LN+segwit vs big blocks, levels of centralization.

[thejaytiesto]

Let's imagine we get LN working nicely with segwit, we get people finally being able to make instant, cheap transactions to buy stuff in a viable way in real-life shops and so on, without having to wait for a single confirmation in a reasonable way (no Roger, 0-conf may have been reasonable back then when nobody gave a fuck about bitcoin, so not anymore)

What would be the fees for on-chain transactions at this point?

Can someone explain with detail and facts (no propaganda) what % of centralization would LN+segwit transactions encumber compared to on-chain transactions under a big block scenario?

I just want to know what is the actual level of centralization of LN+segwit. Andreas makes it look pretty nice here:

https://www.youtube.com/watch?v=gF_ZQ_eijPs

Is it a tragedy if no one can make on-chain transactions anymore without paying a massive fee? Will the blocksize be raised so on-chain transactions are still viable?
Is LN+segwit decentralized enough that on-chain transactions are nonsense? or not being able to make on-chain transactions is an huge deal that only leaves a small elite of rich people being able to move bitcoin at top levels of privacy and decentralization?

And most importantly, why big blockers pretend to argument big blocks (enough to cater for mainstream, coffeeshop-tier transactions) wouldn't pose a deadly centralizing factor? (as the backbone of the network is centralized in less and less widespread nodes, ending up in datacenter farms) isn't this a bigger issue than the % of centralization LN+segwit would pose? since LN+segwit is a layer, not the root of the network.

Why big blockers want to call p2p cash something that would end up controlled by datacenters?
Similarly, why LN proponents call it p2pcash when (presumably, as argued by some which im not sure if they are legit or FUDsters) LN would end up a bank cartel where the hubs are centralized and controlled? (and again, only a few can bypass LN and make on-chain transactions anymore to get top-level privacy)

Even tho it's clear centralizing the nodes is worse than the supposed LN centralization problem (since you at least can still move coins on-chain decentralized due the conservative blocksize), aren't both sides facing an obvious problem?

Am I taking crazy pills?

I don't know who is trolling, who is a paid propaganda shill and who is legit anymore.

[1715000508]

1715000508

[1715000508]

1715000508

[1715000508]

1715000508

[1715000508]

1715000508

[BillyBobZorton]

There's no way to scale bitcoin to mainstream levels on-chain, there's just no possible way, because it would take insane levels of block size, and no, the hard disk capacity and network speed will not catch up on time with the block size needed to deal with all the transactions a global mainstream scenario would need.

Therefore, I don't see any other way but LN. And as far as I know, segwit + LN is totally decentralized.

[FiendCoin]

/sigh

Bitcointalk has really become a shitshow lately.

Firstly, we need both on-chain and off-chain scaling. What we don’t need is unlimited size blocks being controlled solely by the miners or any 1 group for that matter. On-chain scaling should be done as safely as possible with a super-majority of community support.

Secondly, LN is only 1 off-chain solution to scaling being worked on. The whole shill story of LN = centralization is bullshit.

[The One]

Let's imagine we get LN working nicely with segwit, we get people finally being able to make instant, cheap transactions to buy stuff in a viable way in real-life shops and so on, without having to wait for a single confirmation in a reasonable way (no Roger, 0-conf may have been reasonable back then when nobody gave a fuck about bitcoin, so not anymore)

What would be the fees for on-chain transactions at this point?

Can someone explain with detail and facts (no propaganda) what % of centralization would LN+segwit transactions encumber compared to on-chain transactions under a big block scenario?

I just want to know what is the actual level of centralization of LN+segwit. Andreas makes it look pretty nice here:

https://www.youtube.com/watch?v=gF_ZQ_eijPs

Is it a tragedy if no one can make on-chain transactions anymore without paying a massive fee? Will the blocksize be raised so on-chain transactions are still viable?
Is LN+segwit decentralized enough that on-chain transactions are nonsense? or not being able to make on-chain transactions is an huge deal that only leaves a small elite of rich people being able to move bitcoin at top levels of privacy and decentralization?

And most importantly, why big blockers pretend to argument big blocks (enough to cater for mainstream, coffeeshop-tier transactions) wouldn't pose a deadly centralizing factor? (as the backbone of the network is centralized in less and less widespread nodes, ending up in datacenter farms) isn't this a bigger issue than the % of centralization LN+segwit would pose? since LN+segwit is a layer, not the root of the network.

