Topic: LN+segwit vs big blocks, levels of centralization. - page 7. (Read 8897 times)

wrong
if you get out of the mindset of "gigabytes by midnight" then home computing can and will continue to be 'node compatible' even without dev's dictating when the silver spoon of capacity increase should be used

EG
a zx spectrum /amstrad pc of the 90's is not todays skype capable.. so should social communications of the 1990's have been held back and be told to not even bother and instead have just one team of developers do all they can to sidestep the issue with half gestures

the 2009 1mb limit was about raspberry Pi minimal hardware requirements(below home desktop/laptop) and old internet averages.
8 years later things have moved on

we are not in the era of 3G internet, we are in 4G and approaching 5G
we are not in the era of bottomline ADSL internet, we are in Fibre and approaching widespread fibre

we are no longer in Raspberry Pi1 we are in Raspberry Pi3 era

all in all 8mb is deemed as safe and even so. 4mb is more than safe.

out of millions of users pretending that blocks would kill off 7000 full nodes is foolish
what you will find is that features like prunned/stripped nodes will kill off the full node count faster

An increased block size would not prevent centralisation.  It's a way of fitting more "decentralised" on-chain transactions, but as mining centralisation happens it won't matter that transaction are onchain.

Nodes would be centralised as well - more transactions means a bigger blockchain, which in turn means it's more expensive to run nodes and they become centralised.

It's impossible to avoid at least one kind of centralisation happening.

The link between the miner oligarchy and the LN channels is attractive ; but LN nodes need, in the first place, a HUGE STASH, because they need to commit coins to every of their customers links.  I'm not sure that miners are sitting on a huge amount of stash.  I think the link will be commercial: LN hubs (owners of huge stash of coins) will agree to buy exclusive room on the chain (they need this to be able to settle for sure).


I don't see how this can be the case, honestly.  After all, you can open your hub on just any server somewhere in the world, and nobody is to know who you are for real.  

But in fact, all of this is totally deluded in a way, because bitcoin is not a currency.  Crypto is not a currency.  It is a speculative asset, and this shows more and more with the current rise of alt coins.  Crypto, with its built-in deflationary spiral, is optimal as a kind of complicated derivative, and is not at all a means of payment.    So all this 'banking' delusion is never going to happen for real.  We're assisting to the definitive transformation of crypto into a huge derivative-like speculative market, with as a main purpose the greater-fool game and the trader's game of getting money out of the hands of a competing gambler ; and with a profitable industry around it leaching off fees from this gambler's game.  

Of course, as a small application, these speculative assets can also relatively easily be transmitted to others in exchange of goods and services.  But this is a very minor, almost parasitic, application of crypto.  Crypto is derivatives finance in the wild.  It has not much to do with "money to pay for stuff" like banking and VISA.

I ran across a post at btc-hedge.biz/?page_id=ScaleDebate which claims "we are already here" and I was shocked then.  Now one year on and after I read so many discussions and papers, I am sad to realize centralization in some part seems inevitable.  Unless, the internet is so much improved that a 5% or 1% lower percentile of p2p speed is high enough and 95% or 99% people are happy to increase the on-chain blocksize.

Two important issues raised here.

The most reliable of LN channel users will be the miners. They have highly redundant systems and the ability to pick and choose which transactions to settle first in a forced expiration spam event. They can take measures to protect themselves.
The other issue is if reliable LN channel users will fall under 'banking license' legislation. In this case, the cost of running a full node is the least of a reliable LN channel users worries. It is not only costly, but difficult to obtain a banking license.

