Topic: [ANNOUNCE] The Proposal for EnCoin - page 4. (Read 9455 times)

And you continually ignore unsunk costs. Did I not bold enough stuff for you? Have you no concept that one instantaneous flash of the network is not how it will remain forever?
You can build your 20 rig 5W mini-machines, but it's gonna cost you 2 grand. That 2 grand is as effectively put into the market as every watt of electricity. And in the years of you paying off those unsunk costs, standard computer hardware will become more efficient than it was, and you will eventually no longer see any return whatsoever as your fixed mhash/s is dwarfed.

If a significant number of peers are colluding to lower the cost-to-produce, the other peers still working benefit as well (and by more as I showed in the 1mh/s vs 2mh/s example). This will encourage even more people to start minting coins at full blast because there is a profit to be had. No higher power halting-solved computer intervention is required. And no lowering the difficulty means that this problem could only occur over a significant time frame or from the beginning of the network. And if it happened from the beginning of the network, competition will make sure that the market value is what it should be (and if it's not, more people will mine!).

I can only believe you are being intentionally dense now and only selectively quoting what suits you. That, or you have absolutely no sense of economics.

You are correct, in your edit. It is still the same problem even with different starting conditions.

I guess it doesn't matter what "best hash value" means, since you claim it is equally distributed "per motherfricken block".

You claim the supercomputer would dominate because it did more hashes than the other machines over the same period of time. To rectify this problem, you want the super computer to do only the same amount of work as the others, or give away its excess effort to the others. We both agree that the super computer wouldn't want to do that.

You claim, therefore, there will be no supercomputers.

I claim, if the supercomputer is more efficient in hashes/W, it will simply just do the average hash rate for multiple networks simultaneously. Thus avoiding your penalty. I could run a hundred nodes on 5W plug computers doing all the network stuff. Then have each plug outsource the hashing to one supercomputer in China that works cheap.

You can't detect this. Nor can you prevent it. Nor as it seems, can you see it.

This is called an "isolation" attack. Yes, it should be prevented if reasonably plausible, but it is in a completely different class from the rest of the discussion. It fits more in with DNS spoofing, or fishing attacks.

If I can surround a "dumb" client completely I can lie to that client sure. But I still shouldn't be able to man-in-the-middle that client by redirecting his transactions to different accounts. At least not while I'm talking to others.

I think IRC is just making life harder for you. What IRC channel is it by the way?

I was just pointing out that the individual peer rep is what was required to join a net. Since the nets weren't network-managed, I needed some way to get people to make new ones. Again, it's a proposal, not a white paper.


It isn't over time, it's per motherfricken block. Each net works to make as many blocks as it can, just like they would in a bitcoin pool. But instead of breaking it up into smaller proofs of work, freenets can simply use a fixed header (we're not securing a block chain!) with each member starting with a random nonce. Whenever someone finds a valid target, the block containing the value and payouts is created based on the best hash value for each member in the freenet. If it were a share system, the supercomputer will dominate time and time again. With a payout structure, the maximum value per block is set. You can't get paid more by going faster without also paying more to each member of the freenet.

edit: the header will probably be based on each user's public key so that this can't be cheated. Either way the target is still the same.


If everyone had the exact same mhash/s and one guy is 40W more efficient than average, he makes a whopping $3.45 additional profit over the course of a month, assuming 12c/kwh. WOW! If he spent $200 on a fancy new GPU that he had no reason to buy other than to be more efficient, he will take 57 months to break even (and even if he bought it for gaming, he can't use it while he's mining!). During that time, everyone who was less efficient than he will probably have upgraded their computers so that now he has no efficiency advantage anymore. If you can get every single person in the network to collude on using less electricity in lieu of hashing power, then the value of ENC would gradually lower over time based on how many ENC already exist. If you can't get every single person in the network to collude then users will see that if they run full blast, they make more money (and everyone else less!), thus bringing up the difficulty over time and requiring everybody else to match or get left behind.


Either clients are required to have every single transaction or they're not (I prefer not). If they're not, the 51% could take advantage of those by giving themselves money from other accounts without those clients being the wiser, if they are only connected to bad nodes. Yes, it's a stupid attack, but I put it up there because of a discussion in IRC that concluded "lol it relies on trust".

I couldn't agree more with your re-definition of "private key". The section that said distribute the group's private key was a crypto mistake. In crypto a private key really means one and only one person. Once you give it to someone else, it is no longer serves its crypto function.

