You can only do this if you delegate authority and trust, in other words, if you build hierarchies with power and enforced trust.  The DNS system is a hierarchical system, where the top nodes have a lot of power ; but in a no-double-spend system, this power becomes financial.  If you need a hierarchy of trust that you will get the correct information that Jack, who is trying to pay you, has NOT double-spent, how can you know that this hierarchy is not colluding to make you think that indeed, he didn't, while in fact he did ?
How can you be sure that one is not HIDING this former spending of his, the time that you accept the payment ? 
How can you check that nobody is putting more coins in circulation than is officially announced ?
If you have to trust specific hierarchical entities to tell you about that, the decentralized and trustless system is out of the window.

The whole problem of a truly decentralized and trustless system, is that you are not to depend on ANYBODY to be able to check the validity of the payment one proposes you, and the amount of coins in circulation.  It means that anyone, at any time, must be able to check this independently if he wants to.  This is not possible if there is a hierarchical system in place, because in such a system, you are DEPENDING on these centralized authorities, that can tell you whatever they want.

Yes, you can think of a system with different hierarchies of COINS, where you have a master coin that is the reserve currency of master nodes, who each of them are in a constant exchange rate with sub-coins of a different nature which can themselves be the reserve currencies of still other nodes with sub-sub coins.    However, if there are random payments from sub-sub-coin A of subcoin B, to owners of sub-sub-coin C of subcoin D, then, if you want this to be trustless and distributed, the users still need to have all these chains, to be able to check the right-to-spend of coin A, the right to exchange to coin B, the right to exchange to the master coin, the right to exchange of coin D and the right to exchange of coin C.  Yes, you might think that you don't have to bother about subcoins F, H, J etc.... but even that is not true, because you want to check their quantities in circulation.

So in the end, if you want this to be trustless, this is just a different way of organizing the transactions, but in the end, you have to know all of them, if you want to check the total liquidity.

I have been trying these kinds of things for quite a while, and I'm coming to the conclusion that there is no real way to have a system that is truly decentralized and trustless, and at the same time, scales without having an increasing burden on the individual user, who has the choice between delegating more and more trust to central authorities, or having more and more technical costs.

Well, there IS such a system, which I'm favourable for, but I'm not sure it is stable: that is: many small *independent* currencies, with floating exchange rates, and connected through decentralized exchanges.  But my fear is that speculative forces will put a hierarchy into these currencies, bringing us back to the current situation.  Nevertheless, at least, that is a system where the burden per user doesn't increase with adoption: a user decides to use just a few small crypto currencies.  If he needs to pay another user, he has to find a way through exchanges to get his coins, with several intermediaries, converted into the tokens of his counter party.  That's clumsy, but at least, it scales, because the number of steps is logarithmic with the number of users.

You see, this kind of evading dogmatic answer is typical.  If *theoretically* non-mining nodes matter, and have power, you must be able to argument against my Gedanken experiment, by indicating how non-mining nodes, in that case, can enforce the 1 MB rule against miners and users.  If you can't indicate how that would happen, you can only admit that my Gedanken experiment has no logical failure, and hence, that even 99% of the non-mining nodes cannot impose anything, from the moment that the miners agree amongst themselves, and the users want to continue to transact.

The nice thing of that Gedanken experiment is that it separates the "user power" in the market, and the "non-mining node power", which is most of the time, like in your reply, confused, to attribute the power of the users in the market, to the non-mining nodes.  By clearly separating them, you can pinpoint where the power resides, and it is NOT with the non-mining nodes, unless you can technically explain how my example fails.


Well, your argument of authority fails, because Satoshi was of my opinion in fact.... But again, rhetoric arguments like those (authority) have no logical meaning.  Tell me simply how, in a hypothetical situation, where there is a backbone of miners, that essentially ALL agree upon a 2 MB block, and users, that only want to be able to transact, and exchanges, that only want to allow people to transact and get their fee on it, how 99% of the "non-mining node army" that would have decided to keep 1 MB blocks, is going to enforce its rule. 

If you can find a way, then you are right, these nodes can keep the users and the miners in check, and are the true guardians of the protocol.  If you have to admit that users will happily continue transacting with their light wallets, connected directly to the miners, who can happily continue to build the block chain, and all the disgruntled non-mining nodes coming to a halt and nobody is caring, then you must admit that they don't serve any power purpose, and certainly cannot keep in check the miners and the users.



