Topic: [SKY] Skycoin Launch Announcement - page 112. (Read 381579 times)

Honest nodes can move which pool they send their shares too. Unfortunately the larger the pool, the more consistent the revenue for the honest nodes. However, I suspect a similar game theory disincentive will arise in your model also, because influence = power = money.

I had become to get depressed about this and was thinking that decentralization is hopeless. Collective decentralization is not the natural order of how nature organizes. Rather what happens in nature is a plurality of centralized options (which is in effect decentralization), i.e. nature decentralizes through creative destruction. I've been trying to think about how to apply that to money systems, which is difficult because we need fungibility and volatility is eliminated through a common unit-of-account (a plurality of currencies is unnatural).

At this point, I am thinking that collective decentralized currency is a lie. It can't exist. It will always come down to trusting some parties, whether it be the developers or the pools or the influential nodes, etc.. This is the dirty little secret that all the experts know but don't want to admit publicly.

Thus I've started to think more about how the trusted parties can't be coerced by the government, i.e. anonymity. And less about the perfect decentralized consensus, which I coming to the realization that collective decentralized currency is a lie and an enormous waste of time.

I expended a lot of effort to design a hash function which can't be monopolized by ASICs (note it can be accelerated by an ASIC, but in theory can't be monopolized and that is a deeper explanation revolving around making PoW unprofitable for large economies-of-scale by leveraging the fact the home electricity is already budgeted!). I was trying to think of clever ways to decentralize the pools. But when I forced myself to think brutally realistically, I started to come the realization that I was fighting against nature and I would lose.

This is why I say I would like to do something more direct to the point of what is provable and less "pie in the sky".

Research is nice and fun, but reality is where the rubber meets the road.

P.S. If I realize I will no longer use my PoW hash, I will open source it so that everyone can see what I created. If I use it, it will be open sourced after launch (no current work on this, this is all shelved as I am working on other projects). Thus it will eventually be open sourced.

For bitcoin, its performance is actually considerabley poorer than our algorithm.
Top mining pool is parallel to top influential node in our algorithm.
So for the currently bitcoin network with 8000 nodes, it can't tolerant collude by 4 top pools(actually, many ppls suspect that GHASH put its computing power in the shadow to hide the fact it controlls over 50% of the total computing power), the fault tolerant is 0.05%.

Another big advantage is that when some node is suspicious, honest nodes can just unfollow it and then it loose the power.
But in PoW, top nodes can do anything and you have no measure to reject them.
So even after a successful attack, they can attack again and again.

Exactly. But there is a detail you might did not noticed previously, for each node the algorithm works like this: round #1: deciding-->round #2: deciding-->....-->round #50: **decided**.

A node only makes its finial decision after enough rounds, and before that, they are just deciding and decding and deciding, called "opinion synchronization".

Thus we need the trust relationships. Sybil nodes has no honest nodes following them and thus has no impact to the opinion synchronization process.

It's YES, during the *deciding* rounds, but each node has exact one chance in its last round to give its *final* decision.

You are correct I will need to spend more time studying this before I can argue or discuss constructively with you about the model.

My oscillation assumption was based on my assumption that localized consensus is not overruled by global majority in your system, i.e. I don't understand how the local trust capitulates to some global aggregation of the majority. Does the model allow a minority opinion to be overruled such that the minority adopts the majority opinion (e.g. for disagreement over which transaction was first in a double-spend)? My concern is that the minority would continue to uphold their autonomous disagreement because the consensus process appears to be local trust and not global overrule, and thus not global convergence. In PoW, the minority submits to the majority opinion, else it is a fork which the majority ignores.

Counting votes won't work, because that could be subject to Sybil attack.

Does the local trust morph as the majority emerges holistically? In other words, are the peers allowed to change their opinion? I couldn't see this mentioned prominently. It is a crucial point.

In my opinion, the explanation of this should be more prominent in the white paper. You jump straight into math and model details without giving an overview of how you've solved the problem.

I had read that. Isn't that much less % than a 51% attack?

I understand we can't precisely compare the two, since your model is a different structure from PoW. The complexity here is beyond my comprehension at this moment.