That said, I was under the impression that everything else in the PB was both deterministic and validated by every other group. (In the same way that transaction confirmations are.)

FreeNet Block minting allocation rules are based upon the previous PB so other groups should be able to detect invalid transactions. Since they are signed by a particular malicious peer, he should get the death penalty network wide.

Reputation Block seems to be deterministic. Did the peer vote "present" during the PB. This is presumably signed so it can't be forged.

Miscellaneous Block seems deterministic as well. If there were module change majority votes. I'm assuming the votes would be signed to prevent forgery.

That just means you need a way to prevent one person or even a 51 majority from prematurely ending the PB reconciliation before everyone gets to submit their sets of data to the Network wide union.

This is again what I was referring to as the problem of controlling time.

OK, you are going to have to acknowledge that the version I started with stressed at the beginning at some human paid to start a TrustNet. Got the privilege of naming the TrustNet. And could invite his (presumably Trusted) friends et. al. to join his new TrustNet.

In that scenario, it doesn't make a lick of sense to say, Hey Bill let me tell you about EnCoin... Fire up a peer, join someone else's shitty TrustNet until you build up enough Trust to join my TrustNet, then maybe I'll Trust you enough to let you in.

Really, when I read it, I figured it was a caveat to keep the rabble out of the Network you had paid money to start and were busy investing money in building.



OK, I did read "best hash value" as something figuring into the ranking. But ranking hashes didn't make a lick of sense. I presumed you wanted to do some sort of non-linear distribution (value only gets 20%) based upon each peer's Trust (Reputation/Placement/Rank).

You appear to have been serious about what I called non-sense. So let's discuss it. How could you possibly rank random numbers over time? If I have a supercomputer and I generated one block. You have a busted old 386 and you generate one block. Then the hashes of these blocks have are randomly distributed across the difficulty range. You can rank the blocks based on this randomness and say the 386 wins this time. But the overall distribution is still 50-50 as we repeat this exercise.

The goal here seems to be to keep the supercomputer from gaining a minting advantage. So just say all member of the group participating in the PB get an equal share.  Maybe you mean minting members. I'm not sure sure. But that's not important. If the distribution isn't equal, and it isn't based on RP, what is it based on?



I don't know if you meant that as a light tease, but I'm going to take it that way. Then I'm going to work some math to show you are wrong.

Say you are at a difficulty level with 31 leading zeros. That means you have a 50% chance of generating a block in 2^30 hash tries. Sure someone could generate the block on their 1st try or it might take all the way to the 2^31st try. But it averages out to 2^30 Hashes. Note: a block's units are hashes, not hashes/second.

So if m=2^20 then,
1mh/s will take 2^10 seconds (on average) to complete a block. 17 minutes, 0.28 hour.
2mh/s will take half the time, 8.5 minutes, 0.14 hour.
3mh/s will take one third the time, 5.66 minutes, 0.09 hour.

3mh/s for 200W * 0.09 hour = 18.0 Wh/block
2mh/s for 220W * 0.14 hour = 30.8 Wh/block
2mh/s for 180W * 0.14 hour = 25.2 Wh/block

If each of the peers (on average) gets an equal share for doing equal work. Then the most efficient in electrical cost (Wh/block) Wins! Again, electrical efficiency has nothing to do with (h/s).


This seem to acknowledge you recognize what I'm saying as true. But you are ACTIVELY trying to make the more electrically efficient subsidize the less electrically efficient.

It can't be done.

All your logic presumes that someone who is generating a block every 5.66 minutes will, WILLINGLY do more work than those generating a block every 8.5 minutes. This is a bad presumption. Each peer has access to the historical record. Each knows the exact the results of his fellow peer's productivity. If an efficient peer endeavors only to be average in blocks, he wins in DOLLARS!

There is a whole famous book about why your philosophy is a bad idea. Give it a quick read. It's awesome!

5-2
"Peers gain reputation in a similar manner as FreeNets: firstly, be apart of a FN that creates a block; secondly, the peer must be present and sign the new Primary Block at the time of creation—and a point of reputation is gained. Peer reputation is used as a minimum requirement to join more reputable FreeNets. "


8-3
"While the payout structure itself is TBD, the biggest, fastest hashing machine will not be rewarded as much as they would hope. It will be based on a set percentage given to the final placement of each peer’s best hash value. For example: you could throw the biggest super computer in the world in a FreeNet, but if the winning hash value only gets 20% of the pot, this supercomputer is subsidizing everyone else in that FreeNet.

