5 ms·

The idea that marginal costs necessarily approach marginal revenues is plain wrong. The Ethereum wiki [1] says it best: > It's not enough to simply say that ma

The idea that marginal costs necessarily approach marginal revenues is plain wrong. The Ethereum wiki [1] says it best: > It's not enough to simply say that marginal cost approaches marginal revenue; one must also posit a plausible mechanism by which someone can actually expend that cost. For example, if tomorrow I announce that every day from then on I will give $100 to a randomly selected one of a given list of ten people (using my laptop's /dev/urandom as randomness), then there is simply no way for anyone to send $99 to try to get at that randomness. Either they are not in the list of ten, in which case they have no chance no matter what they do, or they are in the list of ten, in which case they don't have any reasonable way to manipulate my randomness so they're stuck with getting the expected-value $10 per day. The author here suggests that grinding attacks could be that plausible mechanism, but in a PoS system with a minimum account age, grinding attacks depend on being able to predict the state of the blockchain's entropy after that minimum age has passed. See [2] for some ideas about how entropy can be generated. Keep in mind that grinding attacks aren't free due transaction fees, so an attacker needs a certain degree of confidence that an entropy state will occur in order for a grinding attack to have a positive expected value; they can't just create a huge number accounts to cover all the possible states. In a well-designed system, the minimum account age and minimum transaction fee will be chosen such that all grinding attacks have negative expected value. [1] https://github.com/ethereum/wiki/wiki/Proof-of-Stake-FAQ https://github.com/ethereum/wiki/wiki/Proof-of-Stake-FAQ [2] https://vitalik.ca/files/randomness.html https://vitalik.ca/files/randomness.html

