News for hackers. Stuff that matters.
Re:Ethereum: Rise of the World Computer by andrewla on Friday February 26, 2016 @02:49PM (#11183674)
I'm of the minority opinion that the Turing-complete nature of Ethereum is an anti-feature. [1]

The features that matter in my mind are things that make Ethereum superior to Bitcoin:

1. Shorter block times without compromising security -- GHOST was proposed for Bitcoin, but it's a long way from happening.

2. Every block contains the Merkle root of the balance tree, so proving account balance becomes a one-node + proof problem instead of an "entire blockchain" problem -- the only way that I can verify that a UTXO is valid is to have the entire blockchain forward from the UTXO block. In ethereum, I just need the root plus a proof at the top of the chain.

3. No reward halving. The fee schedule is constant, which means that the marginal rate of growth is diminishing, but the logic across the board for utility calculations is incredibly simple.

4. Easy offline transaction creation -- the properties in #2 means that all you need to generate a transaction is the nonce(s).

5. (possibly) ASIC-resistant proof-of-work. I'm not completely sold on this one; all it really means in my mind is that the cost of such ASICs is higher on a per unit basis, which means that the power is more likely to be consolidated than the double-SHA256.

[1] Why do I think the whole Turing-complete thing and smart contracts are useless? Because fully-funded contracts are economically uninteresting. Basically, only zero-sum contracts can be encoded. If I borrow money from you at a given interest rate, I can do economically productive activity to make more money, and both pay you back and come out ahead. If I am instead forced to keep the total payback amount in escrow, then I have no capital to use, and the transaction is meaningless. Things like cryptographic ownership tokens (colored coins, etc.) have some limited utility, but as you say, those are easily done on the Bitcoin blockchain.

Re:Ethereum: Rise of the World Computer by wocram on Friday February 26, 2016 @11:41PM (#11186097)
You can definitely enforce non-zero sum contracts, the problem being whether they are actually useful. If you're handing out loans, nothing stops the borrower from vanishing and creating a new identity or otherwise resetting their 'credit score'. It seems like a solvable problem though, you could then depend on some identity or credit oracle, and that could be built on trust (eg. in a government or facebook id), or some meaningful reputation scheme.

I guess the trick here is whether you consider "losing credit score if you don't re-pay" enforcing the contract or not. Doing anything stricter, like programmatically garnishing wages or handling bankruptcy, seems like it would require a hard 1:1 mapping between a person and their ethereum presence, which is much more than trusting an oracle to provide a unique ID or reputation based credit score.