Note in page 5 of the paper: Note here even we follow the same naming convention with traditional mathematical optimization techniques, the meaning is totally different, instead, it’s only the driving force to producing emergent phenomena of a complex network.

Did you check the mathematical model? Actually, our final algorithm is a *compound* algorithm consists of the greedy and they simulated annealing algorithm.
(note we just use the same naming convention)

And even with *pure* simulated annealing alone, your assumpution is still incorrect. Mathematical model(also confirmed by simulation) tells that "oscillate indefinitely and never converge" will only happen when indegree/outdegree is not small e.g. degree>30. If degree=10, then it will still converge rapidly.

It's possible, of course. We assume a byzantine failure model thus any behavior is possible.
Nodes can even collude to commit attacks to mislead the consensus or prevent the whole network to reach consensus.
From simulation under the strongest attack, our alg can tolerant attack committed by 2% top influential nodes or 15% random nodes.

Incorrect. Read more carefully what I wrote the first time:


Apparently you read one sentence and ignored the rest of what I wrote below it.

I clearly indicated that I my disagreement was that capital controls would not be global and would instead be focused on Westerners, and I clearly pointed out that Asia was headed the other direction.

I started to skim (actually read most but very fast) the whitepaper and on page 6 it explains that the simulation did not globally converge so you added simulated annealing (random jumps to escape from local minima).

I studied machine learning in the 1980s, so some of this is familiar to me.

Afaics bottom-up, cascading trust relationships are not always going to be able to reach global consensus. If you insist on simulated annealing, then it can just oscillate indefinitely and never converge. Or you can accept that the network will converge on non-global groupings that don't agree with each other, e.g. a double-spend and they don't agree on which spend was first (because you said you don't want to depend on a delegated timestamp server). Perhaps I am wrong because I am not going to take the time to read and understand every detail of the paper (check the math, etc), but conceptually I am nearly certain I am correct.

And that is not even getting into the peers colluding due to game theory and/or political economics.

Just because you model it without game theory and political science, doesn't mean there isn't an economic incentive to game it and a political motive to organize the "voting". Models can tell you anything you want to hear, and are not fool proof until they've been vetted in the real world.

You are correct I can't comment forcefully until I read the white paper and understand it. But I don't think I need to read the paper to understand that peers vote by signing and appending to the directed graph. Any form of voting can fall victim to the Iron Law of Political Economics.

OTOH I will grant the possibility that you've some how obfuscated such that political organization and gaming the voting is impossible. But I would need to see how. I doubt it.

No, no game theory or political science are involved, in fact, the core algorithm is deadly easy that you can describe it within 15 lines of pseudo codes.
The heart of the algorithm is the *emergent phenomenon*, maybe it's a little difficult for one without background of physics or system science.
But if you coincidentally read one popular book called "Out of Control: The New Biology of Machines, Social Systems, & the Economic World"[1], you may have better idea of such a bottom-up, self-orgainized style.

[1] http://www.amazon.com/Out-Control-Biology-Machines-Economic/dp/0201483408

No, no global clock/timestamp server is needed. Any dependency on global server will break the decentralization.
The logical time is called *round*, each message has a round number, and round can be determined to be timeout *locally*, that's all.

Until you read the paper, but...it's long and need patience.
I think you might begin to read it **ONLY** if you REALLY have interest on it.

Okay but I think that then involves game theory and political science. And you didn't answer as to whether it depends on a delegated timestamp server? And that is much more complex than I have time to analyze now. For example the linked nature article refers to consensus on naming, but afaics there is no economic incentive to game it. And moreover, because I currently don't think that is the highest priority task, i.e. I don't want to build a "pie in the sky" coin.

But I don't discourage you. We need people doing all kinds of experimentation.

Thank you and apologies for interfering.

Why can't you understand the difference between disagreeing that capital flows will get cut off in every nook and cranny and agreeing they will get cut off for Westerners??

I disagreed with his pronouncement that capital flows would get cut off to such a degree that gold and silver would not rise in price. Even you don't agree with that. You think gold and silver will rise. So why are you agreeing with him and against me? I was saying that gold and silver will rise because not all capital flows will get entirely confiscated.