So this encourages people to simply use their normal, everyday computer for making coins. There is little incentive to build a massive, 4x GPU machine to mint coins faster. It will not benefit you much, and the costs of the hardware may take years to recoup."

I realize there is a lot to take in, but this did have its own sub-section under the section "Energy Equilibrium", something that might sound like required reading before arguing that the system has major flaws. Someone getting 3mh/s for 200W is going to be subsidizing someone getting 2mh/s for 220W, and in the end this cancels out someone getting 2mh/s for 180W.

Anyways, one of the reasons I want to randomize freenets more is so that there can't be a freenet of just 190W people or one supercomputer and a bunch of duds. You can't know who you're up against, so you don't have data to manipulate the system.


Which is somewhat why I want to switch to this system. The freenet rep no longer has any bearing except minimum user rep to join. If a user does something irrationally bad, they get like -1000 rep and whatever coins they have on that account. -1000 rep so they can't join other networks with the same wallet. Not that it matters since they can just use a new wallet, but perhaps an IP address can be added to that list. And -1000 so that it can eventually go back to 0 so that it doesn't stay on the list forever and waste bandwidth.

Also, in the revised design, I'm planning on having it so that only one user signs anything for a set period of time (I'm thinking 3 hours) with a private key only they know. Which user is determined randomly each time so that there could not be collusion between freenets to try to sign an invalid transaction. If he tries to sign something the freenet doesn't agree on, they immediately notify the rest of the network. If he is disconnected for more than say, 2 minutes, he'll lose some rep as the FN needs to get a new member to be the signer and notify the other networks.

I'm also thinking the MerchNets will hide behind the freenets (I'm calling them corenets in the next proposal per your suggestion). As in, their IPs won't be publicly available. In the future, in a stable economy, merchnets should dominate the reputation. If most of their IPs are not casually available, it should be a lot more difficult to successfully DoS the network.


They could change whatever they wanted in the PB. Clients not connected to the network at the time wouldn't know. It's a very unlikely possibility, but if it did happen, there had to be a way to stop it. Having randomized freenets makes it an even more impossible situation. the "signer" could be evil, but all that will accomplish is wasting time to consensus.

For the MerchNets, I'm pretty sure they are going to gain rep based on how much in tx fees their account receives (and amount of time in a merchnet). The higher up in merchnet, the greater a % of the transaction fees refunded, but the more minimum amount is required to be in your account. Thus, do evil, lose account balance. muahahaha. Plus it creates demand by forcing merchants to save some money in ENC. Brilliant! imho

One problem I see though is merchants might not like users being able to see their account balances. I suppose the ability need not be added to the client software, but it would be easy to hack it in.
Do you think they will care? I suppose this applies to everybody.

So of course not I realize you are trying to PREVENT people from deliberately building particularly reputable FreeNets. I'm guessing this is to prevent 51% transaction confirmation DOS attacks.

I was never able to identify any other power a 51% majority gave a malicious group. You weren't worried about past posting, and falsifying transactions is still impossible. It doesn't seem to be a minting issue.

If I'm missing an issue, I really am clueless I guess.

But if you are worried about the transaction DOS issue I realized that such an attack is trivial to detect. If you see a valid transaction that doesn't make the next primary block. If that transaction is still valid after PB reconciliation, then it MUST be appear in the next PB or a DOS is happening.

Any way. I hope at least one thing was insightful.

Continuing on a slightly different point.

-----

FreeNet reputation seems intended to be a measure of a network's combined Reliability over time. In my mind, I define that Reliability as made up of two parts Availability and Accuracy.

Availability means at least one of the FreeNet's peers must be online to receive transactions, validate them, and to batch and send those transactions around. The goal for the FreeNet is 100% availability.

Accuracy means every transaction must be validated and confirmed exactly according to the crypto and accounting rules. Each FreeNet is required to be 100% accurate. If even one invalid transaction is confirmed, action must be taken.

So if a FreeNet is 100% available and 100% accurate for some period of time, they reach a max Reputation. I know you had said validation/confirmation errors would really hammer a FreeNet's reputation. I began wondering if it should be a death penalty. That seemed overkill at the FN level.

---

When I thought about these two concepts at the peer level I noticed a difference. Each peer must still be 100% accurate. But each peer is not required to be 100% available. Also, minting effort seems to be more important intraFN than between networks.

