Partnered With:

Quick review of Starknet upgrade: What are the improvements in Starknet v0.12.0 version.

Read this article in 6 Minutes

Direct sensory experience, transaction speed has been reduced from 10-20 minutes to a few seconds, and Starknet finally pays attention to user experience.

Source: Haotian (Twitter: @tmel0211), blockchain security practitioner

Spread the word, @Starknet v0.12.0 has successfully killed the pending upgrade! The transaction speed has been reduced from 10-20 minutes to a few seconds, providing a direct sensory experience. Starknet has finally paid attention to user experience. So, from a technical perspective, how should we view Starknet's upgrade? How does it compare to zkSync and how does it work? What impact will this have on Starknet's future ecosystem development? Let's discuss.

论 zk 技术 starknet 优于 zkSync，但 zkSync 的主网交互地址、生态繁荣度、TVL 等 Metrics 数据都远超 Starknet。why？ The reason is that Starknet has poor user experience in terms of account abstraction and Pending. This upgrade optimizes the cario contract algorithm, rewrites the Sequencer, increases TPS, and crucially shortens the time for transaction "success" by postposing the accepted on L1 time.

That is to say, previously the user had to wait for the txs to be verified successfully on L1 before ending the pending state, which took about 20 minutes. Now, as long as the Stark proof is generated on L2, the txs is assumed to be successful, which takes only a few seconds.

However, the corresponding Accepted on L1 time has been increased to more than 5 hours (as shown below), mainly to ensure the network security of the layering deterministic mechanism and avoid the probability of L1 reorganization.

The original single-layer confirmation mechanism, although the pending time is long, is only a sensory problem. Users can still send other transactions while there are pending transactions. (The key to adding or not adding a plus sign is not here.) It's just that users are used to waiting for a transaction to complete before starting a new one. But eliminating pending transactions will drive high-frequency transactions. To ensure sufficient security, the duration of L1 confirmation has been greatly increased, but users are unaware of it.

How it works? zkSync adopts a two-phase commit verification mechanism. After the Snark proof is generated and verified, it will interact with L1 twice. In contrast, starknet is a single-phase verification mechanism that only submits the final state to L1, rather than the initial submission state. How to ensure consistency before and after state changes? This is actually a point where Stark proof is stronger than Snark proof. Stark proof can deduce the self-verification of the submission state based on the final state.

On the one hand, Starknet incurs greater costs in terms of algorithms and computing resources; on the other hand, it will stack more transactions on its cario contract, which poses a great challenge to its zk circuit algorithm. The more transactions there are, the greater the pressure on state deduction. Once there are individual errors, network congestion may occur. This also explains why the time for Accepted on L1 exceeds 5 hours. The main purpose is to buy more fault-tolerant time for its zk algorithm.

Overall, the upgrade strategy of Starknet is significant, as it will make Starknet more user-friendly and directly improve its data, changing the competition landscape of various L2 public chains. However, you must know that the cost of simplifying the front-end experience is definitely an increase in back-end technology, and the network stability challenges that Starknet will face in the future will not be small. Nevertheless, by filling the gap in the market and user-friendly ecosystem, the future of friendly Starknet is promising!