I was thinking of the voting issue as well. So far, a pattern is emerging.
Let v be the voting function, mapping the amount x in one account to it's voting weight: v(x) -> weight of voting. Let's further assume v is monotonically increasing (the more you have the more your vote weigh in) and differentiable (so we have a v').
There are three cases:
1) v' is increasing => merging accounts is more valuable regarding voting => merging encouraged
2) v' is constant => 1 Nxt = 1 Vote; conservative; no change encouraged
3) v' is decreasing => splitting accounts is more valuable regarding voting => splitting encouraged
From my perspective, distribution of nxts is a good thing. So, higher amounts of fees are a good thing. Therefore, splitting and merging accounts should be encouraged as big shareholders then have to give away fees. These fees are then given to the ones forging and stabilizing the network.
On the other hand, splitting and merging just because somebody wants to act in the best interest of the network seem to me like spoon-feeding that somebody.
So, why not making v dependent of the amount of fee somebody is willing to pay for it. I favor a logarithmic scale: v(fee) = log(fee).
In that case, one could even imagine that these special fees aren't just given to block forgers but are evenly distributed among 'all voters' or 'winners' or 'loosers' of the voting session.
So, voting is encouraged.
Please, can you put your approach in a mathematical expression?
That would be really helpful!!!
Just write it down in one or more terms.
Don't care about if its bullet proof.
If you like, you could add it to collection of voting models?
cheers
Errr, I try. Not sure, what is it exactly you need? Unfortunately, I do not find any formular symbols here.
An example; two accounts x and y:
1) Let's confine ourselves to multiple-choice-polls with some options O.
2) Each account can pay some Nxt an option o. (multiple options allowed).
3) The weight w(o) of each option o is: log( x1 + x2 ... + xn ) + log( y1 + y2 + .. + ym ) where xi are the fees paid for o by x and yi are the fees paid for o by y.
4) The set of options O can be sorted by the weights of each option.
-----------------------------
4) Some aspect which come to my mind recently would be an aging effect. That is each coin spent on an option is worth less and less in the future.
In the example above:
w(o) = log( x1/tx1 + x2/tx2 ... + xn/txn ) + log( y1/ty1 + y2/ty2 + .. + ym/tym ) where txi and tyi is the amount of time passed since spending on option o.
4) would force account holders to repay their opinions if they find them worth it. So diminishing their stake over time.