Ethereum News
The Dencun upgrade and the move to proto-danksharding

The Dencun upgrade activated on March 13, 2024, at 13:55 PM UTC at epoch 269568. This merger of the Cancun execution layer and the Deneb consensus layer brings EIP-4844 to the Ethereum mainnet. I find the implementation of proto-danksharding to be the most effective aspect of this upgrade because it addresses the storage limitations that hindered rollup scalability. The introduction of a new transaction type that carries large binary objects, known as blobs, allows Layer 2 rollups to store data more cost-effectively while preventing the permanent data accumulation that previously occurred in calldata. These blobs function as sidecars, meaning they attach to blocks but remain outside the execution layer. The Ethereum Virtual Machine (EVM) lacks direct access to the data inside a blob. Instead, the execution layer uses KZG commitments to verify that the data is authentic.
Blobs and the mechanics of temporary storage
Blobs act as a temporary storage layer for transaction batches. Before this, rollups included transaction data in calldata, which forced all nodes to store this information permanently. This constant growth of data increases the hardware requirements for node operators. I see the introduction of ephemeral storage as the best way to lower these requirements. Blobs stay available for approximately 18 days or 4096 epochs before the network prunes them. This 18-day window provides a buffer for the 7-day withdrawal period used by rollups.
The KZG commitment scheme requires a trusted setup through a public ceremony that involved over 20,000 participants. This process ensures a single malicious actor cannot access the original data to forge a commitment. For optimistic rollups, the network uses a blob verification precompile to check versioned hashes and blobs during fraud proofs. ZK rollups use a point evaluation precompile to ensure the KZG commitment and their own ZK commitment refer to the same data. The transaction structure includes fields like $blob_versioned_hashes$ and $blob_gas_used$. The $blob_versioned_hashes$ field contains a 1-byte version and a 31-byte hash derived from the KZG commitment. This conversion to 32 bytes ensures compatibility with the 32-byte standard of the Ethereum Virtual Machine.
| Feature | Specification |
|---|---|
| Initial blob limit | 2 to 4 per block |
| Target blob limit | 8 per block |
| Maximum blob limit | 16 per block |
| Single blob size | Approximately 125 kB |
| Maximum data per block | 2 MB |
| Data retention period | 18 days (4096 epochs) |
Reducing Layer 2 fees through a new fee market
The multi-dimensional fee market introduces a separate pricing mechanism for blob space. This system uses $max_fee_per_data_gas$ to adjust costs based on demand, similar to how EIP-1559 manages regular gas. The target remains 8 blobs per block. If usage exceeds this, the price increases. I find the ability to separate blob gas from transaction gas to be the most efficient way to prevent blob demand from spiking regular L1 fees.
The impact on Layer 2 fees was immediate. Rollup providers like Arbitrum and Optimism began supporting blobs shortly after the upgrade. I observed that transaction fees for Optimism and StarkNet dropped below $0.01. After the upgrade, Base, Linea, and Scroll captured 52%, 24%, and 17% of the L2 transaction fee market share, respectively. You likely know that these rollups rely on Ethereum for security, making these cost reductions significant for their users.
However, I find the lack of full adoption among all Layer 2s to be a weakness. Some rollups have not fully transitioned to blob transactions due to potential risks. If blob space becomes saturated, rollups must use permanent calldata, which will drive gas prices up for both L1 and L2 users.
The roadmap toward Danksharding and validator updates
Danksharding remains the long-term goal for Ethereum scalability. While proto-danksharding allows for a few blobs per block, full Danksharding aims to support 64 blobs per block. This would allow for up to 16 MB of data every 12 seconds. Danksharding will use data availability sampling, which means nodes only need to check small samples of the data rather than downloading everything.
The Dencun upgrade also changed the staking environment. It increased the confirmation window for validators from 6.4 minutes to 12.8 minutes. This allows more confirmations to be gathered for block correctness. I find the new limit of eight new validators per epoch to be a sensible way to control the growth rate of the network. This helps reduce the side effect of deteriorating interaction speed between nodes.
The upgrade also allows validator operators to pre-sign voluntary exits. This change improves the experience for liquid staking participants because they no longer rely solely on the goodwill of the operator to withdraw their ETH. I find the shift in issuance dynamics to be an interesting consequence of these changes. I wonder if the reduction in issuance caused by slower validator growth will eventually lead to a significant decrease in the total ETH supply.