A short write-up on "Dual-deadline PTC vote in ePBS". Idea first proposed by @EthDreamer Indicated times in the figure very much tentative and subject to change.

RT @soispoke: The paths of least resistance: Introducing WFR-Gossip. tldr: WFR-Gossip applies optimal transport principles to Ethereum’s go….

Considerations for the BLOB_BASE_COST in EIP-7918.

RT @barnabemonnot: Had a great time presenting :) Slides are in the next tweet 👇.

RT @soispoke: 🦖🦴 FOCIL for Glamsterdam 🦖🦴. In this note, we propose EIP-7805 as a headliner for a fork inclusion in Glamsterdam. We highlig….

RT @weboftrees: @timjrobinson @MikeIppolito_ @tkstanczak EIP-7918:. 1. Ensures that the cost of the purchased blobs stays relevant compared….

When a girl asks you how much you love her and you need to spend the next ten minutes explaining that you said 1.0 because it is the maximum.

1. It would be neat to use the blobSchedule constants if the relationship target/max = 2/3 is expected to change (i.e., to access max on the EL), keeping the linear logic. 2. A non-linear variant proposed by @adietrichs instead thresholds the regular return statement at 0. Fin.

There are also alternative proposals for the return statement (second line of code), when the if clause evaluates to true and the mechanism operates in the thresholded regime. The protocol should then raise (or at least not reduce) the blob base fee.

Here are associated blob base fee histograms for the four-month period from November 2024 through March 2025. The black histogram is the observed fee and the black-and-blue dashed histogram is the fee imposed by EIP-7918 with the adjusted constants.

The blue line is the minimum blob base fee (38% of exec. base fee) and the black line is the observed market-clearing blob base fee for two recent 3-week periods. The maximum of the two lines (darker) would be the new blob base fee (ignoring shifts in the equilibrium).

Relying on POINT_EVALUATION_PRECOMPILE_GAS would increase the minimum threshold quite a bit, warranting further consideration (some fraction of the constant could be considered or the log scaling applied). The blob base fee would at a minimum be 38% of the execution base fee.

The link to the execution base fee ensures that blob consumers pay according to how in-demand compute services of the nodes are at the time they are requested. As per the original design, the link also helps establish the minimum reserve price of a functioning fee market.

The EL costs remain fixed (in the worst case) whereas CL costs fall with higher blob count, roughly according to a log-reciprocal. Parallelization across columns is possible on the CL, but the point evaluation itself is hypothetically also mostly parallelizable during processing.

We can measure a node's compute in "point evaluation multiples". Batch-verifying the 128 cell proofs of EIP-7594 for each blob added to the mempool is 15 times more expensive (red) than the point evaluation. The CL verification cost depends on how many columns a node custodies.

There are still just two lines of code. The original constants establish a low minimum blob base fee that ensures a functioning fee market. The new constants would raise the minimum by charging closer to the workload blobs impose on nodes, still ensuring a functioning fee market.

The rationale is that nodes must perform the same core KZG proof verification when processing blob data as in the point evaluation, with the POINT_EVALUATION_PRECOMPILE_GAS specifying a base cost for this work. In this write-up I analyze the implications.

There have been a few suggestions during the discussion phase for EIP-7918, and I will here detail the most interesting ones. First, @ben_a_adams has suggested that we use POINT_EVALUATION_PRECOMPILE_GAS instead of TX_BASE_COST for the constant, also skipping amortization. 🧵👇

RT @barnabemonnot: The Ethereum L1 is scaling, with active work for short- and medium-term improvements. @dankrad here tells us the tech re….

Beneath the memes, @adietrichs walked us through an impressive range of presentations that show just how seriously Ethereum takes execution layer scaling.