Epoch marker entropy value

[davxy]

Since for the block seal we’re using $\eta[3]$, this effectively establishes $\eta[3]$ as a required value for blocks verification.

Suppose we announce the epoch marker $H_e$ during epoch $N−1$. It includes:

• Fallback validator keys for epoch N (for use if the epochal lottery fails to complete).
• $\eta[1]$ as in epoch $N−1$ (which corresponds to $\eta[2]$ of epoch $N$).

Now, if this announcement must be really for light clients who want to verify block seals, it might be more advantageous to include $\eta[2]$ as in epoch $N−1$ ​ (corresponding to $\eta[3]$ of epoch $N$). This way, $H_e$​, and optionally $H_t$​, would be sufficient for light clients to verify seals for epoch $N$ blocks.

Including only $\eta[1]$ (as we're doing now) requires the light client to use two $H_e$​ markers: the one from epoch $N−1$ (for the fallback keys sequence) and the one from epoch $N−2$ (for the entropy).

What about modifying $H_e$ to also include $\eta[2]$ for convenience?

Including $\eta[1]$ is still useful as it is used to shuffle the keys in case of fallback and it is set as the epochal entropy which can be used to check for core assignments, etc.

[davxy]

This is not a request, but rather a prompt for reflection on the matter

[gavofyork]

That sounds reasonable. Want to make a PR?