I think that like anything, it's exploitable. Just see the example of bitcoin where a mining pool had over 50% of the computing power. If the internet was driven by a block chain, with enough money and time, you could "control it".
Owning 50% of the mining infrastructure is not an exploit. The system is designed with that in mind - anyone in control of that much hash power has no incentive to use it, because the value of the currency would instantly evaporate. They in fact have strong incentives to avoid putting themselves in that position - because it devalues the currency (which they have invested so heavily in), and makes them a target for attack. And that is exactly what we saw when ghash hit 50 percent.
Anyway, it's a false choice. I'll take the hyper-public well described risk of a 50% 'attack' over the risks of counterfeit paper currency manipulated in secret by unelected officials any day. Similarly, it'd take much more time and money to control a blockchain based internet than the one we have now, and if it were exploited in that way, it'd be substantially more likely to be known by everyone, which in itself is a deterrent to exploiting it in the first place.
All that being said, there are probably reasons that other architectures on which to build an overlay network are better than this - but 'exploitability' is not one of them.
>> I think this is a strange way of looking at things. What non-trivial protocol doesn't have known design limitations?
It's still a weakness, not a feature, and when a blockchain is not a monetary thing the pressures and incentives to run infrastructure will be different, it's hard to say that a 50% attack won't be possible or desirable in that circumstance.
>> Moreover, what could you possibly hope for in addition to a self-mitigating limitation, i.e: a theoretical limit that doesn't matter in practice?
It's not theoretical though, pools have got there with BTC. There are incentives not to obviously mess with the currency at that point but perhaps subtle ways to operate to your own benefit could happen at that point.
It's a huge design flaw though. The original intention of the blockchain in Bitcoin was to make it so it was decentralized...except the increase in computational difficulty, the winner-takes-all payouts, and the electrical cost, of course means that eventually the only people running the blockchain are large centralized organizations.
At which point one can wonder, why bother being distributed when we're still ultimately invoking trust in several groups? Couldn't we do this a lot more simply and efficiently by just having everyone delegate a private key signing operation to these groups so that it's just based on maintaining consensus, rather then burning CPU time on computationally hard problems?
Wouldn't people notice it though ? I mean if you hold 50% percent of that computing power, wouldn't someone detect that a large portion of the network traffic revolves around one place ?
Also people would notice that the internet would be "off".
Anyway, it's a false choice. I'll take the hyper-public well described risk of a 50% 'attack' over the risks of counterfeit paper currency manipulated in secret by unelected officials any day. Similarly, it'd take much more time and money to control a blockchain based internet than the one we have now, and if it were exploited in that way, it'd be substantially more likely to be known by everyone, which in itself is a deterrent to exploiting it in the first place.
All that being said, there are probably reasons that other architectures on which to build an overlay network are better than this - but 'exploitability' is not one of them.