Which he was.  The error he committed, if his purpose was to have a decentralized development (which is not sure), to make a heir to his central power: Core.

I'm trying to make see that no "avoid double spend" system can be like that, simply because if you want it to be trustless and decentralized, an individual user needs to KNOW, and hence, needs to be able to check for himself:

1) the rights to spend of every other user he could potentially receive coins from
2) the rights of creation/total balance of the system

and this knowledge grows with the network size, so there's no escaping for this user to have to learn more, update more, check more as the network grows.

This comes about because a "no double spend" system needs to be sure that something DID NOT HAPPEN, not that something did happen.  You need to be sure that someone *did not spend*, not that someone did spend.  The only way to know this, in one way or another, is to get regular updates on spending rights/balances/.... of other users.

This is entirely different for, say, a communications network.  You really don't care that two remote nodes had a communication.  It is no burden for you.  

This is somewhat related to the Gibbs paradox in thermodynamics.  Suppose that you have two separate networks.  If these two networks are communication networks, and you link them together, then this new link doesn't put any burden on each of the participants on each of the subnets ; on the contrary, it opens up new possibilities of interaction between users of both sub nets.  This si why the internet is such a big success: hooking up a new subnet doesn't cost anything to other users.

But suppose now that these two networks are two different crypto currencies.  "linking them together" would mean, making the tokens of one exchangeable for the other one at a fixed rate (say, 1-1), so as to make them fungible.   What happens when we do that ?  Each user now has to have TWO block chains, somehow, united (but the data on both of them is now necessary).   Each user will now receive twice as much blocks/transactions .... So the technical burden on each user of each subnet doubles.

Well, it is interesting that you write "psychological" and not "political".  
There is indeed a very large political incentive to go for decentralized systems, because a centralized systems give a huge amount of power to the central authority.  However, in a system that is supposed to work on an economical incentive base (like "mining rewards"), the laws of economy apply, and I just wanted to point out that the "scaling wars" are using essentially economical arguments, like "it will be too expensive for people to set up nodes", and "LN nodes are economically incentivized by the fees that they can obtain".  If the system is supposed to work by economical incentive (and a monetary system usually will be !), then one cannot escape from economical laws.


Of course, but this hidden cost of central power will not offset the centralizing forces in a decentralized system, on the contrary.  The central forces, knowing they will benefit from this central power, will even be willing to "centralize at a loss", giving even more impetus to the centralizing forces.  So the *arguments* that a specific form of scaling would lead to economic incentives to centralize, are moot, because ALL scaling will have economical incentives to centralize beyond a certain size.

Indeed.  It is funny to see the parallels between the crypto hype and the dot-com hype some 20 years ago.
That said, out of the dot-com bubble did arise things, like google and amazon.  But pets.com didn't survive.
Whatever will come out of crypto, when all the dust settles, it is not what it was set out to be.

I wonder what is the origin of our inability to communicate.  I guess we both think that the other party is being deliberately of bad faith, but there must be another reason.

I'm telling that the peculiarity of a decentralized, trustless "no double spend" network, is that the technical cost to each user has to grow (essentially linearly) with the number of users on the network, contrary to other types of decentralized networks, like the internet, which was taken as an example.

This is because the big difference between a class of decentralized networks like the internet (where the cost can be almost constant per user when it grows) and a cryptocurrency type of network, where the trick is to know that there haven't been double spends, is that each user needs to know the spending rights of ALL OTHER users ; so the more users there are, the more each user, individually, needs to know, which augments the costs of storage, computing and bandwidth for each user.  The only solution is that users delegate more and more this knowledge to centralized authorities, that do this checking for them and which they have to trust.

I would think that that argument is quite simple, clear to understand, and evidently true, and indicates why decentralized systems like bitcoin cannot scale at constant user cost, and that "increasing the network" doesn't pay for itself, contrary to a network like the internet, where more users don't put a burden on existing users.

The fact that decentralized "no double spend tokens" networks have a cost per user that scales more or less linearly with the size of the network, and a centralized, similar network, can essentially work at constant cost per user, means that there will be 'centralizing forces' in a decentralized network, from the moment that the rising cost-per-user of the growing decentralized system crosses the fixed cost of a similar, but centralized system.

