I skimmed the immutability article. The argument is essentially that a country like China could wreak havoc on Bitcoin by acquiring 51% of mining power. His calculation is a bit outdated -- current mining revenue is over 7 billion [1] without even counting transaction fees -- but I guess the argument is still somewhat plausible.

More importantly, the argument doesn't apply to proof-of-stake blockchains, which I believe are the future. Unless a government is able to purchase a 51% stake in the currency, but that would require hundreds of billions for a currency as large as Bitcoin.

[1] https://www.google.com/search?q=12.50+%2F+%2810+minutes%29+*...

Proof of stake has other flaws. One issue in particular is that forks are much, much cheaper in proof of stake. This means that you can achieve "51% attacks" with decent probability with a much smaller amount of mining ability.

I've been working on a blockchain design involving PoS, and it's certainly more difficult to get right, but it's a solvable problem.

51% attacks normally require 51% stake. It's true that large stakeholders will occasionally have several winning block scores in a row, which lets them create a viable short fork, but it's not really a problem. Short forks just mean that users need to wait for several blocks if they want a high degree of certainty that a transaction is final.

51% attacks could also be more feasible if an attacker can manipulate entropy, but the usual safeguard is to derive entropy from verifier addresses, as NXT does, so that manipulating entropy will cost a verifier their block reward. Then require verifiers to have a certain minimum account age, high enough that grinding attacks become infeasible.

Certainly more complex, but with the right design PoS works fine while saving electricity, and the cost savings are passed on to all stakeholders.

You're assuming that didn't already happen.

The majority of the supply was produced rather trivially for a measurably inexpensive amount of captial cost to mint it prior to 2015.

One could surmise the reason crypto isn't banned yet is because nation states have the resources to mine majority stake easily due to the majority supply being created in a short time frame.

It's possible that they acquired a majority of hash power when doing so was cheaper, but in order to maintain their majority, they would have had to invest several billion since then. See [1] which shows the growth of the overall network hash rate.

I'm not saying it's unrealistic, just that it would be more expensive today than what the article says.

[1] https://blockchain.info/charts/hash-rate?timespan=2years

51% is not necessary.

A case study in censorship though the pools has proven this:

https://steemit.com/ethereum/@dhumphrey/f2pool-manipulates-u...

51% is necessary to create viable forks with arbitrary lengths. What that mining pool did was much less harmful: they just censored other parties' transactions during an ICO.

Unless a coordinated majority is censoring your transactions, you can't really be censored, your transactions will just be slightly delayed sometimes. Which is normally very harmless.

The only problem here was a poorly run ICO, where prices were dependent on the order of transactions. That wouldn't be the case with e.g. a reverse Dutch auction similar to the process for Treasury bills.

So, you are using bitcoin's mining revenue to make a claim on PoS coins?

Assuming you are correct, here's something you dint consider - most of the PoS coins need a starting point - mainly the genesis block. These coins are acquired through ICO mechanism. Example NEO. While the current price is 122 USD @ 8 billion in marketcap. But, it started at 0.3 USD and raised USD 28 million. So, China need not buy now 4 billion of coins now, it could have spent 15-20 million to acquire more than 50% of the staking coins.

Even if they acquired a 51% stake on the cheap, they would still own $4b worth of NEO today. Revealing that they hold a majority stake would destroy NEO's price, since it breaks the security model, resulting in a $4b loss.

If you want to run a trust dependent system, just use a database. A database works quite well in most cases.

You need trust as soon as you want to interact with real-world entities anyways.

This is only true during the transistional period we are in now.

> A global distributed ledger is less cumbersome to develop and more efficient to run, when it uses a transactional NoSQL database, than a blockchain.

In other words, a distributed system is less cumbersome if it is centralized.

Blockchain is a very expensive technique, but it's the first attempt at developing document timestamping system (one that doesn't need trusted third party) that produced at least mildly robust protocol.