Pre-computing work - to limit or not

Are you in favor of limiting pre-computing of work or not?

If you answer this poll, perhaps you'd like to follow the Work Generation series.

  • Yes
  • No

0 voters

This limitation of course could take a variety of forms, but likely wouldn't involve removing it completely. It could be limiting to a single pre-computed transaction per account, limit based on periodical changing value on the network, or other methods.

per-account would solve vertical spam but not horizontal, i think if each precomputed work only lasted an hour or so, that would be optimal (e.g. network value changes every 10 mins, after 6 changes work is no longer valid; if it was just hourly then a precompute made at :59 is no longer valid in a minute)

Yeah the horizontal issue is a tough one and the limiting of pre-computing based on time is one of the top suggestions to solving that. If that were to be put in place we'd have to ensure balance the between pressure towards on-demand work vs. the extra resistance to a pre-computed attack.

Feedback from services would be key and I would suspect many would be looking for a broader timeframe than an hour for valid PoW so they can do their own batching of work generation less frequently, especially since they would likely be using cloud services that come with startup times for each work generation session. Plus many may be able to solve for the peaks and variations in work needed with a bigger timeframes (more predictable trends of work needs across 24 hours is likely vs. a few hours, etc.).