gas: Analysis of l-Smoothing

Why l-Smoothing?

There are a few arguments for l-Smoothing over fee-burning: maintains current miner incentives on state-growth, a significantly smaller and less contentious change to economic policy, avoiding the creation of large amounts of hashrate on hashing brokers during the market discovery phase, and improved security versus fee-domination.

Why Should EIP-1559 Care About Contention?

Aside from the lowered security following a contentious fork, a contentious hard-fork is the sign of a governance failure because some stake-holders were ignored to such a degree that they chose to try and go their own path instead of continuing to work in the current environment. This is important because contentious hard-forks have the potential to cause users to withdraw from the network in order to watch where the dust falls. In an environment where there exist other blockchains delivering on the promise of PoS with lower fees, it is important to consider that not all users will return.

Since contention is something, which is not a positive aspect of any change, it makes sense to pick options that minimize contention as long as they do not materially impact the proposal and do not weaken the security of the project. In the case of l-Smoothing, Vitalik Buterin himself has said that the difficulty of adding l-smoothing is likely being overstated. I endeavor to show below that l-smoothing is better or comparable to the performance of a burned base-fee and tip model.

The Hashing Broker Threat

David Stanfill elaborated on a point by BitsBeTrippin here stating that the sudden and inorganic dcrease in profitability would drive hashrate in to hashing brokerage services in order to try and avoid the struggle of switching between multiple coins while the market reaches a profitability equilibrium. This means that an incredibly large amount of hashrate will become avaiable at slightly above market rates for an attack on the network by any interested party. This is a major concern since trust in Ethereum is partially based in the lack of major security breaches in the Ethereum protocol.

Incentive Analysis and Security Concerns of l-Smoothing

Block-Stuffing Attack

I've performed analysis of block-stuffing attacks in another hackmd.io to avoid distracting from the other analysis. The main takeaways are that it is not viable in the current Ethereum environment, and it suffers from game-theoretical instability that makes it too difficult to perform. In the event that it does become a concern, a proportional burn of the l-smoothed reward can make it non-viable in case the ethereum environment shifts.

State-Growth Incentives.

I've written about this issue in a separate hackmd.io in order to avoid bogging this note down. The takeaways are that the current implementation of 1559 creates an incentive to increase the gas limit to lower overall transaction fees, This will create state-growth which is unsustainable for the hardware of current nodes. Since miner tips are primarily expected to be a flat rate that is greater than their marginal cost of inclusion, a lower transaction fee means that miners get to claim a larger proportion of the transaction fee. l-Smoothing avoids this issue by ensuring that miners are able to harvest a stable proportion of the transaction fee, which means that their state-growth incentives don't change whether the transaction fee is high or low.

l-Smoothing in Relation to the Dangers of a Fee Dominated chain

A few opponents to l-smoothing and paying miners transaction fees have brought up the paper by Miles Carlsten that states that fee-dominated blockchains are uniquely vulnerable to near-head reorganization attacks in order to steal blocks with fees substantially greater than the preceding block. The l-smoothing proposal has a lower risk in comparison to the current model of Ethereum rewards as well as in comparison to the burned base-fee proposal.

In the current Ethereum model, blocks can have a fairly large variance given that major smart contract interactions occasionally need to outbid many other transactions in order to be placed in a block. Under the fee-burn model the chain becomes more vulnerable to this flaw. Most transactions that pay large fees pay them in excess of the cheapest transaction in the block. Post-1559, we can expect that to translate into a willingness to pay a large miner tip to outbid other users since that is a willingness to pay above the standard rate. Since most blocks are expected to be 2.0125 Eth (basefee plus marginal cost of gas) it has a much lower resistance since the next block is predictably low and relatively close to the uncle reward. We can calculate the what proportion of the hash-rate needs to be controlled by 1 coalition to make that profitable by taking the inverse of the squared the percentage the coalition controls and multiplying by the difference between uncle reward and the next block's reward (.2625 Eth).

The l-Smoothed proposal combines the higher average block rewards of the current model with the reward stability of a burned base-fee model, which strikes the most effective balance. Every block's next reward should be within 1/n of the last-block's reward, and the inclusion of the base-fee in the block reward means that even greater tip instability is required to make it profitable to attempt a near-head re-org.

Tragedy of the Commons and l-Smoothing

Adrian Sutton raised a good point about the potential for l-Smoothing to suffer from the tragedy of the commons, which is worth addressing. Since the l-smoothed reward is independent of work done inside the block, there exists an incentive to avoid doing any work. For example, if a lazy miner decided to submit an empty block, then they can reduce their uncle risk as well as receive the l-smoothed reward. This is bad since it could reduce the total transaction throughput and create an incentive to avoid actually contributing to the network.

In this instance, the mining tips are our savior. Since miner tips must be above the marginal cost for a miner, the expected profits for a miner that chooses to actively participate in the Ethereum network are higher than those that opt to disengage from the Ethereum network. Even more so, when a lazy miner opts to disengage from the network, then the base-fee will drop and allow for the next block to be double-filled with a net-neutral impact on the base-fee. This means even more profits for miners who fix the damage done by lazy miners.

Minimum Viable Issuance

Vitalik Buterin's point in his discord post that was referenced earlier is that Ethereum should pursure minimum viable issuance. l-Smoothing is not in opposition to that goal, and it evens offers more nuanced opportunities to pursue that goal in a graceful fashion. Since a proportion of an l-smoothed reward can be burned, it becomes possible for the impact of a reduction in the transaction fee reward to be performed over time by slowly raising the burned proportion. This important since it allows the simulation of an organic decrease in hashrate similar to a fall in price as opposed to the sudden drop caused by sudden reductions in miner reward. This avoids the dangers from large amounts of hash-rate suddenly becoming available through hashing brokers as discussed earlier.