Vitalik: EIP 1559 will alleviate the three major problems currently facing Ethereum, and the new solution will be more easily accepted by miners

Note: EIP 1559 is considered to be the next major surgery for Ethereum. In this regard, Ethereum co-founder Vitalik Buterin compiled a FAQ, which first introduced the three major problems with Ethereum's current transaction fee mechanism and how the EIP 1559 proposal would alleviate these problems. In addition, he also explained the limit value setting of EIP 1559 and proposed a modified version of the plan to address possible miner objections.

1. What is EIP 1559?

EIP 1559 is a proposal to change the Ethereum fee market. The main changes are as follows:

1. The current 10 million gas limit will be replaced by two values, one is the "long-term average target" (10 million), and the other is the "hard cap per block" (20 million);

2. Transactions require a base fee (BASEFEE, which will be burned), which is adjusted by block. The goal is to determine a value so that the average block gas usage remains around 10 million;

Essentially, our goal is not to have all short-term fluctuations in block space demand manifest themselves as fluctuations in transaction fees, but rather to have some of the fluctuations translate into fluctuations in block size.

2. Why is EIP 1559 beneficial to Ethereum?

(The answer is copied from an old post)

Currently, there are three major problems with the transaction fee market:

1. Mismatch between volatility of transaction fee levels and social costs of transactions: Mature public chains have sufficient usage, so blocks are full, and transaction fees on these blockchains are highly volatile. On the Ethereum platform, the lowest fee is usually around 2 gwei (note: 10^9 gwei = 1 ETH), but sometimes it can be as high as 20-50 gwei, and even once exceeded 200 gwei: https://etherscan.io/chart/gasprice. Obviously, this will cause many inefficiencies.

2. Inefficiencies of first-price auctions: See https://ethresear.ch/t/first-and-second-price-auctions-and-improved-transaction-fee-markets/2410 for a detailed description. In short, under the current method, when a transaction sender publishes a transaction with a fee, miners will select the transaction that pays the highest fee, and then everyone bids for it. This method is well known in the relevant design literature to be inefficient, so complex fee estimation algorithms are required, but these algorithms usually do not work well and often lead to overpayment. See also https://blog.bitgo.com/the-challenges-of-bitcoin-transaction-fee-estimation-e47a64a61c72 for a description of the challenges involved in the current fee estimation method by a Bitcoin core developer.

3. Instability issues of blockchains without block rewards: In the long run, blockchains with a limited total amount (including Bitcoin and Zcash) intend to reward miners entirely through transaction fees. However, existing known results show that this may lead to many instabilities, such as incentivizing miners to mine "sister blocks" of transaction fees, opening up more powerful selfish mining attack vectors, etc., and there are currently no good mitigation measures.

EIP 1559 has the following advantages:

1. It mitigates the economic inefficiencies caused by the social cost mismatch caused by fee fluctuations. There is a fairly subtle economic argument here, see pages 16-20 of the paper mentioned in https://ethresear.ch/t/draft-position-paper-on-resource-pricing/2838 (although I recommend reading the whole paper) for details. Intuitively, this fee adjustment mechanism looks like a fixed fee in the short term, while in the long term it represents a cap. The argument made in Martin Weitzman's 1974 paper suggests that in the context of essentially coexisting public chains, a fixed fee may be better than a cap.

2. It replaces auctions with a fixed-price sale (except in the short term when blocks fill completely until fees catch up), eliminating the inefficiencies of the first-price auction mechanism and making fee estimation very simple: calculate the fee f for the next block if you can afford it, otherwise don't.

3. It creates a mechanism similar to a perpetual block reward, alleviating many of the instability issues of fee-only blockchains (blockchains without new block rewards) without actually issuing them in perpetuity.

3. Is the current fee market really that inefficient?

Yes, in regular blocks the difference between the average gas price and the 10th percentile gas price is about 3x the median and 5-8x the mean. People are overpaying unnecessarily.

Everyone who doesn't overpay suffers a delay of 1-2 minutes, or even longer, that doesn't actually benefit anyone. Whether a given unit of load arrives on-chain at time N or at time N+60, the total load on the chain is the same. At least under normal circumstances, there is no actual social benefit for participants to "express low time preference" in the fee market mechanism; it's pure deadweight loss. We'd all be better off if more transactions were included immediately, which is what EIP 1559 allows.

4. Will EIP 1559 put too much pressure on nodes and miners during periods of high market demand?

Even in the short term, EIP 1559 will at most increase the block size by a factor of 2. Each "full block" (i.e. a block with 2x TARGET gas) will increase the BASEFEE by 1.125, so a series of constant full blocks will increase the gas price by a factor of 10 every 20 blocks (~4.3 minutes on average). Therefore, this heavy chain load will not actually last more than ~5 minutes.

Note that currently due to differences in block production speed, a period of double load lasting 5 minutes occurs randomly approximately every ~63888 blocks (about 10 days). Therefore, the introduction of EIP 1559 will not introduce any unprecedented load on the system.

Furthermore, the gas limit is just 10 million and not higher, which is largely justified not because of hard network limits (uncle rates are near all-time lows) but out of a concern for the long-term nature of:

1. Centralization risk from higher uncle rates: If the uncle rate spikes to 20%, it will greatly favor large, well-connected mining pools;

2. Limit state size;

3. Difficult to sync after a short period of offline time;

In all three cases, what matters is not the upper capacity limit during a very short window, but the long-term average capacity. An uncle rate of 2% in odd hours and 18% in even hours would have the same effect on all three cases above, since the uncle rate is always 10%. Since EIP 1559 still caps long-term gas usage to an average of about 10 million per block, it does not affect the long-term average.