This also means that all scaling "solutions" will have in them, automatic economies of scale that push towards centralization of some kind.   The type of centralization will be different, but its effect will be equivalent.

There is simply no solution to scale a decentralized no-double-spend system beyond a certain size, because it becomes non-competitive with respect to a centralized version of it, towards which it will naturally evolve.

You see, that's a valid argument to say that if miners are divided, users can chose, but it is not an argument against my Gedanken experiment that tries to demonstrate, theoretically, the futility of non-mining nodes as a power structure that "keeps miners in check". 

You are now arguing about a fork by divided miners.  A fork is simply decided upon in the market, like ETC/ETH, and also has nothing to do with the "power structure of non-mining nodes".  So it is no argument against my statement that "non-mining nodes have no power at all, and don't keep miners in check". 

My texts are coherent.  But they are not polished, they are long, because raw.  The effort to polish them would be 10 times the time I take to write them (I type about as quickly as a speak, to the time it takes for me is about the time it takes to read them out loud).

I don't really need people to read them.  They help me formulate, for myself, arguments that would have remained implicit in my head, and by having to write them up with the idea that *someone* might read them, I am pushed to formulate them explicitly, which helps me get the logic, for myself, right.     At the same time, it is also a challenge, and if people are interested - sometimes they are - they can argue against my positions.  In as much as I understand and appreciate their rebuttals, I learn something, and I re-organize my own understanding of the whole system.  That's the utility of this forum to me: as a test bench for my own understanding.
My customers get my polished understanding, where I have to be much more consensual, careful and precise.  But that's not free.  However, I need to be aware of the potential arguments and rebuttals of certain positions.  That's what I try to get out of this forum.


You see, this is an empty argument.  My argument is technical and economical, because the entire question is economical: the cost of growth in a decentralized versus a centralized network, and that cost is technical, because the arguments put forward are technical/economic arguments: "the network/disk/computing cost of big blocks" for instance, and your argument that a decentralized network pays itself for its expansion.


Of course not, both are most intimately related, as you should be able to see when you read more than the first paragraph of one of my previous posts.  The *technical cost* of a centralized network always scales at least as favourable, and in most cases, much more favourable, than the scaling of the cost of a decentralized network ; simply because the cost *per user* of a decentralized (trustless transaction) network has to rise (often almost linearly) with the scale of the network ; while the cost *per user* in a centralized network can essentially be kept constant as a function of the scale of the network.



Well, if you read what I just wrote, you see that I argue exactly the opposite, so this is perfectly on topic.

You cannot compare this with "the internet" because the internet is not a trustless network where every node needs to know the balances (or other "still right to spend" proofs) of ALL OTHER USERS, which is inherent in a trustless "no double spend" system.

If we have a global internet mesh, when Joe is connecting with Jack, then Mary doesn't need to know anything about that.  However, in a trustless double spend system, Mary DOES NEED TO KNOW, at the end of the day, that Joe spent his coins to Jack, so that she doesn't accept them.  So the  more Joes and Jacks there are on the network, the more Mary needs to know.  While Mary doesn't need to know if there are many Joes and Jacks when it is about communication.  Mary doesn't need to know that Joe sent a big file to Jack.  Mary does need to know that Joe spent his coins on Jack.

This is the big difference between a communications network, and a trustless, decentralized "no double spend" system, and the reason why the technical cost, in such a network, has to rise, per user, with the size of the network.  While a centralized network doesn't have that problem.

I fully agree with that.  If several software implementations, independent one from another, and on "equal footing", would have arisen, then one would have seen the clear distinction between the protocol on one hand, and the software implementations on the other.  A bit like HTML on one hand, and mosaic, netscape, firefox, internet explorer.... on the other.

But Satoshi, the central decider of bitcoin, put up a Pope, his heir: core.  As such, this centralized force continued to exist.  When other elements in the power structure of bitcoin arose, and put their hegemony into question, people were surprised.

It is not a matter of blackmailing.  Most users (that is, the people buying and selling bitcoins against goods, services and fiat) use light wallets/onlline wallets, and only care about their transaction getting through, and couldn't care less about any politics and technical strategies of bitcoin.  Most exchanges want users to trade, exchange, deposit, withdraw, because they take fees on all of that.  So, apart from totally crazy modifications, they just want "a node that works", and will point their light wallets to whatever works.  The 5000 or so bitcoin ful nodes that do not mine, are insignificant with respect to the millions of bitcoin owners/users/traders.