How quickly you chose to sidestep the issue

Obviously, if we "run twice as fast" to develop "other solutions", the block size issue becomes a non-event, negligible or even completely irrelevant (provided we succeed at that, of course). But you didn't mention that in your original post, so I challenged exactly that case as you put it, i.e. when only the block size gets increased (in a "conservative way") as a solution. As to me, that wouldn't be a solution at all since that would be basically prolonging the commotion by essentially freezing it, primarily because the "centralization risk" you mention in your post is already unmanageable. I guess this is exactly what rogue miners are hoping for (if they can't just freeze "the lay of the land" as it is now)

I agree fully.  The problem with LN is not so much the "fractional reserve banking", but the fact that a user is tied to a banker, who can charge fees for his services, can report any spending and earnings, can require KYC/AML, and can give you a permission or not to transact.  Of course, you can always settle and free yourself from your banker... at least, if you can obtain room on the chain, which is maybe not free, but may also be sold out exclusively, if LN hubs agree on exclusive block rights with miners.

My idea is even that the above evolution is "natural".  There are no evil minds behind this, it is the natural evolution of the power structure of payment systems.  This is how we got here in the first place, not because a few men with cigars met in a room making evil plans, a few centuries ago.  Banking is just a natural emergent phenomenon in payment systems.

The fundamental problem, however, is that at that point, crypto has nothing to offer any more.  Traditional banking is then superior, because these are institutions with much more experience, trust, and legal backing.  In fact, the fiat system is then a better system than the 3-layer system with "guys on the internet", because there are rules, legal audits, legal recourse and so on, that "guys on the internet" will not have.  You can say what you want, but I trust my bank 1000 times more than my exchange.  I even trust my bank 1000 times more than the security of my wallet, or the Chinese miners, or the Core devs.

He did.  From memory, he wrote something like "only people mining new coins need to run full nodes".  (one of his first e-mails on the site of nakamoto institute).  


Yes.  To have a "decentralized" system, you need 3 non-colluding parties, none of them having more than 50% of the hash rate/consensus voting power.  So that is the bare minimum.  But you need to be sure that:

1) they don't collude (not even on non-agreed things but by independent judgement, like "we don't accept transactions from dark markets" or something of the kind)
2) one cannot make them "an offer they cannot refuse"

That's the hard part.


Thank you.  Someone else understands.

Decentralisation of what? That is the question.

Satoshi's vision was of a backbone of big nodes, with millions of SPV wallet users. It's quite possible he did not realise just how much the hash power would consolidate in the hands of a few actors, but clearly he didn't intend for everyone to run a full node. I don't know if he made any statements on the value of 'proxy' nodes.

I think 'decentralised p2p cash' probably meant freedom from the interference of the bankers and government. A handful of mining nodes without a diverse distribution of hash power ownership and location is clearly the biggest threat to this idea of decentralisation.

Proxy nodes distribution of the blockchain only decentralises the blockchain history, not its main operational aspects.

EDIT: Although under LN, proxy nodes may become stable LN channels. This is the real political motivation for keeping the blocksize small, to be able to claim LN is a decentralised solution with the carrot that stable channel users may earn routing fees. It's essentially a full reserve banking system through these stable channel users. Eventually, the settlement layer will have to have an increase in capacity and the number of stable channel full reserve bankers will consolidate. The conspiracy theorist part of my mind leads me to think scaling war 2 will require a 3rd layer solution, the users will have to move to an off chain BTC token backed by BTC reserves - the transition from BTC has p2p cash to an electronic version of fractional reserve banking will be complete. It will become a settlement system between BTC bankers, with users using BTC backed tokens at the top. It is useless to resist.

The maximum block size parameter is not the parameter that leads centralization, but for some or other reason, this myth must remain intact.  The technical burden (networking, block chain storage, CPU time....) that leads to centralization is simply *the size of the network*.  You can shift this burden from storage (block size) to networking/CPU to some degree, but the technical burden per "decentralized user" is simply proportional to the size of the network (the number of users/the number of transactions). 

The only way to make that burden lighter, is to centralize.  This is already the case, for all people using light wallets are not decentralized users, but depend partially on the full nodes they trust.  If there are millions of bitcoin users, and only 5000 full nodes, we already see how people have "delegated trust".