5. What will high usage scenarios look like with EIP 1559 compared to the current situation?

Consider a “mathematically ideal peak” (e.g. this could happen in real life, like a sudden market event leading to a lot of arbitrage opportunities on a DEX, then there is a lot of bidding activity on liquidated CDPs, etc.), then there are N*10 million gas worth of transactions, each of which has a very high gas price, and all of them are broadcast.

Under the current fee mechanism, this would lead to the following:

1. The next N blocks are dedicated to filling in new priority (spiky) transactions;

2. After that, all other transactions, as well as transactions that people sent after the peak, will be implemented in order of gas price from high to low;

3. An ordinary user needs to wait for more than N blocks;

Now, let’s consider the case of using EIP 1559:

1. The next N/2 blocks are dedicated to filling up with new spiky transactions, and these blocks have twice the gas of normal blocks;

2. If all other transactions were sent with a gas price cap equal to the old gas price, then the next N/2 blocks would be empty and everything would return to normal. But in reality, higher priority transactions will set a higher gas price cap and be included in the block first, followed by other transactions;

In the former case, ordinary users have to wait for more than N blocks, while in the latter case, ordinary users have to wait for N/2 blocks.

Therefore, even including the “recovery period” after the peak, during which the block capacity will be smaller than normal, most transactions will be included in the blockchain more quickly.

Below is a very crude simulation (lots of weird assumptions here, but it's hard to model a complete system with supply and demand curves and waiting times), the spreadsheet source is here.

6. What can EIP 1559 do in the event of larger and longer-lasting peaks (for example, peaks within a day)?

There isn't much that can be done. The BASEFEE will go up, and there will be a short period of time where some transactions will come in faster, but after that the fee market will operate just like it does under "normal conditions", just at a higher fee level. The main benefit of EIP 1559 is that when fees are higher, the harm of regular fee market inefficiencies becomes more significant, so having a well-functioning fee market becomes more important.

7. Why is limit = target value * 2? Why not multiply by 4 or 8?

Going beyond 2 is easy. The higher the limit/target ratio, the greater the fee market efficiency gains of EIP 1559. It depends on how severe short-term spikes we are willing to accept, 2x is a fairly conservative number, and we could even launch with a multiple of 2 and increase that multiple over time, but only if the network can perform well in the presence of short-term spikes.

8. Why are miners included in transactions?

The EIP includes a "tip" fee that transaction senders can include, which will be sent to miners. It has two functions: first, if there are suddenly many more transactions than expected, miners will include transactions with higher tip fees first, so the fee-based priority mechanism exists as a backup. Second, it compensates miners for the risk of mining uncle blocks.

The tip fee level to compensate for uncle block risk is calculated to be about 0.8 gwei (uncle blocks receive an average reward of 1.67 ETH instead of 2 ETH, so there is a loss of about 0.33 ETH. The uncle block rate increases by about 0.025 for a 10 million gas block compared to an empty block, so the expected cost of 1 gas is =330m/10m*0.025=0.825gwei).

This tip fee is independent of the base fee (BASEFEE), so users can confidently set 1-1.5 gwei through their client implementation and expect their transactions to be accepted by miners.

9. Don’t miners have an incentive to collude to lower the base fee (BASEFEE) by making all blocks less than half full?

In general, the effectiveness of this strategy is limited because unless almost everyone colludes, a transaction not included in a block will be included in the next block, so the impact of this operation on the long-term base fee is negligible.

What miners can do, however, is a kind of "monopoly pricing". Suppose the sender of a transaction is willing to pay some extra fee to avoid delaying a block. Miners can refuse to accept transactions that do not contain the minimum tip fee T. They will lose some fees, but if the sender values ​​the possibility that you will be the next miner and increases the fee with enough consideration for their transaction, then miners may benefit from it, but in general, this strategy is seriously disadvantageous to miners.

Note that even if a miner adopting this strategy is successful, other miners will increase their revenue by more than their own (because other miners will free ride on the higher tip fees due to their actions), so this is not a centralization vector.

This doesn't drive the BASEFEE down to zero, instead it reaches an equilibrium where the BASEFEE is still the majority of the fee and the tip makes up the rest. This is because, unless miners are colluding (in which case we have a much bigger problem), miners will incur the full cost of not including a transaction and only gain some benefit from a higher tip.

If the community still believes that the risk for miners to adopt this strategy is too high, we can direct part of the revenue from EIP 1559 (e.g. 50%) to a pool, from which a small portion is used to increase miners' block rewards. This ensures that miners benefit from the high base fee, further reducing the likelihood of this attack.

The following is a schematic diagram of the proposed EIP changes:

1. Define account 0x35 as FEE_SMOOTHING_BUFFER and define FEE_SMOOTHING_CONSTANT = 8192;

2. Add an additional item to the block reward (base block reward and uncle and nephew block rewards are added at the same time). Set smoothing_reward = FEE_SMOOTHING_BUFFER.balance // FEE_SMOOTHING_CONSTANT. Transfer smoothing_reward wei from FEE_SMOOTHING_BUFFER to block.coinbase;

3. After applying this block reward rule, 1/2 (rounded up) of the EIP-1559 fee from the block will be added to the balance of FEE_SMOOTHING_BUFFER. The rest (i.e. the remaining 1/2) will be burned;

Note that in the context of Proof of Stake (PoS), it is best to enforce secret leader elections, as well as penalties for early disclosure, to prevent validators from simply accepting high tip fees and reaping the full benefit, since transaction senders will quickly know which validators are creating blocks.