So if the large majority of these 5000 nodes simply decides not to download the only block chain available out there, then it is simply as if they stopped working.  The users will point their wallets to the few nodes that still process and confirm transactions, that is, the miner pool nodes.  Exchanges will want users to be able to deposit and withdraw, so they better upgrade their nodes such that they follow the "live" block chain. 

I'm pretty sure that nor users with their light wallets, nor exchanges, are going to sit down, stop their nodes, and wait for an eventual block according to their rules, that may never come, and stop their business/freeze their holdings for that undetermined time, while others happily continue.

The remarkable thing is that at the same time, bitcoin reaches an ATL in crypto currency market share:

http://coinmarketcap.com/charts/#btc-percentage

At this moment: 61%. 

Well, that is the standard dogma, but I have challenged different people in finding a logical hole in the gedanken experiment that proves the contrary, and the only thing they come up with is what I'd call rhetoric, that is to say, non-logical arguments (like saying that I must be paid by someone, that I'm an idiot, that I'm wrong, that this and that such, but never a logical argument).

The closest come arguments that confuse non-mining nodes with users in the market.  But that's wrong, because users are people doing transactions, and they mostly use light wallets.  There are millions of bitcoin users, and only a few thousand of full nodes.  So full nodes are at most a promille of the user community, and are not representative of the users nor of their market weight.  As such, one clearly needs to distinguish the economic weight of users (who only care about the value of their token, and their ability to transact) and the "large majority of non mining nodes".  So non-mining nodes are not "the users", or "the market".  

My gedanken experiment is as I outlined above: suppose that all 20 important mining pools (and their miners) decide to make ONLY a block chain of 2 MB blocks, and that 99% of the non-mining nodes refuse that, but that the users only want to continue transact their tokens from their light wallets or online wallets, no matter what (to separate the effect of non-mining nodes from that of users).  As such, they will use whatever means they have, to continue to use bitcoin transactions, and don't care about any of these politics.

Tell me how those non-mining nodes are going to keep in check the miners and the users and force them to remain on the 1 MB blocks, if the users connect their light wallets directly to the miner nodes, and the miner nodes are connected directly amongst themselves.  We can also assume that exchanges cannot allow themselves to have a stopped node, as they would lose all their customers to the competition, so the exchanges are also going to use or a light wallet directly connected to the miner nodes, or upgrade their full node to the 2 MB protocol.

So, now we have:
- all miners making a 2 MB chain and they can continue do so because they send their blocks directly amongst themselves
- all users connect their light wallets directly to the miner nodes and send them their transactions
- exchanges are the less than 1% of non-mining nodes accepting 2 MB block chains and continue working
- all other (> 99%) non-mining nodes refuse these blocks, and want to "keep the miners in check, and menace them with a UASF"

Tell me how these non-mining nodes are going to succeed, which is your thesis, and would be the proof of the fact that non mining nodes keep the miners in check.

Once you finally realize (as Satoshi already told us in 2008) that "the only people needing to run full nodes are the people mining new coins", you might realize the futility of their existence in the power structures of bitcoin, and once you realize that, you might re-consider all arguments that are based upon the need of their maintaining in place and the compromises that go with that.

It is very simple.  Demonstrate, in the above Gedanken experiment, the full power of the network of non-mining nodes, keeping miners and users in check.

If there are N addresses on the block chain, the probability of this happening randomly (that is, not done on purpose) is of the order of N^2/ 2^160.

This is why, when N is of the order of 2^80, chances are high that it happened.  When N is of the order of 2^60, chances are of the order of 1/2^40 that it happened.

As others pointed out: not really.  

However, the possible number of addresses is not really 2^160, but rather 2^80, and in reality much smaller.  The reason for that is what is called the "birthday paradox".    If you chose 2^80 different addresses at random, out of 2^160, then you start to have a significant probability (several 10%) that two of them are equal.  Which would mean that the private key of the one, also works on the other one.  To make this probability small, one should be a huge factor below 2^80.  Say, 2^50 or so.  That's still of the order of one million billion addresses.  So it is reasonable to limit each human to a few tens of thousands of bitcoin addresses if we want bitcoin to last for a few generations.