But the technical burden is absolutely not what has driven centralization in bitcoin.  Bitcoin's *economic* parameters, and especially its PoW, have introduced much, much more centralization than the technical burden of having a big disk and networking.  No matter of how much you guys are putting your head in the sand, non-mining nodes are not a contribution to the decentralisation of the network, because they only act as P2P proxy servers of the miner-produced chain, which comes from essentially 20 nodes.

Bitcoin at this point, is a 20-node affair (with 5 of them being majority).

I didn't know that.  That's outright ridiculous, because that means that most payments won't be able to use the LN.  LN would be the "coffee network" then.  If channels are only $60,-, and they need to carry the ratio between "VISA-like" and "block chain" amount of transactions, say, 100-fold, which means that every channel needs to transact at least 100 * (average number of hops in the LN) * 2 transactions before settling, the LN is supposed to have transactions of the order of maximum $0.3.
(if there are bigger amounts, this will settle channels quicker than the ratio between the number of LN transactions, and the available space on the chain).

I don't understand where you're disagreeing here. What I wrote basically is that the "maximum block size" parameter, which is crucial for the decentralization of the system, should not be increased in a non-predictable ("unlimited") way because it's probable it will lead to a more centralized node structure.

So the "conservative" way I'm advocating would be to rise the limit slowly, based on available numbers about internet bandwidth and CPU and RAM price evolugion ("Nielsen's Law", "Moore's Law" etc.) but slow enough to be able to lower it if it's too dangerous. (It's not important that it's explicitly BIP 103, but it could also be a BIP-100-based solution).

We can advance and "run twice as fast" in other areas than block size - for example off-chain/sidechain/child-chain/sharding/extension blocks solutions.

But it is much worse than that.  If you open a channel with your dad, and a channel with your wife, you need to lock in half of your funds with your dad, and half of your funds with your wife, and what can be transacted in one direction is limited to that amount.  Moreover, you have to be careful not to transact all the time "from you to dad", or you will exhaust quickly your channel.  So you have to make sure that one time, you transact *towards dad* and another time *from dad*.    If dad is on a "highway" and your wife is too, then your modest amount of locked-in coins will not suffice to do the link between the highway on which your dad is, and the one on which your wife is: you will quickly run out of funds and have to settle the channel, with expensive on-chain transactions.

However, if you are a rich guy, you will be able to put up a lot of funds to "dad" and to "your wife" and your channel will live a lot longer before being exhausted.   The settlements will be much farther in between.  As such, you will be able to have more competitive LN fees than the modest guy.

This is why the "economies of scale" in LN are essentially proportional to your stake.  Mind you, you do not get BENEFIT proportional to stake, no, your ability to compete goes with your stake.  That's the equivalent of saying that the *efficiency of a miner* would go with the hash rate he has.   Now, miners have about similar efficiencies, almost independent of their hash rate, but of course the more hash rate they have, the more gain they have.  With LN, the more stake you have, the higher the *efficiency* of your channels.

With such almost linear economies of scale, only the biggest hubs can be competitive in the LN fee market.  This is why if ever the LN is up and running and people are squeezed out of the on-chain transactions by the scarcity of the transactions and the height of an on chain fee (if it is not a matter of exclusive room), it will centralize much faster than the mining population centralized, because the economies of scale are much more important in the LN, than in the mining business.

I partially agree. 

I don't think the routing information will be monetized.  That will be freely available.

It's just that there won't be many efficient routes except through big hubs and most will stop trying...and it will further centralized. 
A mesh network of everyone connecting to everyone else requires everyone to have at least 2 channels open with unrelated parties.
I can't just open a channel with my dad and another one with my wife.  If everyone did that, no one would be able to connect
to anyone but their own family...so everyone needs to connect to 2 random people and then what do you do when you need to
settle?  Also, how long will it take to get critical mess for such a decentralized network?  It's not impossible but it seems very
impractical.  People will just use hubs.... so the whole thing is promised as this p2p solution and its not.