So if a peer solves minting blocks he might increase in rank. If a FN member peer is not 100% available he might lose rank. But if a FN member peer is not 100% accurate, the FreeNet either has to kick him out, or shut down. No other FN should trust a net that is even partially non-compliant.

If a guy is shopping for a new FreeNet, I'd want to know if he got kicked out of his previous one!

---

You seem to be seeing network membership as more flowing and flexible then me. I'm interested in hearing more about that. But this is the original picture I got in my mind when I read the proposal.

Clearly, I don't feel qualified to really come to terms with most of those details. But I can offer what (I think) are insights.

I always visualized your reputation idea as two layered. You have FreeNets which gain reputation as a whole. And you have individuals, who are ranked (for lack of a better term) by their FreeNet Peers. To make a lame example, I may be a "made man" in the Italian Mob. But if I quit and go to the Russian Mob, my rank doesn't necessarily travel with me.

I assumed in between when someone asked to become part of a FreeNet and when they were accepted into the FreeNet there was some sort of vote based on the existing FN Peers. I guess I'm suggesting that when you apply to be a member of a FreeNet. That the application contains an "rank" that you are applying for. If members don't think you are proven enough for that rank, they reject you.

To make an example, I might say "Yo! I'm currently ranked 100% in Fred's (90 RP 3 block/day) FreeNet. I'll quit Fred, if you let me join Ryan's (180 RP 4 block/day) FreeNet at 75% rank. That way, I don't have to start from the bottom again.

I was presuming, the FreeNet reputation dictated the total number of coins the FN could generate in a block. And the intraNet rank somehow affected how those coins were split among members.

IMHO instead of thinking about it all day long JUST DO IT so it can compete with SC2, Tenebrix and Fairbrix etc. as a cpu only coin etc. please 

This would be fantastic. Staying on one topic at a time would be an added boon. Makes it a little easier to follow which of the 4 different purported attacks we are discussing, and a bit less easy to claim one meant a different one when I have an answer. Just sayin'.

Before I start a new thread however (with rev3.1 that addresses the issue of this thread), I want to figure out a way to do the freenet reputation differently.

As I had it, users would gain reputation with the freenet, and freenets would require a minimum user reputation based on its own reputation. This has a few flaws. Users could be "grandfathered" in to a higher FN reputation bracket for one, giving them an unequal weight in the consensus.

I thought maybe a FN's rep could be the aggregate of its users, but this has flaws as well. As users come and go, broadcasts would have to be sent to adjust the network's opinion of that freenet. Could get spammy if there are a lot of users. Although having a minimum of 30 RP or so to be considered part of the consensus is one option to mitigate that. Additionally, it would be difficult to judge if >50% of the reputation has confirmed a transaction. If a lot of high RP people are missing (assuming we go by total rep of the network, whether present or not), >50% reputation confirmations would be impossible.

So I'm kind of thinking of network-managed freenets (this would also apply to the "mercnets" or whatever the merchant-based ones would be called). Say there are RP level networks in the range of 0-29, 30-59, 60-119, something like that (each one in each range is weighted equally). Whenever a 'net hits a certain number of users, it will force a split into two. Say at 80 users, it splits into two 40 user nets. If a 'net drops below a certain number of users, say 15, then the net's reputation is no longer weighed in the consensus (users could set the option to automatically join another 'net if this happens). This means no confirming of transactions, and probably no minting blocks either. Every 10 minutes or so, each member of a 'net would have to sign a transaction block so that other 'nets can be sure no funny business is going on.

New 'nets that are created from a split won't have reputation weight until after the next PB so that current 'nets always know how much reputation is required to confirm a transaction in a given PB. They will be able to mint blocks though.
'Nets that die and effectively disappear would still be required for consensus so that intentional or unintentional network splits have no bearing on whether or not a transaction gets confirmed. At the next PB, 'nets that don't show up will be "set aside" for a certain period of time, say a few PBs or so, so that if they do return, whatever happened between the two networks can later be reconciled--but in the mean time, they are not required for consensus. They will begin confirming transactions under the new setup (perhaps they wait 1 PB?), but with a warning to users that a massive split has occurred and details on why this might have happened.

Splitting a network would be controlled by a pre-determined, random function so that it could not be subverted. Anyone not conforming to this function gets reported.

What do you think?

Wasn't a baited question. I'd never given it two seconds thought.

It was a witty retort to your smug reply.

