Topic: Proof of Capacity as a replacement for Proof of Work. (Read 542 times)

Yes, I also agree with that. However, the reason why many think there is a Nothing-at-stake problem in PoC is probably the way how it's mined today. As there are only minor coins (Burst being the biggest but it's already a >top100 coin) existing now, not many resources are "burnt" to mine proof-of-capacity coins. So people can mine with their free HDD capacity, and thus the earned rewards are practically "free coins". In this situation, it is very cheap to mine on several chains and thus to attack the coin, cheaper than to mine a minor PoW altcoin. (One could call this attack "little at stake" instead of "nothing at stake").

This however changes when mining is carried out in an industrial fashion, as then the capacity to handle maintainance costs (HD failure, principally) become the factor which make a miner (and an attacker!) competitive in the mining game.

Why do you think PoC isn't a "competitive environment"? There is no guaranteed reward in current proof-of-capacity algorithms. The only coin I know that really had a "guaranteed reward" (but for transaction processing) was Timekoin, which mainly is defunct because it was too inflationary to gain any value (and also had other problems, as the double spending problem was not really solved).

I haven't thought much about the resource usage difference between guaranteed-reward and "competitive" concepts, because  "guaranteed-reward" concepts for coins are problematic from other points of view, mainly because of the difficulty to prove that you really participate, and the complexity/scalability once the full node number increases.

The NaS problem arises not on 2 separate chains, but on different branches of a single chain. If two blocks at the same height are found on top of the current block with the most work, then in PoW you must split your computing resources to work on both branches. But on PoC there is negligible cost to extend both.

Thanks for that explanation, but I am still somewhat confused why that would be. In the case of mining multiple chains with PoC, there is not a "Nothing at Stake" problem that I can see.  If you mine on 2 chains, you have to devote capacity resources to 2 chains, splitting how much you can earn. Only 1 of these chains will eventually win, and then all the value you earned on the second chain is lost. Had you devoted twice the capacity to the winning chain, you'd be much better off.  It is completely different than the PoS model, where there is absolutely no resource usage required.

What I am trying to figure out is if there is any reason why resource usage in a non competitive environment such as PoC (or indeed an Proof of Useful Work algorithm) is any different than resource usage in a competitive environment, such as the existing SHA-256 PoW, that allocates rewards statistically rather than on a guaranteed basis.  They seem to be identical to me from a mathematical point of view, but there is a group of people who insist they are not. The N@S problem is strictly a PoS issue, and does not enter into this particular question.

@DDAsics: Monsterer is referring to the "nothing at stake" (N@S) problem, which affects mostly PoS algorithms (but in his opinion also affects PoC). If there is no value at risk if you mine more than one chain, this opens the way to some attack vectors. In PoW, attackers always have costs when they want to carry out a double spending attack, but in PoS they can try so continuously, practically "for free". Also in PoC there must be value at risk to mine more than one chain, if it doesn't want to suffer from N@S.


It is not exactly the same "value-at-risk level" (to call it that way). But if PoC was carried out in industrial fashion the marginal cost of maintainance would become dominant in the "attack cost" calculation, while in Bitcoin/PoW systems it's a negligible fraction of the total attack cost. Maintainance costs are external costs, unlike in (pure) PoS where the "attack cost" is tied to the price of assets created inside of the blockchain. Maintainance costs go up in the case you try to mine two or more chains at once, so there is value at risk, albeit not the same amount than in PoW.

I am doing a deep dive into the psychology behind PoW algorithms. You make a very interesting assumption here, but I would really like to analyze this more in depth, because I think it is more complex than what you have implicitly assumed.

Can you expand on exactly "why" risk of loss is important during block generation? What do you think the result would be if PoW was not a competition, and instead everyone was guaranteed to be allocated a reward for doing work, rather than only a statistically probable percentage of a larger reward, which effectively acts as a proxy for the same thing?  Seems to me those would be mathematically equal in the limit, and I am trying to understand the underlying psychology for this belief?

You are not the only one who has it obviously, so I really want to figure out if there is merit to it which I do not yet understand, or if it is simply a value choice because that is how it has always been done?

Re: your second sentence. The difference between PoC and PoS, is that capacity requires you to expend resources outside of the system you are securing. With PoS, nothing in the world could make an oligarchy give up control if they did not want to.  Thus, 51% of a PoC system would always be a temporary condition in the presence of a determined outside force, where as 51% of PoS would be permanent and entirely subject to the whims of those who control it.

I don't see where "risk of loss" really enters into the equation.

Would anyone care to expand on this and help me out?

That is not value-at-risk in the same way as electricity cost of mining; what you are referring to is the marginal cost of maintenance, which of course all hardware has especially regular bitcoin mining hardware.

