[EM] Heitzig consensus ...

[Kristofer Munsterhjelm]

On 31/07/2020 03.09, Forest Simmons wrote:
> In general it is easier to find a lottery that is unanimously
> preferred over the default lottery than to find a deterministic
> alternative that is unanimously preferred over the default lottery. 

So a good consensus mechanism should both find as good a lottery as
possible, and then try to arrange that the deterministic alternative
is the best one.

I guess there's a tradeoff between randomness and consensus acceptance,
so to speak. The whole reason a deterministic consensus is a good thing,
and better than a random consensus, is that it leaves less to chance. If
we didn't value determinism as such, then deterministic consensus would
be no different from any other unanimously accepted consensus lottery.

On a related note, how would you guard against a "repeated referendum"
strategy with repeated voting, e.g. in a parliament? Suppose a faction F
wants its favorite implemented at all costs, so it proposes F in every
possible election/proposal and relies on that as the number of elections
goes to infinity, the probability of passing this favorite goes to one.

Clearly, if F is so dead set on getting its favorite and nothing else,
there can't be a consensus. The "exploit" is that F, in essence, gets to
play over and over until the favorite *is* passed.