Why big blockers want to call p2p cash something that would end up controlled by datacenters?
Similarly, why LN proponents call it p2pcash when (presumably, as argued by some which im not sure if they are legit or FUDsters) LN would end up a bank cartel where the hubs are centralized and controlled? (and again, only a few can bypass LN and make on-chain transactions anymore to get top-level privacy)

Even tho it's clear centralizing the nodes is worse than the supposed LN centralization problem (since you at least can still move coins on-chain decentralized due the conservative blocksize), aren't both sides facing an obvious problem?

Am I taking crazy pills?

I don't know who is trolling, who is a paid propaganda shill and who is legit anymore.

There is plenty of room to increase capacity by increasing blocksize. Scaling will be improved. As for every day transaction by billions of users, we are not even there yet. Every software started small. It got bigger over time. No one started a software, big and to meet the needs and demand, 10 to 30 years in the future. Not Bill Gates. Adoption takes time. Blocksize limit of 1mb is preventing adoption. New users see a "civil war" and then go ermm mmm nah. LN is in my opinion rubbish. We have plenty of time to research new ways of scaling that is decentralised. That means in the hands of the Full Nodes and miners. Not some centralised entity.

The refusal to raise the blocksize is the prime reason for present "civil war."

[d5000]

As I've already explained somewhere here, both approaches have centralization risks. But in both cases, if technology evolves rapidly, then it's more decentralized.

Big blocks can be decentralized if internet bandwidth, storage and CPU/RAM costs (needed to validate the tons of transactions in big blocks) go down significantly. At this very moment, a 100-MB-block Bitcoin could be enough to serve 1 billion users (the maximum potential I see today: first-world households and the upper-middle class of third world countries), but that would be definitively too much for consumer hardware, mid-scale businesses in first-world countries could probably handle up to 8-10 MB.

A decentralized LN depends mostly on availability of 24/7 stable internet connections, less on bandwidth. Because LN is not 100% trustless, clients must be online 24/7 to avoid being scammed and to enable a decentralized routing. That's not a problem in big cities and well-connected countries, but it could be one in rural areas, especially in third-world-countries.

But there is not only "big blocks" and "LN". My favorite solutions, sidechains and extension blocks, would be a compromise of both approaches. A Bitcoin with 10 sidechains with similar characteristics to the main chain (1 MB blocks) would offer about 10 times the original capacity (11 times minus the settlement transactions), but most nodes would have to validate and store only the main chain and a sidechain. So it would still be possible to run Bitcoin on consumer hardware with high capacity without sacrificing the low-trust character of on-chain transactions.

You could scale to 1 billion users with 25 sidechains of 4 MB blocks. That would be surely possible today (but not necessary, volume is not that high still).

That's why I eagerly await the RSK start and hope that a "two-way-peg" enabling mechanism is added to Bitcoin soon.

[The One]

As I've already explained somewhere here, both approaches have centralization risks. But in both cases, if technology evolves rapidly, then it's more decentralized.

Big blocks can be decentralized if internet bandwidth, storage and CPU/RAM costs (needed to validate the tons of transactions in big blocks) go down significantly. At this very moment, a 100-MB-block Bitcoin could be enough to serve 1 billion users (the maximum potential I see today: first-world households and the upper-middle class of third world countries), but that would be definitively too much for consumer hardware, mid-scale businesses in first-world countries could probably handle up to 8-10 MB.

A decentralized LN depends mostly on availability of 24/7 stable internet connections, less on bandwidth. Because LN is not 100% trustless, clients must be online 24/7 to avoid being scammed and to enable a decentralized routing. That's not a problem in big cities and well-connected countries, but it could be one in rural areas, especially in third-world-countries.

But there is not only "big blocks" and "LN". My favorite solutions, sidechains and extension blocks, would be a compromise of both approaches. A Bitcoin with 10 sidechains with similar characteristics to the main chain (1 MB blocks) would offer about 10 times the original capacity (11 times minus the settlement transactions), but most nodes would have to validate and store only the main chain and a sidechain. So it would still be possible to run Bitcoin on consumer hardware with high capacity without sacrificing the low-trust character of on-chain transactions.

You could scale to 1 billion users with 25 sidechains of 4 MB blocks. That would be surely possible today (but not necessary, volume is not that high still).