Faster storage has better mining perfomance. Related to Burst implementation you have more chances to find better deadline (=best share in PoW) before others so you outperform slow miners and may virtually steal they part of profit. Actually this may be very good point for everyone to use the top hardware, higher infrastructure cost but more profitable.

Tape is useless as PoC (at least at current implementation), PoC mining requires random access to the data within a time range. For the same reason on-the-fly mining like botnets is also not effective, it can't prove capacity with large volumes, but just small chunks. So botnets still requires to use the same way as regular miners and they also need to use dedicated and filled storage for working PoC.  I do not feel and threat here from kind of botnets, they are not effective at all with PoC IMO.  Especially kind of floor level of the capacity may be defined in the Consensus to protect network, just kind of extra protection layer.

The one thing that I haven't yet understood about the PoC concept is the difference between storage that is slow vs one that isn't and how a system would cope with such 'bandwidth' differences.

Think of something like that AI Doctor from Captain America Winter soldier - an AI saved on hundreds and thousands of miles of tape. Okay maybe that's a very extreme and fictional example but technically that's space that could be plugged into a PoC system - how would something like that affect the efficiency?

PoC could be still more energy-efficient. I didn't investigate in depth, made only some general assumptions (about energy use in hard drive manufacturing):

- attack cost (achieving 50% of the validation power, e.g. hashrate) determines the security of the chain. Thus, to become equivalent of Bitcoin security, all PoC farms/botnets together would have to be so large that the attacker needs the same amount of money to achieve 50% than he would need to 51% attack Bitcoin.
- PoW mining cost is not much higher than energy cost (hardware costs are only a little fraction)
- PoC mining cost, in contrast, is more tied to hardware cost
- energy consumption makes up a relatively small fraction of hardware manufacturing cost (certainly less than 50%, probably less than 25%)

Thus, we could assume that PoC, for the same "attack cost" level, needs, at most, 50% (probably 25%) of the energy of PoW mining, but on the other hand needs more other resources. Because of that reason, I don't know which one is "better for the environment".

(I know Paul Sztorc's article, but I don't consider it convincing, because not all human labor damages the environment.)

Would need to be extremely large datacenters to avoid a botnet attack as everyone has a hard drive in their computer. At least with memory hard GPU mining a botnet would only mostly exploit CPUs instead of high end GPUs. 

Doesn't industrial fashion like mining undermine the whole reason of using PoC in the first place? What specific problem is being solved here?

Here are some basics for people to investigate before suggesting other consensus mechanisms -

https://blog.sia.tech/fundamentals-of-proof-of-work-beaa68093d2b

http://www.truthcoin.info/blog/pow-cheapest/

http://www.truthcoin.info/blog/pos-still-pointless/

I wrote some larger posts about that somewhere in the Altcoin section (btw, I would really like a coin-neutral "Cryptocurrency Technology" subforum, because this section is normally dedicated to Bitcoin). There is less at risk than in traditional PoW, but not nothing; the point is that if PoC was carried out in an "industrial" fashion like Bitcoin mining then the loss of failed hard drives would become a major factor.

Nevertheless, short-range nothing-at-stake can be a problem, but long-range attacks should be costly.

That is only a problem while "PoC hashrate" is low; if it was carried out in industrial fashion like mining, legit actors would operate such high capacities that botnets would not have a major influence.

What is the point of replacing energy + ASICs with hard drives ? Do you really want a single botnet to be able to 51% attack your coin so easily?

Anti-POW people haven't also immersed themselves in the topic, and deeply understand why "the higher the mining cost, the more secure the blockchain, the more valuable the coin".

I was aware of that Reddit thread. Sorry for the long time to make the reply, I was on vacation. I will try to get nixops and answer you in here. I think some people who flamed you away didnt know anything too and didnt want knowing any weaknesses and limitations in Proof of Capacity lol.

PoC is not a replacement for PoW; there is nothing at risk of loss during block generation. At best it is an alternative to PoS.

You just want clicks in your video, or do you want a discussion?

I hate those posts that have zero information and just a link. Why don't you write a few words about it?
Hardly any veteran user is going to open this link to discuss anything...

I recently asked some questions on grinding attacks in

https://www.reddit.com/r/burstcoin/comments/a98tj4/grinding_attacks/

but got no satisfying answers. One replier got quite upset when I suggested that proof of capacity should not be proof of disk bandwidth.

This is nice , I always wanted burst to grow , so much unused storage, so much energy savings.

Daniel, nixops, Jones did a nice presentation about Proof of Capacity in this video. It starts in the 48 minute mark.

https://youtu.be/OoxIslpOiOg

Post your question if you have one. I will ask nixops to reply to all of them in the thread. Ty.