zby 9y ago
The quoted text does not support your thesis - come on! It admits that C=R might be true - but argues that this is not enough.
- dlubarov 9y ago
  I see how you could read it that way based on the first half-sentence, but if you read the rest, it's saying that C=R is not always true.
middayc 9y ago
> then there is simply no way for anyone to send $99 to try to get at that randomnes can't I send one of the 10 people $99 today, and he pays me out in 10 days? Can't that person rent the fact he/she is on the list of 10 out. I am new to economic terms, but it seems your example even better shows his MC_rent + MC_nonrent = MR. Where he says in p2p systems MC_rent should be 0, because it's exclusionary. Again .. your example's main feature is that it excludes all but 10 people.
- dlubarov 9y ago
  > can't I send one of the 10 people $99 today, and he pays me out in 10 days? But if he already received the $100, why would he trade his $100 for your $99? Are you saying the recipient might want $99 now rather than $100 later? Let's say that the donor doesn't give any advanced notice; the recipient doesn't know they've been selected until they see $100 appear in their bank account.
  - roenxi 9y ago
    The example given is a bit fuzzy about what it is showing and what assumptions are being made. This example is pretty bizarre because we have actors behaving economically irrationally (the $100 provider) and beneficiaries (I'll call them listees) who don't seem to have volunteered. I think it might be meant to simulate the creation of new money, but I don't accept the situation as being similar enough to be sensible (because no money is being created). Mr-give-away-$100 is burning resources (his cash reserves) to create an asset for 10 people (the potential to be given money randomly in the future, which is a strange financial asset). That asset is absolutely saleable on a secondary market and will be priced at some value south of $100. We can deduce who is on the list by observing who gets the $100 payments and deduce the probability after a few days of observation. The listees have incentive to provide this information as asset + secondary market is better for them than asset + no secondary market. Upper bound: Nobody is going to pay >$100 for a promise of $100 (I lie, that does happen, but only in rare and extreme circumstances. We'll ignore those). Lower bound: If the listed individual needs money NOW for some reason, which does happen in practice, then they may be willing to sell at any price >$0. Mechanism: A contract. They work well for events that happen in the future. Motive: The listees are giving up a small amount of value to smooth out their income stream. Results: When the $100 is transferred out of Mr-give-away-$100's account it will immediately be transferred to someone who bought it for some value $100 > X > $0; at a guess probably in the high $90 range. The person who was gifted a free asset will get rich, but that can be realised before the actual payment is made.
lalaland1125 9y ago
This is actually addressed in the article. See section: IS A “WORK-INDEPENDENT” PROTOCOL POSSIBLE? > If any cryptosystem is to periodically release coins, without immediately creating an incentive to “waste” an amount equal to the value of those coins, the cryptosystem is going to have to release the coins in a manner which is totally independent of all possible human activities. The coins will have to be rewarded on a completely effort-blind basis. The coin-reward must have a Spearman correlation of zero with everything that mankind could influence. If you can make it so that there is no way for human influence to affect who gets the coins, then of course marginal revenue will not necessary approach margincal cost. However, it's quite hard to design a system that doesn't have at least some way to influence who receives the rewards. And once you have some way of paying to modify the rewards, then marginal cost start creeping closer to marginal revenue ...
- dlubarov 9y ago
  You're right, I missed that, but > The coins will have to be rewarded on a completely effort-blind basis He's right that an effort-blind system is impossible, but that doesn't matter as long as the expected costs of influencing the block reward exceed the expected benefits. And it is is fact possible to design a system where the expected costs will always exceed the expected benefits, so putting effort toward influencing the block reward is unprofitable. The profitability of a PoS grinding attack depends on the transaction fee, the minimum account age, and the certainty with future blockchain states can be predicted. In a well-designed PoS system, those variables will be chosen such that grinding attacks have negative expected value.
  - smallnamespace 9y ago
    > And it is is fact possible to design a system where the expected costs will always outweigh the expected benefits, This is the key point, and also why the author's analysis is much too broad: as stated, the argument applies to all social structures in general. For example, under the author's analysis, PoW itself should not be able to exist, because if $10 of economic value is generated by the blockchain, then at least $10 of energy needs to be burned in order to keep the blockchain secure. This is wrong; it ignores the high real-world cost of dispersed attackers to acquire the (dispersed) surpluses from an attack. Yet we see that in practice social structures exist and produce net surpluses greater than their maintenance cost (in fact, these two facts are equivalent, since a social structure that did not provide an energetic surplus would not survive in dynamic equilibrium). The ability to punish rule-breakers as a large multiple of the magnitude of their attempted rule-breaking is what allows social structures (including markets) to gain the necessary leverage; otherwise there can be no surplus derived from creating structure, because to get $1 of social surplus, you need to burn $1 of surplus. Therefore, the real question is whether dPoS creates structures with higher leverage to punish cheaters than PoW.
CryptoPunk 9y ago
The Ethereum wiki is wrong on this point in my opinion. Even disregarding grinding attacks, the value diverted to coin purchases ultimately has a cost in goods and services. It diverts economic activity to non-economically productive activity in cycling capital into and out of deposits. I think Proof of Stake could potentially be better than Proof of Work, but the point about cost being equal across validation methods is correct in general in my opinion. There are specific circumstances where it is not true, like if producing the mining resource creates negative externalities. Where I think the article is wrong is in neglecting other aspects of consensus algorithm efficacy, like the potential security benefits from Proof of Stake totally aligning the incentives of owners of mining capital (which in the case of PoS is the network coins) with the success of the network.
- dlubarov 9y ago
  > the value diverted to coin purchases ultimately has a cost in goods and services I see what you mean, but that opportunity cost applies to all currencies. Every dollar we hold in currency (be it fiat, PoW or PoS) could have been invested in productive businesses. People tend hold substantial amounts of currency anyway since it's convenient. If we assume that people are going to hold about $10 trillion in currencies anyway, then it makes sense to ignore that opportunity cost when comparing PoS to PoW. > Proof of Stake totally aligning the incentives of owners of mining capital (which in the case of PoS is the network coins) with the success of the network. I agree -- aligning validators' and stakeholders' incentives has some nice benefits. It means we can have validators vote on parameters like minimum transaction fees, and the outcome of the vote should reflect what's best for stakeholders as a whole. In a PoW system that wouldn't work well -- the miners would likely vote for minimum fees that were higher than required for secure validation.
  - CryptoPunk 9y ago
    >>I see what you mean, but that opportunity cost applies to all currencies. I don't mean that. I mean the actual process of converting liquid wealth into/out-of ether staking deposits, and of sacrificing liquidity to maintain stake deposits. That is not a free activity. Given the earnings accrued to anyone with staking deposits, there will be competition to increase the share of one's wealth that are held as staking deposits, and that requires increasing frequency of these activities. The cost of all of these activities will approach the revenue generated by holding ether.
api 9y ago
> there is simply no way for anyone to send $99 to try to get at that randomness. Sure there is: black bag your house and bug your laptop, or remotely compromise it. If the original article's argument is basically correct then it means that for any alternative to PoW such as PoS compute-grinding work may be replaced by attempts to game the system or attack the system's security. I'm also skeptical of the claim that PoW is necessarily more expensive than fiat currency issuance. To account for the cost of fiat you must include the cost of all the security, administration, and governance that is required to enforce the integrity of the system. A dollar costs a lot more than the paper it's printed on (or numbers in a database).