That's why I eagerly await the RSK start and hope that a "two-way-peg" enabling mechanism is added to Bitcoin soon.

A billion users won't happen overnight. Start small with secure quick confirmation. A small business would be a "test case" to see if successful. I looked at sidechain and does look good. Need more research on it. Downloading 8mb is 2 to 4 seconds on my computer and true for many. Worrying about those with 56k line is pointless.

[dinofelis]

I have to say I never understood the priorities in the worries expressed in this debate, and I think it is because people start from some pre-conceived dogma to which they think they have to cling, and adapt all the rest around it.  I know that what I'm saying is considered controversial, and that I'm called a shill for that sometimes (I'm not), but I have posted it several times, I've been arguing over it several times, and I have never heard any convincing argument to the contrary.  It is this, in my eyes quite obvious, statement:

"the amount of non-mining full nodes is absolutely no measure for the decentralization of a coin." (at least, with a PoW coin).

I already explained several times why I'm convinced of that statement, and that is because the only *source* of data, of block chain, is a consortium of miners, which these nodes can simply copy, and put at disposal, like a proxy server.  It would entirely matter if mining nodes were totally distributed in the network, and pop up and disappear in the P2P network.  If non-mining nodes were potential mining nodes, but just decided, for the moment, not to mine.  In that case, the number of nodes WOULD be a measure of decentralization.  But with a consortium of miner pools, that are made up, at most, of, say, 20 nodes (5 are sufficient for >50% of the hash rate), and with the immense difficulty associated to mining (so that you just don't invent yourself a significant solo miner), non-mining nodes have only 2 options:
- copy faithfully the data that those few miner nodes produce
- refuse to copy it and come to a halt.

The miners having invested huge amounts in their mining equipment, the network equipment is only a small fraction of their investment, and as such, these miner data centres are totally capable of serving the whole user base with light wallets ; the few thousands of non-mining full nodes that serve the light wallets of many users are only providing a proxy service, but don't add any *decision power*.

When I write this, I always get remarks about the *users*.  Yes, the users are important economically, but if there are millions of users, and only a few thousand of full nodes, these full nodes are in no way representative of these users. 

So, the actual situation, which will not improve, is that the only entities keeping miners in check are other miners, and as long as the consortium of miners comes amongst their members, to a consensus of a block chain, that's the only chain that exists and that full nodes can copy, whether they like it or not: it is the only source of block chain data that can provide a block chain with all that PoW in it.  As such, the current architecture of the network is not P2P, even though on paper it looks like it: it is a few central nodes (the pools) that are the "backbone" and the collective source of block chain ; they are the only true "guardians" of the protocol rules and keep one another in check ; and they produce collectively a block chain, which is the only one available for you to use, or not to use.  This block chain is then diffused through a P2P network that helps "serving" the light wallets of users, but these light wallets could get the block chain information just as well directly from the miners.  The non-mining full nodes can just copy, or not copy, this block chain, that's their only options.

I know that things get more involved when miners disagree amongst one another, but they don't.  The large majority hash rate produces only one single block chain, to be taken or to be left.  If ever miners disagree, and they make two or more block chains (forking), then the P2P network of nodes can copy one or the other and chose ; but in the end, it are not these nodes, but rather the users, with their light wallets, and exchanges, that will vote with their money.  Even if 99% of all P2P nodes pick chain 1, if the richest users and exchanges want to pick chain 2, they can, by pointing their wallets to those miners (pool nodes) that produce chain 2, and the market will evaluate chain 1 and chain 2, independent of how many non-mining nodes copy chain 1 or chain 2.  So even in a case of disagreement between miners, and having several alternative chains available, what the non-mining  P2P nodes do, have no importance in this decision.

As I said, this would be totally different if just any node could start mining all over the P2P network, but these days are gone.   There are 20 source nodes of block chain and that's it (and 5 have a majority).  Whether there are 5000 copy cats or not doesn't really matter.

Once this dogma has fallen, the discussion becomes totally different.  Big blocks are mostly no obstacle to the 20 data centres of the miner pool nodes.  Everyone who really wants to, can obtain a copy of the block chain, but that's a bigger investment than having your home PC run as a node ; however, as that running home PC doesn't really matter in the decentralization of the network, one shouldn't care about that.  A big exchange might like the idea to have a full node.  People hung up to the privacy of their transactions might like to invest into the network capacity and processing capacity to set up a full node.  But for far most users, it doesn't matter.  Most of them can use electrum or other light wallets, and connect to the few full nodes that are around, like those of exchanges, and, of course, the miners themselves.