This is in fact more limiting than one might think, if one also wants to adhere to "no address re-usage".  It would mean that each of your transactions, in your whole life, would need a new address.  Now that starts to be problematic, because if you do 10 transactions a day, you do 3000 a year, and in 3 years time, you've used up all of your address quota.

But then, we could allow for 2^60 addresses.  That would increase slightly the "collision probability", but it would still be very small (of the order of 1/2^40).  At that point, each human would be entitled to 1000 times more addresses, which would solve the issue.

The direct connection to a merchant is not even LN, "micro payments" exist already on bitcoin in a single channel.  The "lock in" comes when the block chain is too small to contain most transactions, and hence an on-chain transaction is difficult to obtain, and very expensive (huge fees).  If on-chain transactions are easy and cheap, there's no reason to go through the LN hassle.  So we start from the premise that settling is difficult and very expensive (this is the case if LN truly scales, and allows for hundreds or thousands of transactions per on-chain transaction).  As such, once you've set up your funds in a LN channel to a partner, you cannot really settle easily: all your transactions will have to go through that channel or you will have to pay a huge fee (maybe of the order of the amount you put in the channel).  So the huge price to settle makes that you are effectively locked in.  If an on-chain fee costs, say, 0.1 BTC, and you have 1 BTC in a channel, while an LN transaction costs you about 0.001 BTC, you better not settle.  If you connect and settle 10 times, your BTC is gone entirely.  So once you committed to a LN channel, you're essentially locked into it until you spent all your content.

It might even be that LN hubs and exchanges have exclusivity deals with miners, buying in advance their block room.  As such, as a non-institutional user, maybe you won't even find a miner that will accept your settlement transaction.  Who knows.  Or only at exceptionally high fees, if the chain is really full.

Which is as of today, essentially impossible in a decentralized system and fungible units of account.  And if you are talking about me as the dude, no, you didn't invalidate what I was saying there, and as you couldn't find any rational arguments, you started accusing me of being unfair in the discussion.

The point in that discussion was slightly different, but related to the point you raise here.  My argument there was that a centralized network (centralized, in the political sense, of central authority, but can be technically distributed, with nodes under central control) can ALWAYS perform technically as well as a decentralized network (which has no choice but to be also distributed), simply because the set of possible decentralized networks, concerning their technical implementation, is a SUBSET of the set of possible centralized networks.

By that, I simply mean that if you can think of a decentralized network that can handle a certain activity, then you can set up a centralized, distributed network with exactly the same topology and hardware-like investment ; which, at that point, can of course handle the same load.  But moreover, in a centralized system, one has much less of a hassle of solving the trustlessness and the consensus problem.  So every decentralized network can technically also be a centralized network (the nodes are simply under one single command instead of having different untrusted owners).  But a decentralized network cannot take on every possible centralized structure, as each owner needs to have his own node of some kind.  This is why the technical solutions of a decentralized network are a subset of the set of centralized (but eventually distributed) networks.

But on top of that, the task to be run by a centralized network, where all of its nodes are of course trusted, is much simpler than the task of the decentralized network that needs to use protocols to protect from trustlessness, and come to a consensus, which is not an issue in a centralized network.

An argument was that the "processing power/network of the decentralized nodes is for free" while the centralized system has to explicitly handle this.  My point was that the hidden cost per decentralized node is higher than the per-customer cost in the centralized system (exactly for the same reasons as above that show that the network set is a subset).

However, the larger the network, the higher the per-user cost in any case in a decentralized system, which brings us to the impossibility of your statement above.  The reason is that in a decentralized system with fungible units of account, one way or another, the whole network, every user needs to be aware of the balances of every other user in some way or another.  This is not the case in a centralized network, where only the 'central authority' needs to know that.  As such, it is unavoidable that the burden of communicating the balances of a growing number of users and a growing number of transactions to a given user, increases as the number of users increases.  
The burden for me to know what the 10 other people on the small network did, is smaller than the burden for me to know what 1 billion people did.  This is unavoidable.   One can only hope that the value of the large network is worth more to me, than the extra cost of having to know what all these other people did on the network.
A form of centralization can scale down the needed cost of having to be aware of all the other user's balances.  In a totally centralized system, there's no such cost to a single user.  I profit from a larger network at no extra cost.  The more decentralized a network is, the more my cost increases with the number of users.