zby 9y ago

The quoted text does not support your thesis - come on! It admits that C=R might be true - but argues that this is not enough.

dlubarov 9y ago
I see how you could read it that way based on the first half-sentence, but if you read the rest, it's saying that C=R is not always true.

middayc 9y ago

> then there is simply no way for anyone to send $99 to try to get at that randomnes can't I send one of the 10 people $99 today, and he pays me out in 10 days? Can't that person rent the fact he/she is on the list of 10 out. I am new to economic terms, but it seems your example even better shows his MC_rent + MC_nonrent = MR. Where he says in p2p systems MC_rent should be 0, because it's exclusionary. Again .. your example's main feature is that it excludes all but 10 people.

dlubarov 9y ago
> can't I send one of the 10 people $99 today, and he pays me out in 10 days? But if he already received the $100, why would he trade his $100 for your $99? Are you saying the recipient might want $99 now rather than $100 later? Let's say that the donor doesn't give any advanced notice; the recipient doesn't know they've been selected until they see $100 appear in their bank account.
- roenxi 9y ago
  The example given is a bit fuzzy about what it is showing and what assumptions are being made. This example is pretty bizarre because we have actors behaving economically irrationally (the $100 provider) and beneficiaries (I'll call them listees) who don't seem to have volunteered. I think it might be meant to simulate the creation of new money, but I don't accept the situation as being similar enough to be sensible (because no money is being created). Mr-give-away-$100 is burning resources (his cash reserves) to create an asset for 10 people (the potential to be given money randomly in the future, which is a strange financial asset). That asset is absolutely saleable on a secondary market and will be priced at some value south of $100. We can deduce who is on the list by observing who gets the $100 payments and deduce the probability after a few days of observation. The listees have incentive to provide this information as asset + secondary market is better for them than asset + no secondary market. Upper bound: Nobody is going to pay >$100 for a promise of $100 (I lie, that does happen, but only in rare and extreme circumstances. We'll ignore those). Lower bound: If the listed individual needs money NOW for some reason, which does happen in practice, then they may be willing to sell at any price >$0. Mechanism: A contract. They work well for events that happen in the future. Motive: The listees are giving up a small amount of value to smooth out their income stream. Results: When the $100 is transferred out of Mr-give-away-$100's account it will immediately be transferred to someone who bought it for some value $100 > X > $0; at a guess probably in the high $90 range. The person who was gifted a free asset will get rich, but that can be realised before the actual payment is made.

lalaland1125 9y ago

This is actually addressed in the article. See section: IS A “WORK-INDEPENDENT” PROTOCOL POSSIBLE? > If any cryptosystem is to periodically release coins, without immediately creating an incentive to “waste” an amount equal to the value of those coins, the cryptosystem is going to have to release the coins in a manner which is totally independent of all possible human activities. The coins will have to be rewarded on a completely effort-blind basis. The coin-reward must have a Spearman correlation of zero with everything that mankind could influence. If you can make it so that there is no way for human influence to affect who gets the coins, then of course marginal revenue will not necessary approach margincal cost. However, it's quite hard to design a system that doesn't have at least some way to influence who receives the rewards. And once you have some way of paying to modify the rewards, then marginal cost start creeping closer to marginal revenue ...

