Going multidimensional - an empirical analysis on gas metering in the EVM

Note that the conversation about this proposal continued here: A practical proposal for Multidimensional Gas Metering

I hadn’t read the proposal thoroughly when I replied here (my bad) but the other thread has more insights.

As it stands, I think this proposal would be detrimental to Ethereum for two reasons:

  1. It wants to create four separate gas limits whose individual values will be “similar to today”[1]. IIUC that means all users will be able to use four times as much overall gas, I don’t think that’s safe. If the current gas limit is tuned to accommodate a certain type of hardware for full nodes, quadruplicating it means full nodes will need much beefier hardware. That in turn will generate higher centralization.
  2. What I explained in both threads: nodes are encouraged to prioritize transactions with evenly balanced resource usage to get a better chance of using up all four gas limits and earn more rewards. Basically a new MEV criterion, making the whole MEV problem worse.

Fixing #1 is relatively easy by dividing each limit by 4 but would cause a regression compared to the current state: if you get an extensive run of transactions that are intensive on a single dimension, today’s Ethereum will happily run them but after splitting the gas limit in four it would only run a fourth of them (or rather, it would take four times the number of blocks to run them).

As for #2 I think the only way to fix it is to allow wallets to pay higher fees for specific dimensions, but that would consistently result in higher fees than today unless a transaction is perfectly balanced.

I think multidimensional gas metering is bound to create regressions one way or another.


  1. A practical proposal for Multidimensional Gas Metering - #8 by dcrapis ↩︎