If you are interested in offline staking solutions and a comparison between some of them (TPoS, LPoS and DPoS)
here is an article my brother wrote about StakeNets / XSN solution (trustless with merchant):
https://medium.com/stakenet/cold-staking-tpos-vs-lpos-vs-dpos-ebeef4f71c48he's comparing with NXTs LPoS (leased) and BTSs DPoS (delegated) ... might be interesting for NXT holders as well
I think the main difference lies in the incentive system to form large pools or not.
- In DPoS, there is
only delegated staking, no "individual" staking of non-delegates is possible. This has some advantages - mainly the stability of the validator set, which is crucial in PoS systems - but also disadvantages. As all "staking power" is in the delegates (which are like forging pools/merchants), a cooperation of 50% of the delegates could be fatal. Also there's the risk of the blockchain "stopping" when less than 66% of the delegates are online.
- LPoS makes delegated staking non-trustless, as the payout is not automatic. The sense behind that is that pools should not become too large, and people should worry about to whom they "lease" their staking coins. The disadvantages are obvious: The pool operator could run away with the coins (although in NXT/WAVES system this isn't really dangerous atm because the forging rewards are small).
- TPoS makes delegated staking trustless, by automating payouts. The advantage is that the merchant cannot run away with the minting rewards. However, this has also a backside: the incentive to form large pools ("merchants") "staking" with most of the coins is bigger, because one hasn't to worry about trusting a particular node. The larger the pools become, the more "oligopolic" the network becomes, and thus the higher the vulnerability to stake-51%/N@S attacks. XSN mitigates that by requiring a static IP for every "stake contract", but this could be a weak protection in the context of IPv6.
As a conclusion: I slightly prefer TPoS over both other solutions (NXT's LPoS coming second, and DPoS third because of its vulnerability to social engineering) but the "large pool problem" continues to worry me.