As for ideas. I defend your good ones. I defend no one's stupid ones. I don't mind starting another thread to actually discuss principles leading to a stable currency. Perhaps, someone who didn't reply to every question with, "It's obliquely referenced in my 28 pages of incoherent blathering about minting reform." would foster some discussion.

There is no stated goal of requiring 10kwh. It was a number chosen that was based on the assumption that the average computer would dedicate about 200W of electricity to making coins. Competition and market factors would determine the final price.
Bitcoin does quite well at achieving an average award of 50 btc every 10 minutes; it is little different from an average computer achieving 1 ENC every 50 hours. Lack of difficulty reduction is a simple answer to anyone trying to game the system.
And thanks for pointing out that no code is available.


Did you forget that there is an entire planet out there?


Baited question after baited question seems like you possess the mentality of a 12 year old! And the argumentation skills to boot! So quick to attack, yet so very few ideas. I wonder if it's jealousy.

That is the part I, and everyone else, have been trying to explain to him. That exercise is futile.



These effects are true, but they turn out to be temporary in the sense that the price will re-stabilize. It will not tend toward zero or infinity. It will simply pick a new level and continue from there.

The thing that made that side effect less abhorrent to me was the relationship of the price of electricity to the outside economy as a whole. Electricity is a government regulated commodity.

If there is a "glut" of electricity then the "fuel charge" part of your electric bill tends to go down. However, random fees and transmission line maintenance costs tend to rise. As with water, it doesn't matter how much conservation everyone decides to do. The gross monthly billing to consumers is never allowed to fall so far that it bankrupts the utility.

Conversely, electricity tend to rise in proportion to general inflation. If your electric bill goes up, the price of bread is up, and *hopefully* your salary will rise to match. (I know salaries tend to lag. Such is life.)



While I agree with you about this specific proposal. I agree with Etlase2 that stable money would be good. I also think using electricity as a benchmark worth further investigation and discussion.


But naming your currency as if it was an Enron product, seems a silly, silly thing!

Very complex design. Does not achieve the stated goal of making one ENC require 10KWh - nor could any cryptocurrency. Does not achieve the revised goal of making 1ENC require 50 hours of computation from an "average" computer. No code available.

Even if achievable, the goal of pegging currency to electricity is a bad idea from a macroeconomic point of view. Electricity price hikes choke the money supply increase clamping economic growth. Electricity gluts create inflation.

Sorry, but this design is a nonstarter.

Probably has to do with 3 pages of runaround arguments that are already covered in the proposal.

And En, predictably, stands for energy.

Anyway, good luck with it. I'm pulling for you to make it work. By the way, what does the "En" in EnCoin stand for?

I do appreciate you pulling me back here to talk about something that is not austrian economics, the wonders of the gold standard, or how hoarding bitcoins is just like buying stock in a new company! That used to be really tedious.

It was nice to hear someone else is a fan of stable digital money. Really, a year ago you could get shouted off the site just for mentioning it. It is a bit sad to see that you are not getting more interest.

So we're back to the hardware that doesn't exist again. k

gl commanding the marketplace with 20 enc a month instead of 15 or 15 for the cost of 12. you'll make millions

Oh, I really thought you were trying to say something insightful. None of my example said anything about running at 50% of the hash rate. That is some random fantasy you made up.

I'm saying that a 190W (2mh/s) machine is going to be significant. And I'm saying that a 200W (3mh/s) machine saves a full third. A 400W (16mh/s) GPU machine saves 3/4s, even if the 400W number looks bigger.

Each block at a single difficulty level takes on average 2^(D-1) hashes to find. Where D is the number of leading zeros. This has nothing to do with (/second). The block doesn't care if you do it in one second or one hour. If all of my peers are averaging 1 block/day. Then if I can average the same block faster, I have hours to save electricity by loafing. There is nothing anyone can do about that.

You seem to think you can average everything out and each peer's individual variance doesn't matter. Woot! Ignore it! Tell everyone it won't matter! Yell it to the world!

I promise to keep my mouth shut. I'll do the same hashes as my peers. Get paid same ENC as my peers. And win, because my costs are lower. I will command the marketplace. I'll cash out faster because I can take the existing "highest bid", but still end up with a higher ROI then those waiting for a "lowest ask" that will never come.

Woot! Trust me! You'll never know the difference. This is the last thing I'll ever say on the subject. I may even go back and delete any mention of it from this thread. Knight will be so proud!