This wasn't supposed to be an argument about the relative exploitability of the attack-surfaces presented by implementations of the technologies of PoW and PoS consensus systems, though; but rather about the economics inherent in "cornering the market" in PoW vs PoS.
PoW and PoS both allow a single party to unilaterally control the network when they're working exactly as intended — just under conditions that are impractical in major networks. That impracticality can be measured in economic terms — as a dollars-and-cents cost to acquire the assets necessary to take control.
It's harder to buy out 51% of validation stake of a network of size N, than it is to buy the GPUs and the electricity to power them to do 51% of the mining for a network of size N. GPUs are technically unbounded — you can just build your own fabs, if you really care — and so the marginal cost of acquiring another GPU+electricity is constant. But stake is a pie; the marginal cost of acquiring more of it increases as less of it remains available on the market. (And many people will just never give theirs up, for ideological reasons.) Thus, it's harder to Sybil attack PoS than PoW.
The hardness works both ways - these with stake can decline to process what they deem ungood and there's nothing one can do about it.
You see "hardness of Sybil attack", I see "rule of unelected comittee".
The very fact that GPUs are not pie, that GPUs are minted, gives an opportunity for a change if things do not go well for actual majority of users-as-persons.
And PoW, more GPUs is a quality bound to physical mechanics. In order to build more GPUs, you need more fabs capable of doing so. You need a more educated populace capable of building and improving GPUs. You also need better energy generation to not destroy the planet and provide energy for an increasingly more energy hungry society. This leads to improvements via physical conditions of survival, resulting in better macrosocial outcomes.
PoS says “you have money so you make money for no reason and you get resource allocation privileges absent binding to physical properties”. In other words, one receives all benefit without any effort. Yet effort is implicit. Which means it become abstracted and convoluted, allowing one to offset effort to no-stakers.
In other words, decision making rights lacking any consequence. Also known as a caste system. Also freezing the castes making advancement highly improbable, and deeply / forever embedding dynasty algorithmically.
There then exists no forces within the body of the mechanism to improve compute capacity nor energy gen efficiency. This, taken to the limit, is a highly undesirable outcome for advancement to the next stage of existence. In fact, it sets us back by a potentially significant margin.
The result is nothing short of digital aristocracy.
Sure, in the short term it will be a secure global compute that can operate more “efficiently” for some time. But look around you. Look at what happens when you let people make decisions without earning it / understanding the full impact of their decisions.
Yes, and? This is orthogonal to PoW vs PoS. In either case, you can treat the resulting cartel as "a single party" for purposes of modelling them.
It's also not really a scenario worth considering.
In order to cooperate, the miners/validators need to have full alignment on their goals. Which is very unlikely for Sybil attacks, which almost always destroy systemic economic value — including the value held by the attackers. You might be able to find e.g. a government who wants to do a Sybil attack to accomplish goal X, and are willing to burn down $N billion dollars of their own staked assets to do so; but can you find a bunch of other governments or large corporations who also want to accomplish that same goal X more than they want their own stake-value? And who can be trusted to coordinate and not leak the plan?
A Sybil attack is basically like a bank heist where the attacker(s) have to pull up to the bank in their own personal aircraft carriers (destroying said aircraft carriers in the process, from the hull damage of sliding along several city streets), without anyone realizing they're coming. You can maybe theoretically do it as one entity with full internal goal-alignment. It's much harder to do it as multiple entities.
One relevant question is which system allows for the pooling of resources easier. In PoW you have mining pools, and a member changing pools is trivial as changing an address. In PoS you have large organizations like Coinbase running staking for users. Don't know how withdrawls will work, but that could be a very significant amount of eth. Coinbase would have to vote a certain way to comply with laws.
> 1.4. Governance and Voting. For certain Digital Assets, the underlying protocols offer stakers the ability to vote on matters related to the governance of protocol-level issues. Coinbase may or may not support voting for such assets, and may cease supporting voting at any time in its discretion. Coinbase will comply with your instruction to vote your Supported Digital Assets to the extent Coinbase or its affiliate supports voting for such Supported Digital Assets. In certain cases, Coinbase may vote on your behalf where Coinbase or the applicable protocol does not support delegated voting; in those instances, Coinbase will vote with the protocol’s recommendation.
But what about jurisdictions that compel them to vote in certain ways?
> To be eligible to stake ETH, you must: [...] Live in a jurisdiction eligible for ETH staking.
> Which regions are eligible for ETH staking? [...] United States (excluding Hawaii & New York)
In other words: they're just not going to offer the service in any place that compels them to vote a certain way.
I presume most delegatable validators have chosen a similar option: avoiding the problem entirely by not taking anyone's money if they're from those jurisdictions. Just as most ICOs in 2017, when the US didn't have regulatory clarity on crypto being a security or not, just avoided the problem entirely by preventing (direct) sales to US-market investors.
But what happens if/when a large amount of user ETH gets staked in a single country, in a few entities? End users don’t have the motive or skills to move their ETH to platforms or systems that better protect decentralization. The downsides of PoS is worth exploring.
PoW and PoS both allow a single party to unilaterally control the network when they're working exactly as intended — just under conditions that are impractical in major networks. That impracticality can be measured in economic terms — as a dollars-and-cents cost to acquire the assets necessary to take control.
It's harder to buy out 51% of validation stake of a network of size N, than it is to buy the GPUs and the electricity to power them to do 51% of the mining for a network of size N. GPUs are technically unbounded — you can just build your own fabs, if you really care — and so the marginal cost of acquiring another GPU+electricity is constant. But stake is a pie; the marginal cost of acquiring more of it increases as less of it remains available on the market. (And many people will just never give theirs up, for ideological reasons.) Thus, it's harder to Sybil attack PoS than PoW.