 

I'm also inclined to think so on the whole

But there should be strong theoretical basis to avoid such centralization. Remember, everyone could mine their bitcoins on their home computers, but we still ended up with just a dozen miners and heavy monopolization of the mining market. Besides, I don't think that the price is going to rise just because people will be using their bitcoins for setting up payment channels (i.e. due to supply squeeze). Most certainly, these channels will be organized primarily by those who are right now just sitting on their coins

Of course you have.  You have to know whether a given "right of spending" (in bitcoin's language, an UTXO) has not been executed somewhere.  From the moment that you delegate that knowledge to "someone", that "someone" can collude with the right of spending owner (or is the same one), and tell you that the right to spend still exists while it is not the case.

Also, you have to know how many "rights to spend" are in circulation, to verify that nobody is counterfeiting and debasing the whole system.  If Joe and Jack have been double spending for half a year, even if you never have anything to do with Joe and Jack or their coin history, they are debasing the system you're using, which is a good thing to know.  The only way to know that Joe and Jack are not double spending, is to have a cryptographic proof of that.

If you give up on those two issues, you have given up on decentralization and trustlessness.  


But this is really not the problem.  DNS is totally different, because it gives you information of things that exist, not on things that don't exist.  The DNS system is a hierarchical system that tells you what IP address corresponds to what name, and yes, things like Kademlia can do it in a decentralized way.  But with a coin, you don't (only) need to know "what IP address corresponds to what name", but also that "no other name corresponds to that IP address", in such a way, that the delegated server has an incentive to tell you that such is not the case, that this could actually right now be assigned even though you don't know, and that the server having that information may very well not be online.

This can only be the case if there is a common knowledge of which you can verify yourself, continuously, whether spending of the same right to spend is being transacted somewhere else right now, or whether an address is "locked in" (style LN / sidechains / ....).

I don't think there is any way out, unless you give up on trustlessness (and you start trusting entities telling you whether you should, or shouldn't, accept a payment - with all the danger of those entities actually being the ones trying to scam you) or on decentralisation and permissionlessness (even if those trusted entities are to be trusted, they are single points of failure whose role you couldn't just, on your own decision, not take over).

In fact, you'd end up with a system that looks a lot like the current fiat system, with a lot of cross checks, reporting, auditing etc... even though it might be somewhat automated cryptographically.

Essentially, if you think that users should have their account at some or other "delegated node of trust", that's nothing else but the banking office where these users have their account, in a way.  If these entities decide to tell you that owner X has nothing in his account, then you will not trust owner X's transactions, even though he may have stuff in his account, but his "node of trust" to which his account was delegated, decided to punish him for one or another reason...

thats the utopia, not the promise

whoever holds the LN seed chooses and directs the "route" LN users follow.

it will end up being hubs charging 1cent
rather than hops where each hop charges 1 cent.

EG
channel Google<>frank
channel frank<> dave
channel dave<>cryptinvest
channel cryptinvest<>bitstamp

so you think if google wanted to deposit funds into bitstamp
frank dave cryptinvest get 1 cent each..
costing google 3cents to get to bitstamp

what ends up happening is that people end up not wanting to pay fee's per channel hop because it all adds up. so hubs start connecting everyone to it
                    google
                       ^
                       v
cryptinvest<>bitstamp<>dave
                       ^
                       v
                   frank

now everyone can fund bitstamp for free because they are connected to bitstamo and google can pay cryptinvest for 1cent via bitstamp
thus making bitstamp get lots of fee's due to being a hub

i can even envision the LN dns SEED acting as 'real estate' agents selling a listing for a price.
EG to get to be listed as a possible route you have to pay a fee to the LN dns to get listed
which only hubs with lots of channels connected would afford

much like the utopia of solo mining turned into pool mining.. efficiency + cost saving + other factors = things not rmaining as the utopian dream people hoped for

The emergence of rapid transactions will not contribute to centralization. Everyone BTC holder can open a payment channel to earn a commission - > comission fee very reduced. LN will have an open source code. Cost of BTC winn grow faster because market offer BTC for sell reduced - many BTC will blocked in LN channels + pay volumes fast transaction INCREASE in 100x.