Stable committee size

[bajtos]

Ideally, we want the size of each committee to be stable, independent of the current size of the network of SPARK checker nodes. The exact committee size requires calculation, but we think a number around 30-40 should give us a high enough probability that if we find a majority, then it's honest.

Fixed committee size is tricky to achieve in a decentralised manner because we don't know how many nodes will be online in the current round, plus nodes can come offline or online in the middle of the round.

The initial design fixates the number of committees created. This does not scale automatically and requires manual interventions to keep the number of committees aligned with the network growth.

We should research a better solution.

[juliangruber]

Can we assume that fluctuations will be small enough, so that he next round size will be estimated to be equal the last round size?

[bajtos]

Can we assume that fluctuations will be small enough, so that he next round size will be estimated to be equal the last round size?

I think we need to gather & analyse some data before we can answer that question. I genuinely don't know.

For example, I can imagine the network size can be very different during day-time in the Americas vs day-time in Asia if we have different number of participant on those continents.