Expected ranges of EB-related parameters

[SupernaviX]

Cross-pipeline EB references are controlled by the 𝜂 (chain quality) and L (stage length) parameters. An EB produced in pipeline i is allowed to reference one EB from each of pipelines i-⌈3𝜂/L⌉ to i-3.

If 𝜂 == L, each EB can only reference a single EB (from 3 pipelines ago). This is what our default configuration uses today (both of these parameters are set to 20). If 𝜂 < L, EBs can't reference other EBs at all and we're back to Short Leios.

• Is my interpretation of all of the above correct?
• Is 𝜂 < L a coherent scenario we'd want to support? Some of Brian's testing has used longer stage lengths.
• I know that L is a parameter of Leios which is fully under our control. Is 𝜂 also a Leios-specific parameter, or is it meant to be a property of the underlying chain (i.e. Praos)? Is 𝜂 something which we are free to tune? The original Leios paper says "Intuitively, a chain quality of 𝜂 means that there is an honest block among any sequence of blocks that was produced during 𝜂 time slots," which makes it sound like 20 (expected Praos block production rate) is the correct value for that.

[pagio]

η is supposed to be rather large to cover for the case that no honest RB ends up in the main chain, e.g., due to selfish mining or some other attack. So you can take η to be larger, reflecting the fact that under attack honest RB in the main chain may be somewhat rare, and we want to cover for this and not lose certified EBs.

[SupernaviX]

When you say η is how long it takes an honest RB to end up in the main chain, does that also mean it should be high to account for "quiet" periods when it happens to take more than 20 slots to produce an RB?

[pagio]

In a worst-case scenario where the chain is under attack, η can be a lot larger than 20 slots (e.g., multiple honest RBs do not end up in the chain).