I would love to see you write something about the scalability of DPOS vs POS vs POW when it comes to decentralization.
Green is Income
Yellow is Profit
Blue is Cost
Assuming a fixed validation cost per transaction and a fixed fee per transaction, there is a limit to the amount of decentralization that can take place.
Assuming the validation cost exactly equals the fee a network is completely centralized and can only afford one validator.
Assuming the fee is 100x the cost of validation then the network can support 100 validators.
What you will quickly see is that systems like Nxt and Peercoin must have very excessive fees if they would like to maintain the illusion that everyone can be a validator and earn fees at scale. What this means for Nxt / Peercoin is that anyone with less than 1% stake cannot validate profitably unless their fees are higher than our DPOS chain.
So if these chains assume 100 is to centralized and start promoting they have 1000 validators then their fees must be 10x. If they grew to be the size of bitcoin ($10 B) then only those with $1M worth of coin could validate profitably and most would consider that an elite club. If they reduce the minimum stake to be a validator to $1000, then their fees would be 10,000 times higher than our fees.
Now we assume that every one with less than the amount required to validate doesn't participate and we assume a "reasonable" distribution of wealth, you will quickly see that unless they have insane fees, there will only be a handful of people with enough stake to validate profitably.
What we can conclude from this is that the only way for POS to work is to delegate it. In the case of Nxt they can pool their stake by some means and ultimately this will end up like DPOS prior to approval voting with a variable number of delegates. Those doing the delegating wouldn't actually receive any income like you see with mining pools because the validation expenses will be consuming the vast majority of the transaction fees.
End result is that decentralization has a cost proportional to the number of validators and that costs do not disappear. At scale these costs will centralize any system that does not support delegation. We should design this kind of centralization in from the beginning so that it can be properly managed and controlled by the users instead of evolving in some adhoc manner as an unintended consequence.