The two centralisation solutions to, say, bitcoin are:

- only a few central miners.  Then my burden is not increasing: my light wallet acts as my banking app on my smart phone. The miners are the central authorities keeping the block chain and hence, people's balances, and I only need to be aware of my own transactions (light wallet).

- only a few central LN hubs.  Similar, if I can engage LN with a light wallet.  The central hub keeps the information of all its customers in all its channels, and I only care about my channel.

The standard way of solving this is:
- normal banks.  

So, a fungible currency in a decentralized system will always have a cost per user, rising with the size of the network (naively, it is linear, but one can try to find sub-linear tricks) ; while this cost will be constant in a centralized system and not rise with the size.

This is why decentralized networks seem to outperform centralized ones when they are small, and run into cost problems when the grow bigger, to become totally non-competitive when they dream of going mainstream, unless they centralize, which economic forces in the system will make happen in any case.

So, ironically, the structure that allows a network expansion "that pays for itself" is a centralized one.  Each user simply has to pay for his own burden on the centralized structure (with a fee), and that burden doesn't increase if the network grows.  In a decentralized network, adding a user increases the burden for every node in one way or another.

My understanding is that this is NOT how LN works.  There are no "LN notes" as far as I understand.  There are essentially only threats of broadcasting transactions that are exchanged.  As far as I understand, an LN channel is nothing else but a growing stack of mutual threats to settle in the disadvantage of the cheater that are NOT broadcast, together with a "last correct balance" transaction, that can be broadcast by any of the parties when he wants to close the channel. 

So there's not much of a chance to do fractional reserve banking that way, because there is no "IOU" (your "LN note") that is to represent a coin ; there is only a stack of mutually signed transactions one bricked into the other so that when broadcasting anything else but the last settlement, you automatically allow the counter party to take all your channel funds ; but without broadcasting them, they can't.

To the OP, as far as I understand, there is no "fractional reserve banking threat" by LN, but rather a "banking threat": namely that the dynamics of the LN is such, that it can only be beneficial to some central, big hubs to keep LN channels open, and that people have no choice but to lock in their funds to one of these central hubs.  As such, these central hubs can apply AML/KYC rules, they can require the payment of fees of all kinds, they can censor certain of your transactions, and of course, they now all about every of your transactions, which they can report to authorities.  But FRB is not part of it.

BTW, FRB is not applied any more in the current banking system, because there are no "reserves".  The central bank reserves are nothing else but buffers for inter-bank settlements, which can be augmented at will by any bank.

Coming from you, I take that as a compliment

Yes, this is an (educated?) guess of mine.  It was exactly bitpay's old charts that made me believe that at least until 2016, there wasn't much volume in bitcoin as a payment system.  Of course, I cannot really know what is the true volume of bitcoin for payments, but I know that all the volume on exchanges is mostly NOT for payments.  And I remember that the 2016 charts of bitpay indicated a minuscule volume as compared to the volume on exchanges.  Of course, I don't know how many people pay for stuff without going through such a payment processor.

https://qz.com/931810/cheapair-and-bitpay-data-show-rising-bitcoin-btc-payment-volumes/

In December 2016, there was a total volume of $169 million processed by bitpay.

That is less in a month than the DAILY volume of bitcoin on exchanges.  That puts bitpay at 3% of the bitcoin volume, and, if we follow Fisher's formula, the market cap sustained by "currency" of bitcoin, at 3% of its announced trading market cap.  Now, let us assume that bitpay is only capturing one half of the bitcoin commercial volume, we are at 6%.   Add dark markets, its principal commercial application, and we are maybe at 15%, but that is a generous guess.

I think that if the choice is between on chain and off chain transactions, if there are no limitations, everybody is going to prefer on-chain transactions, because with an on-chain transaction, you are free.  Your funds are not locked up in a channel with a well-determined partner.  It is an affair between payer and payee (and miner), and there are no "intermediaries".   I think that the reason why those in favour of segwit don't want larger blocks, is that they need an incentive to move off chain.  If the chain can handle it, there's no reason to leave the chain.  And if there is no hard limit feeding the fee market, there's no reason that these fees would be much higher, than the fees for those, locking up funds in channels with doubtful partners.