RE: Are You Seeing the Truth? What Is Your Vision of Steem?
So maybe non-developers shouldn't be making statements about what is or is not trivial?
Yup, hence "I think" since I'm not familiar with the practicalities in code. Or perhaps multiple linear approximations hugging the mild superlinear curve could do the trick. Not very professional though but I think it'll serve the purpose.
That is one reason why I support the simpler (if somewhat imperfect) measure of deducting some voting power from each comment (or post). It amounts to a mild 'superlinear' (in practice even if not in strict mathematical terms) as the effect of the one-time deduction is burned off (twice the vote power would always produce more than twice the rewards, with the effect approaching linear at high rewards).
Seems like an equivalent of a simple increase in the floor-threshold, haven't thought about this through that well.
There may still be implementation complexities there. For example, how to calculate large votes that cross many transition points without introducing more calculation overhead (votes and their associated payouts are the most frequent transaction on the network by far so their computational cost is important), or creating opportunities for gaming by placing specific calculated votes around the transition points. I don't know there are, but I do know that any code going on a blockchain needs to be carefully analyzed first because those trying to exploit it will often do so. I'm mostly guessing but factually Steemit and its devs (unquestionably the most experience and familiar with the code) have definitely said that tinkering with the curves is NOT trivial.
Slightly different (but IMO the difference is quite important) because of what happens when the threshold is reached. With a deduction it remains non-linear on the low end. For example, if the threshold is 0.10 and you can vote up to 0.11 with your own (or delegated, bot vote power, etc.) then you get full linear value of your vote. With a deduction of 0.10 you get only 10% of linear value on your vote (a vote of $0.20 gets 50% value, $1 gets 90% value, $10 gets 99%, etc. so linearity is approached on larger per-comment payouts, which seems to be what most advocates of 'mild superlinear' want).
Still, just increasing the threshold would help somewhat with spam (and processing overhead of small payouts) and wouldn't be bad either.
(BTW, comments and posts are equivalent in the code, I use the word 'comment' above and probably elsewhere when referring to both.)
yup i think i understand your suggestion now regarding the roundabout way of simulating modest superlinear, traf has also said the same it seems. as for transition points in the case of multiple linear approximations, probably won't be as great of an exploit considering what's happening at Steem's current state..