The capital flows will be confiscated, but not 100% to the point where only food becomes money every where on earth and gold and silver are useless. I am sure you agree with me. So I don't know why you are making trouble with me?


Please do not quote me out-of-context then accuse of saying capital flows will not get cut off. Here follows what I wrote, and I clearly stated that capital controls will not get entirely cut off, but also agreed they would to a large extent for Westerners:


Which I further clarified for you as quoted below, when you indicated your misinterpretation of what I had written above:

The account of "skycoin" is actually shared by more than one core developers.
I am another guy who wrote the consensus paper.

The consensus algorithm is cross discipline in some sense, the key words are "naming gaming", "opinion dynamics", "social consensus", "complex network" etc. and the topic is actively studied in the fields of statistics physics and system science. Most of the related papers are published in journals like "Physical Review E", "science", "nature" rather than ACM transactions.

here is one related paper: http://www.nature.com/srep/2014/140704/srep05568/full/srep05568.html
another one(considering attack): http://journals.aps.org/pre/abstract/10.1103/PhysRevE.84.011130

The basic principle is to drive the **immergent phenomena** of a complex network by *only local* information(no global clock or any other global information are needed).

Disagreeing with a quote that says "capital flows get cut off" surely looks like you reject that argument. Now you say you agree. So which is it?

Ok let's see then what develops... Thank you also.

Agreed and now is not the right time to cater to me. If my ideas or interests fit with what develops, I can come back at that time.

anonyMint, I welcome your thought processes, and as you can see in this forum - everyone else value it.
Skycoin IPO is geared to raising a modest sum of money to pay [in BTC] to coders, and a majority of skycoins held by OP to code contributors and hardware design perhaps, for this project.

To the best of my comprehension, Skycoin is so wide, that there are bound to be many elements (e.g. file sharing, mesh communication, anonymous messaging, personal blockchains, etc.) that will cater to your interests and vision... and since post-ipo, OP will have the funds, there is bound to be many building blocks to develop, where you can contribute and gain considerable rewards as one of the earliest significant contributes.

Please stay tuned...

I agree with chompyZ. I'd prefer you expend scarce time down your current path and it would take me more time to convince you (I mean I could never convince you to change your philosophy, you have to learn trying), than to code it all myself.

I did this discussion to illustrate to those who keep wondering why I don't work with others, how difficult it is to work with others when my thought processes are so out-of-sync with the way most people are thinking on these forums.

I don't agree that Ethereum is doing very well. I think they will be an enormous bust and they are just expending a lot of zealot money while receiving nice salaries, capital gains,and having fun attending conferences and socializing. The fundamental issues were never solved.

I have my own ideas and path to forge. Best of luck to you.

P.S. I do note you did not (yet) refute my technical skepticism about whether your algorithm can converge on global consensus.

And apologies for cluttering your thread. I'd prefer not to comment on any more altcoins. If anyone messages me about such and I don't reply, you know why. I don't enjoy making enemies or creating animosity.

---

No! Not even close to working together. We disagree. Your simplistic idea of my unstated design and marketing strategy matching his is ludicrous to me because I know we are not even close to synergy. There is no way to marry the two. He would have to mostly quit most of what he has developed up to this point and start over. There is no way he would do that.

Caveat being if he could prove to me that his algorithm guarantees global consensus. Perhaps there might be some way to incorporate all his existing code and design.

Skycoin: imho, just ipo and launch. Whoever wants to join and contribute thereafter is welcome. Don't try to reach 'consensus' with other coders regardless how talented they are. At this point, instead of propelling the project, it only stagnates it.
After the IPO, talented coders will be able to easily review the code, see how it works in practice, and tap into your vision. imho ONLY THEN it is reasonable for other coders to come up with concrete ideas and justification to replace any Skycoin procedure. As you claimed, Skycoin consensus is a 'container' that we can later plug-out and replace, so lets not waste valuable time.
No project has ever been launched and left as-is, untouched. There will be changes, modifications, new modules, and so forth.

At the end of the day- you are the project OP / lead dev.
Lets rock...