Decentralization and Security: Why is PoS an exception to the rules of economics?
By Jan Xie and translated by Toya
The ability to provide security for decentralization is essential to the value blockchain technology promises. The level of security a blockchain can provide is a quantifiable property, which can be approximated by the cost of running an attack. How much that cost is reflects how much security there is—security is the good or service produced by the set of consensus forming nodes who serve as producers in a blockchain network.
Basic principles of economics dictate that, for producers in a perfectly competitive market, the marginal return will equal to marginal cost at equilibrium. In other words, when the consensus algorithm creates a market with perfect competition, the equilibrium where supply meets demand will be where: cost of running an attack = level of security = cost of production
Again, for every unit of marginal security supplied, producers need to pay an equal amount, as the cost of production. This seems to be at odds with a key claim PoS makes, that it is able to supply the same amount of security at 1/10 even 1% the cost. One must wonder why PoS is an exception to the rules of economics.
Possible scenario number 1: cost of production for security is the same in PoW and PoS, yet part of the cost gets translated into something else in PoS. Take DPOS, which typically subscribes to a ranked voting system, for instance, a party seeking to become a validator needs to first collect enough votes in order to be one of the first Nth preferred candidates in a hierarchy on the ordinal scale. It is reasonable to assume that there will be cost associated with getting elected—what about this cost? How much will it be? Will it be the same for different people/institutions? Is it good or bad if it differs? Will it be a piece of public information? If such information is not public, is that a good thing or a bad thing? Here I quote:
Production cost is made public in the most transparent way possible in PoW.
PoS is not in Fact Permissionless
A permission requirement for consensus undermines, at a very fundamental level, arguably the most important promise blockchain systems offer ,in that its existence in a blockchain would entail that the control over the network will be contained indefinitely within a certain group of people.
Another possible scenario: PoS does not create a perfectly competitive market.
In the case of PoS, the future consensus forming body is determined by the currently existing consensus forming body. Any node which is not already a part of consensus forming body seeking to participate can only become a part of it when at least one transaction (e.g. staking, voting. etc.) is completed, whereby the choice of whether or not this transaction will be validated is held by existing consensus form body. In the event of current validators indefinitely holding off validating the transaction in question, a new node may never be able to participate in consensus. Meanwhile, liveness faults, i.e. withholding or omitting a valid transaction, can be indistinguishable from a simple network latency thus difficult to detect. I have yet to see any solution that solves this problem to my satisfaction. PoS algorithms in general distribute block rewards in a way that is proportional to the amount of token at stake, which could only serve to exacerbate the situation, given the degree of token concentration in most systems.
PoW, on the other hand, ensures the network is completely permissionless in so far as absolutely no permission from current miners is necessary for any new miner to join and start mining as long as she/he is willing to gather the necessary resources, i.e. mining rigs and electricity. One reasonable question might arise here: would the requirement for resources such as mining rigs and electricity constitute a form of permission?
Yes, in a very broad sense, requirement for these resources amounts to a barrier to entry hence can be seen as a permission requirement. For better or worse, this is the most decentralized form of permission requirement in existence across the wide spectrum of existing Proof of X algorithms for the simple fact that resources pertaining to production of mining rigs and electricity enjoy a much higher level of decentralization than various token distributions. As a matter of principle, one should always aim for decentralization where possible, or else one might as well fall back to the good old centralized systems.
Perhaps here lies another question: what if we were able to design a PoS algorithm such that participation in the formation of consensus does not necessitate any transaction on-chain and instead only requires having token in one’s possession. Would that be sufficient to solve the “premission begging” problem?
The proposal under discussion actually introduces two additional problems:
- Token holders will necessarily have different levels of commitment to contributing to consensus in a system where possession of tokens automatically grants participation in consensus which, in other words, is to say there will be no commitment required since there will neither be cost nor penalty to non-participation. Designing a well-functioning consensus mechanism that meets such condition is indeed very challenging.
2. There is the nothing-at-stake problem, which I won’t dive into here.
Layer 1 Should Rely on PoW
If a layer 1 blockchain were to be the foundation of a network of crypto-enabled economic activities, then only a blockchain that is permissionless should serve as this layer 1 —an infrastructure that is global and neutral.
On the other hand, a permission requirement for consensus undermines, at a very fundamental level, arguably the most important promise blockchain systems offer, in that its existence in a blockchain would entail that control over the network will be indefinitely contained within a certain group of people.
In sum, if it is our wish that layer 1 be as decentralized and secure as possible, then the one and only viable option is PoW. That is not to say PoS has no place where it is appropriate, rather it is simply does not meet the requirement for layer 1.
Translated from: https://talk.nervos.org/t/nervos-pow/1639