Mining is supposed to provide security by making the ledger difficult to forge.... however, once there are 6 blocks on top of your transaction you really don't care about subsidizing additional mining. After all, by holding your shares (coins) you do not ask anything of the network.
What if we were to shift the burden of security from all bitcoin holders, to just those want to make transactions. Move 100% of the security to transaction fees. This is the long-term model that Bitcoin aims for right?
Assuming 1 MB blocks... 25 BTC per block, 256 bytes per transaction, and $600 per BTC, then we can estimate that the average transaction fee should be .006 BTC or $3.66.
Now most people would argue that charging that kind of transaction fee is insane and I would tend to agree. Factor in that most blocks do not contain 1 MB of data and the actual average fee-per-transaction would be near $7. You can bet the network would not be used for every day transactions or micro payments!
Now there is really no reason for the transaction fees to be so high, the network would still be secure with fees of just $0.36 per transaction.
So I would like to suggest a new kind of blockchain which is a modification of Bitcoin.... dynamically adjust the transaction fees to target a desired bandwidth for your decentralized system. When the volume of transactions is higher than target, raise the price and when it is lower, reduce the price.
You are still securing the network with hash power, but now the hash power is only rewarded for including transactions.
Now lets add TaPOS to the mix. With TaPOS the network is automatically secured against VERY LONG alternative chains and the block chain eventually becomes irreversible. TaPOS is like an automatic checkpointing system.
There remains only one kind of mining attack that could result from 'centralization'... denial of service. This is an attack where miners refuse to include certain transactions (ie: by government decree). This kind of attack is easily mitigated by refusing to relay blocks that exclude valid transactions you have known about for more than 1 block.
Lastly you can reduce the block interval down to 1 minute and set the maximum chain fork to 5 blocks. This will result in the network quickly reaching consensus on which chain is the best, it would become irreversible (except under extreme cases where the physical network splits). The long-splits would be resolved by TaPOS and not by hash power.
Burn 50% of the transaction fees as dividends and you now have a system that is secure, pays dividends, and is decentralized.
Obviously it would be better if the network had 0 mining at all... after all the miners are only performing a single role: selecting the next block / short term security. This role could just as easily be performed by a central (trusted) signing authority without undermining the security or irreversibility of the network. After all this trusted signing authority would only have one power.... to create blocks... and that power cannot be used to double spend due to TaPOS + the fact that everyone would KNOW they were the ones behind the attack. It also couldn't exclude transactions without the network forking and a new trusted entity taking charge.
Given that blockchains have not been deemed illegal it may actually make a lot of sense to have a central party overseeing the creation of blocks to maximize performance and dividends for everyone involved.