Rewriting EIP-1559: A Victory for Block Proponents and Builders?
Original author: James Prestwich
Original translation: Luffy, Foresight News
Before EIP-1559
Long ago, in the world of cryptography before EIP-1559, miners determined block size through a simple voting mechanism. With each block, miners could adjust the gas limit up or down by a maximum of 1/1024. As a result, the gas limit would move towards a weighted average preferred by the miners when producing blocks. This mechanism has existed since the birth of Ethereum.
With the introduction of EIP-1559, the gas limit has doubled, but the method of setting the limit has changed. From the time after EIP-1559 to the merge, miners voted on the gas target, and the gas limit was defined as twice the target. For example, if the pre-1559 limit was 10 million (mm), and the post-1559 target was 10 mm, then the post-1559 limit would be 20 mm. EIP-1559 provided effective block size increases but came with significant consequences (in the form of base fee adjustments). Miners under the 1559 system could vote on the gas target, which is not very different from the previous adjustment system of moving 1/1024.
Now, 1/1024 (0.09%) may seem small, but it means an increase of around 5% every 50 blocks, or around every 12.5 minutes, or a doubling every 3 hours. Imagine taking a short nap and waking up to find the gas limit has doubled (or halved)!
A Brief Review
EIP-1559 has 2 inputs and produces 2 outputs:
Block producers set the target Gas for each block by adjusting the target of each block by 1/1024. In this way, the Gas limit (twice the target) is indirectly set. Block producers set the base fee by combining the Gas target with the actual gas usage. When the actual usage > target, the base fee increases; when the actual usage < target, the base fee decreases. There is a complex elastic relationship here.
Please note that the target usage does not affect the actual usage. In fact, each block producer sets the target (through adjustment mechanism) and the actual usage (by including transactions when building blocks). In reality, each block producer controls two inputs of EIP-1559, and therefore can control the outputs (for specific blocks). The target in any given block is the result of the producer's latest move in the "setting Gas target game" of infinite iterations. Therefore, over time, the behavior of the average value will converge to the equilibrium value in this game.
EIP-1559 and Merge
Merge has two main impacts on the Gas limit mechanism:
Due to slightly shorter block time, the cycle for Gas limit to grow by 5% is reduced to 10 minutes. This means that approximately every 2.4 hours, the Gas limit has the possibility to double (or halve).
Miners are replaced by "stakers" or "validators" or other parties.
The first one further accelerates the already fast compounding process, but this may not be important.
I believe the second one is more important. But it is not isolated, but part of a larger system trend. We'll discuss this issue later.
What prevents the block target from increasing?
You may wonder why the block target does not increase. Wouldn't it make transactions cheaper for users and allow validators to earn more money? If these things sound good, why is the target 15 mm instead of 15, 000, 000 mmm? Why is the current balance a target of 15 mm (limit of 30 mm)?
The main reason is that it has always been set according to the suggestions of the core developers. It is widely believed that we set it at a level that we know users can handle within a reasonable time frame. Large high-gas blocks can DoS nodes and prevent confirmation. Validators (previously known as miners) are required to build blocks and validate blocks from others. If the blocks are too large to reliably build or validate within the 12-second block time, validators would lose money in case of confirmation failure. This would also cause network instability as nodes may not handle gigabyte-sized blocks well.
EIP-1559 mentions this in a relatively short security consideration:
The EIP increases the maximum block size and problems could arise if miners cannot process blocks fast enough, forcing them to mine empty blocks. Over time, the average block size should remain roughly the same as it would be without this EIP, so this is mainly a short-term capacity burst issue. One or a few clients may not handle short-term capacity bursts and errors well (e.g., out-of-memory conditions), and client implementations should be able to handle properly individual blocks of maximum capacity.
We tend to think that validators have sufficient profit incentives to keep block limits within a reasonable range, and considering client conditions, a target of 15mm seems reasonable.
Is EIP-1559 still secure with PBS?
MEV-PBS changes the game. With PBS, proposers no longer directly bear the costs of block construction. They outsource the whole process to builders. In fact, common MEV-PBS setups, such as Flashbots' Relay, hide the block from validators until it commits to propose it. As part of the building process, validators communicate their preferred gas limit to the relay.
Builders use specialized, optimized software (not off-the-shelf clients) to construct a block and then transmit that block (adjusted with the gas limit) to the relay. The relay ensures that the builder's block complies with the proposer's requirements.
This brings an interesting situation. Proposers are now prohibited from paying for the construction costs of blocks. Builders specifically bear these costs and need not be too concerned about larger gas limits. So in the MEV-PBS world, why should gas targets be kept lower?
Not secure yet
Proposers always have an incentive to infinitely increase the gas target to eliminate base fees but are limited by the cost and risk of doing so. On the other hand, builders want to confirm blocks and do not encourage building blocks that are too large to be validated within confirmable time periods. In other words, builder's profitability is walking on a tightrope. They need to carefully balance building larger actual blocks to extract more with the impact of block size on confirmability. If builders generate blocks that are too large to be validated by validators, they risk losing the entire block reward. This means builders have a strong incentive to keep the actual block size within a reasonable and verifiable size, regardless of the gas target.
On the other hand, the proposers do not need to consider such a balance. They can simply infinitely increase the Gas target for protocol execution and let the builders find a profitable balance point. This can eliminate the base fee while still limiting the actual block space.
By increasing the Gas target (and protocol-enforced restrictions) while keeping the actual block size within a normal range, the proposers not only can extract more MEV through abnormal block sizes, but they also eliminate the base fee while artificially limiting the supply. If the builders keep smaller blocks, then users will have to pay Gas fees, returning to a fee market before EIP-1559. In reality, they can "presume" to increase block space to patch EIP-1559 and then refuse anyone from using it.
EIP-1559 seems like an obvious victory for the proposers and builders. They can both extract more MEV for themselves in exceptional cases (by occasionally building huge blocks) and convert all base fees burned into tips paid through artificial supply limitations. Additionally, the cartel is incentive-compatible because neither party seems to profit from deviation.
This leads me to the central question of this article.
When will the proposers and builders raise the Gas target to 100 mm+?
Given the constraints of confirmability, builders should keep the actual block size near its current value regardless of the target size. Some reasonable small blocks can achieve maximum profit. However, for the proposers, some outrageously high Gas targets can maximize their profit. As the positions differ between the proposers and builders, they can collude to raise the Gas target while shrinking the actual scale to achieve maximum profit.
As long as the actual Gas in the block remains around 15 mm, wouldn't a 15000 mm Gas target be more incentive-compatible than a 15 mm Gas target?
So, whether it's the block builders or validators, pushing up the target Gas but keeping the actual generated blocks smaller will result in higher earnings.