dlubarov 9y ago
You're right, I missed that, but > The coins will have to be rewarded on a completely effort-blind basis He's right that an effort-blind system is impossible, but that doesn't matter as long as the expected costs of influencing the block reward exceed the expected benefits. And it is is fact possible to design a system where the expected costs will always exceed the expected benefits, so putting effort toward influencing the block reward is unprofitable. The profitability of a PoS grinding attack depends on the transaction fee, the minimum account age, and the certainty with future blockchain states can be predicted. In a well-designed PoS system, those variables will be chosen such that grinding attacks have negative expected value.
- smallnamespace 9y ago
  > And it is is fact possible to design a system where the expected costs will always outweigh the expected benefits, This is the key point, and also why the author's analysis is much too broad: as stated, the argument applies to all social structures in general. For example, under the author's analysis, PoW itself should not be able to exist, because if $10 of economic value is generated by the blockchain, then at least $10 of energy needs to be burned in order to keep the blockchain secure. This is wrong; it ignores the high real-world cost of dispersed attackers to acquire the (dispersed) surpluses from an attack. Yet we see that in practice social structures exist and produce net surpluses greater than their maintenance cost (in fact, these two facts are equivalent, since a social structure that did not provide an energetic surplus would not survive in dynamic equilibrium). The ability to punish rule-breakers as a large multiple of the magnitude of their attempted rule-breaking is what allows social structures (including markets) to gain the necessary leverage; otherwise there can be no surplus derived from creating structure, because to get $1 of social surplus, you need to burn $1 of surplus. Therefore, the real question is whether dPoS creates structures with higher leverage to punish cheaters than PoW.

CryptoPunk 9y ago

The Ethereum wiki is wrong on this point in my opinion. Even disregarding grinding attacks, the value diverted to coin purchases ultimately has a cost in goods and services. It diverts economic activity to non-economically productive activity in cycling capital into and out of deposits. I think Proof of Stake could potentially be better than Proof of Work, but the point about cost being equal across validation methods is correct in general in my opinion. There are specific circumstances where it is not true, like if producing the mining resource creates negative externalities. Where I think the article is wrong is in neglecting other aspects of consensus algorithm efficacy, like the potential security benefits from Proof of Stake totally aligning the incentives of owners of mining capital (which in the case of PoS is the network coins) with the success of the network.

dlubarov 9y ago
> the value diverted to coin purchases ultimately has a cost in goods and services I see what you mean, but that opportunity cost applies to all currencies. Every dollar we hold in currency (be it fiat, PoW or PoS) could have been invested in productive businesses. People tend hold substantial amounts of currency anyway since it's convenient. If we assume that people are going to hold about $10 trillion in currencies anyway, then it makes sense to ignore that opportunity cost when comparing PoS to PoW. > Proof of Stake totally aligning the incentives of owners of mining capital (which in the case of PoS is the network coins) with the success of the network. I agree -- aligning validators' and stakeholders' incentives has some nice benefits. It means we can have validators vote on parameters like minimum transaction fees, and the outcome of the vote should reflect what's best for stakeholders as a whole. In a PoW system that wouldn't work well -- the miners would likely vote for minimum fees that were higher than required for secure validation.
- CryptoPunk 9y ago
  >>I see what you mean, but that opportunity cost applies to all currencies. I don't mean that. I mean the actual process of converting liquid wealth into/out-of ether staking deposits, and of sacrificing liquidity to maintain stake deposits. That is not a free activity. Given the earnings accrued to anyone with staking deposits, there will be competition to increase the share of one's wealth that are held as staking deposits, and that requires increasing frequency of these activities. The cost of all of these activities will approach the revenue generated by holding ether.

api 9y ago

> there is simply no way for anyone to send $99 to try to get at that randomness. Sure there is: black bag your house and bug your laptop, or remotely compromise it. If the original article's argument is basically correct then it means that for any alternative to PoW such as PoS compute-grinding work may be replaced by attempts to game the system or attack the system's security. I'm also skeptical of the claim that PoW is necessarily more expensive than fiat currency issuance. To account for the cost of fiat you must include the cost of all the security, administration, and governance that is required to enforce the integrity of the system. A dollar costs a lot more than the paper it's printed on (or numbers in a database).