This means that the main argument against "everything on chain" actually holds no water.  Bitcoin, at this moment, is an affair of ~20 mining nodes, that's all.  All the rest are at their mercy anyhow, and can play copy cat and proxy server, or shut up.  These 20 nodes can handle big blocks.   
The only argument that *does* hold water against shutting down all non-mining full nodes, is just a matter of privacy.  If everyone has to connect directly to the 20 miner nodes, you're much easier identified than if you propagate your transaction through a P2P network.  One could continue to use a P2P network for transaction propagation, though.  No need to have a block chain on disk for that.  Or one can use tor or a VPN.  This has nothing to do with  "checking the block chain".

But of course, it is not because only 20 miner nodes can handle big blocks, that one needs to go for that solution, if a better solution is available.   LN is maybe a good solution, but it seems rather evident that LN will not be different, in network structure, than the 20 miner nodes.  We may have the 50 LN hubs.  However, what I don't like about LN, is that you are way more dependent on your LN hub than you are dependent on your miner node.  With miner nodes, a transaction is a transaction.  You can send it to miner 1, and if he wants to ask you things, you can then try miner 2 at no cost.  When you are hooked up to hub 1, however, you need to settle and open a channel to hub 2, if ever hub 1 puts conditions. 

So it seems to me that LN is more centralized, and is less permissionless than the 20 miners that make bitcoin now, unless it is easy to settle and re-connect, but that means that you can actually do your thing on chain.

I think that the chain capacity should never be much lower than the projected LN capacity, or you are locked in to your LN hub (because of costs, and difficulty to have access to the chain).

[Wind_FURY]

/sigh

Bitcointalk has really become a shitshow lately.

Firstly, we need both on-chain and off-chain scaling. What we don’t need is unlimited size blocks being controlled solely by the miners or any 1 group for that matter. On-chain scaling should be done as safely as possible with a super-majority of community support.

Secondly, LN is only 1 off-chain solution to scaling being worked on. The whole shill story of LN = centralization is bullshit.

Can the block size be decided algorithmically depending on the demand? Why do the miners have to control it? That would make it open to a lot of bad actors colluding.

[dinofelis]

Can the block size be decided algorithmically depending on the demand? Why do the miners have to control it? That would make it open to a lot of bad actors colluding.

Of course.  Monero has such an algorithm for instance.  It is maybe not the best possible, but at least it solves the issue in a specific way.

https://monero.stackexchange.com/questions/35/is-there-any-block-size-limitation

[d5000]

A billion users won't happen overnight. Start small with secure quick confirmation. A small business would be a "test case" to see if successful. I looked at sidechain and does look good. Need more research on it.

You are right that a billion user "won't happen overnight". That's why we can go gradually to reach that goal - first let's test RSK's "semi-centralized" sidechain which will start in May if all goes well, then try to get the code for a decentralized variant into Bitcoin (see the "Drivechain" proposal), and then try to add another sidechain, and so on.

Downloading 8mb is 2 to 4 seconds on my computer and true for many. Worrying about those with 56k line is pointless.

It's not only downloading - transaction validation needs also CPU and RAM.

"the amount of non-mining full nodes is absolutely no measure for the decentralization of a coin." (at least, with a PoW coin).

I already explained several times why I'm convinced of that statement, and that is because the only *source* of data, of block chain, is a consortium of miners, which these nodes can simply copy, and put at disposal, like a proxy server.

You POV is interesting - maybe the quantity of "non-mining user nodes" is really less relevant as most people think. However, I have two comments:

- non-mining user nodes are also useful because they add redundancy to the decentralized storage of the blockchain - if the 20 miners should go offline some day (that's not completely impossible, if we take into account that most miners are located in a single country - a large electricity blackout in China could be the reason for such a "catastrophic event"), then other nodes could jump in and serve as "data source".
- the discussion about UASF is showing that user nodes actually do have some powers: they can delay the propagation of blocks that do not fullfill some requirements (e.g. are not voting for their preferred soft fork solution). If only miners validate and transmit blocks in a kind of "miner intranet", then this (in my opinion) important element of the power balance (it reduces miners' power a bit) becomes unavailable.

