. This proposal is even a reflection on the technical direction of Ethereum over the past 10 years. Before delving deeper, let's take a look at how Ethereum is doing now.

As a blockchain, Ethereum innovatively combined programming language with blockchain, as is well known, this is the essential difference between Ethereum and Bitcoin. Bitcoin technology has a transaction memory pool, and miners are responsible for taking the transactions with high fees from the pool, can sort them according to their own rules and calculate the root hash, and finally run the PoW consensus algorithm, which is equivalent to a vote, no matter how many miners participate in mining, only one miner's work is effective and gets the block reward.

The principle of Ethereum is similar, the difference is that the content of Bitcoin transactions is transfer, while the content of Ethereum transactions is actually a call to the code. So in addition to ordering and reaching consensus on the content of all transactions, Ethereum needs to complete another important calculation within the short block time of more than 10 seconds, which is the root of the global state after the transaction execution. These two key Merkle root data (there are also receipt trees) are fed into the consensus algorithm to generate a new block.

Here lies a problem, the calculation of the transaction Merkle tree itself is not a heavy workload, even when the transaction volume is very large, it can be calculated by multiple CPUs or clusters. But the traditional EVM can only execute transactions sequentially on a single-core CPU (parallel EVM was proposed to speed up execution), so the main factor limiting the TPS/capacity of a smart contract blockchain is not the block size, but the efficiency of EVM execution.

What is L2? Many people already have some concept of it, it works like L1, but not as expensive. Here we need to go deeper, L2 does not run a consensus algorithm. Everyone knows that the consensus algorithm is very expensive, in a PoW block, the cost of consensus is roughly equal to the power consumed by all miners in a single block time (to be more precise, there is also the cost of hardware).

Does PoS have no cost then? No, capital needs to earn interest, the ETH staked for staking needs to pay interest almost the same as a bank, so that the capital will choose to stay in ETH for staking. Then who pays this part of the cost? The gas used by users of the chain is the tax.

We've gone a bit far, L2 doesn't have its own consensus algorithm, it has to rely on L1 to guarantee its security, otherwise it can't be called a public chain, only a private chain. L2 doesn't run a consensus algorithm, but it still orders the transactions, this is the centralized sequencer, using a centralized sequencer to order and reach consensus on the content of transactions is very cheap, just choose a boss and let him sign.

The only problem is that the sequencer cannot go down, the data center cannot lose power. To make L2 a public chain, the ordered transactions and the global state information of L2 need to be stored on L1, this is the concept of DA (data availability), mainly referring to the original transaction information (this part does not need to be too rigorous, just understand it). So the L2 minions paid a portion of the protection fee to the L1 overlord, and soon they will grow into feudal lords, threatening the Son of Heaven, because the tax of L1 is much more expensive than L2, so users all go to L2.

What is the idea of "based rollup"? Many people say it is a new form of L2, but we directly say that "based rollup" denies L2. Let's look at an obvious problem that Ethereum has had over the past ten years: it has put the EVM inside the blockchain nodes.

As mentioned earlier, the block needs to order the transactions (user-issued signed messages) and run the consensus algorithm within the block interval. Ethereum has expanded this step, adding an incredibly expensive operation, executing the virtual machine. Of course, when Ethereum was only 7 TPS, we couldn't see the expensive nature of the virtual machine operation, but when we need 700, 7000, or even Visa and Alipay level 70,000 TPS, can you imagine that an ordinary PC can sequentially execute 70,000 EVM programs in one second?

As the TPS in the block gets higher, we can recognize that this design is fundamentally impossible to have a future in scalability. Ethereum's act of embedding the virtual machine into the blockchain is great, but it is a mistake in direction.

Based rollup proposes that sequencing is completed by L1, and it does not go into much detail on other parts. However, if sequencing is done by L1 and the execution of transactions is also done by L1, then Based rollup is essentially saying nothing and is the same as the existing Ethereum. So Based rollup tactfully assigns the work other than sequencing to other entities, whether they are called L2 or indexers, which are an execution layer independent of L1! In simple terms, this is kicking the EVM out of L1. We are very pleased to see the major shift in Ethereum's thinking, and next, we will use an even deeper subtraction theory to explain why this is the right thing to do.

What is the idea of the subtraction theory? When Minting was booming last year, everyone thought Minting was just a way to increase the income of BTC miners. When the Minting craze spread to chains with their own virtual machines, such as ETH, we mocked it as a form of self-castration, then cut it off and put on a fake XX. But then we understood the superiority of this approach from the perspective of consensus. On the 1st day of 2024, we began to think that Minting had accidentally hit the mark and brought us back to the right direction of blockchain technology.

As mentioned earlier, blockchain consensus first requires freely selecting some transactions from the memory pool and ordering them by gas. If it's EVM, you also need to execute these transactions in order to verify that all execution gas does not exceed the block limit, confirm that there are no wasted transactions, and finally complete the block production.

From BTC to ETH, blockchain designers have been adding features. From ETH to CFX, and then to Monad, they are still adding features. So, is there a way to simplify the blockchain even further like BTC? Is it possible to do less is more and create a more optimized blockchain through subtraction? Our subtraction theory proposes that the EVM execution and blockchain transaction verification part should be removed from the consensus steps.

Simplify the blockchain into a machine that reaches consensus on input transactions, bluntly remove the EVM or any other computationally intensive components from the blockchain nodes, and as long as users are willing to pay the gas fee, the blockchain nodes themselves will not do more verification work, just directly put the transactions into the block.

Does this subtraction theory cripple the blockchain's smart contract functionality? No. The subtraction theory merely removes the virtual machine from the blockchain nodes, but all virtual machine computations are outsourced to another node called an indexer (the concept of indexer already exists in Minting), so the indexer can have elasticity and ample time to execute transactions, rather than being required to execute thousands of transactions within the short few seconds between two block productions. Such indexers can run on a single machine or on a cluster, and can have parallel algorithm optimizations. And even if the indexer execution crashes, it will not affect the stability of the blockchain nodes.

Is the subtraction theory safe? Yes, it is safe. Because our modeling of the blockchain is based on the classic data structure state machine. Mathematics tells us that as long as we replay the transactions in order, we can always get a unique state (even if there are hardware failures or computational errors, we can use other means to identify them).

By reviewing the technical development trends of recent projects, we find that many industry pioneers and excellent researchers have gradually reached a consensus on the optimization path for the future of blockchain. We hope that through this, blockchain technology can achieve a breakthrough and reach the scale of Internet technology as soon as possible.

Disclaimer: As a blockchain information platform, the articles published on this site only represent the personal views of the authors and guests, and are not related to the position of Web3Caff. The information in the articles is for reference only and does not constitute any investment advice or offer, and please abide by the relevant laws and regulations of your country or region.