LN is maybe a good solution, but it seems rather evident that LN will not be different, in network structure, than the 20 miner nodes.  We may have the 50 LN hubs.  However, what I don't like about LN, is that you are way more dependent on your LN hub than you are dependent on your miner node.  With miner nodes, a transaction is a transaction.  [...]

So it seems to me that LN is more centralized, and is less permissionless than the 20 miners that make bitcoin now, unless it is easy to settle and re-connect, but that means that you can actually do your thing on chain.

I agree, and that's why for me sidechains are a better scaling mechanism than LN - it preserves the "trustless transaction" paradigm.

[dinofelis]

"the amount of non-mining full nodes is absolutely no measure for the decentralization of a coin." (at least, with a PoW coin).

I already explained several times why I'm convinced of that statement, and that is because the only *source* of data, of block chain, is a consortium of miners, which these nodes can simply copy, and put at disposal, like a proxy server.

You POV is interesting - maybe the quantity of "non-mining user nodes" is really less relevant as most people think. However, I have two comments:

- non-mining user nodes are also useful because they add redundancy to the decentralized storage of the blockchain - if the 20 miners should go offline some day (that's not completely impossible, if we take into account that most miners are located in a single country - a large electricity blackout in China could be the reason for such a "catastrophic event"), then other nodes could jump in and serve as "data source".

Yes, that is true: as a kind of "collective indelible memory" non-mining nodes do have the function of "keeping the memory".  However, one can wonder what that memory is worth, without miners.  After all, we don't care much about a static block chain ; this is not something like, say, the "memory of the Holocaust" which has to be transmitted from generation to generation, and kept intact even though evil forces would like to modify it.  No, a block chain only has a meaning as a dynamic entity that can grow.  As such, suppose that all miners collude to change the block chain, and produce a new one, which has a modified history, and nevertheless, more PoW than the old "true" one.  One can say that there has been a successful 51% attack on the immutability of the block chain.  Ok, but suppose now that all miners agree with that, and that this modified chain is the only one available and growing.  What is now the use of your stored, true, old, block chain, with less PoW than the current dynamical one, and which is not growing at all, which is a dead piece of data ?
Ok, at least you *know*.  True.  You know that the immutability of bitcoin is gone.  So what ?  If you possess, say, 1000 BTC (as well on the old, dead chain, as, happily, on the new one) ; are you going to consider them worthless, or are you going to try to sell them (on the new chain of course) anyhow ?  What are most bitcoin holders going to do ?  "leave bitcoin", or try to do as if nothing happened ?  
Your scenario of a big power failure in China is indeed one of the useful cases of "a block chain copy preserved in New Zealand".  Although of course, the miners in China may have a backup of the chain too... unless they have been bombed or something.

But let us imagine another scenario: a big *internet* failure in China's outside link.   Suppose that the rest of the world is cut off from China's network for, say, a week.  Suppose that nodes have copies of the block chain, outside, but almost nobody is mining.  The high difficulty outside China and the low hash power make that only a few blocks are added outside.  But inside China, people are happily mining, trading, transacting.  The block chain grows almost normally in China.
And then, suddenly, the link is established again.  What happens ?  All the external blocks are orphaned.  All external transactions didn't take place, after all.  Whether this chain is saved on hundreds of nodes or not.  The few Chinese miners with all their hash power have been building a longer chain with more PoW, and the outside nodes can throw away all their blocks since they got separated from them.
(BTW, if ever this happens, it is the moment to spend all your BTC if you are outside of course !)

So, the true power of "knowing the block chain", apart from a big bombing of all the miners and their backups, seems to be non-existent after all, apart for informing you.

- the discussion about UASF is showing that user nodes actually do have some powers: they can delay the propagation of blocks that do not fullfill some requirements (e.g. are not voting for their preferred soft fork solution). If only miners validate and transmit blocks in a kind of "miner intranet", then this (in my opinion) important element of the power balance (it reduces miners' power a bit) becomes unavailable.

My argument has always been that there is OF COURSE an "intranet of important miners" because every second counts.  A miner wants to receive *as quickly as he can* the blocks from other miners, because otherwise, he's wasting hash rate mining on an old block.  So every miner has all the reasons to be connected with a very good network connection to the most important other miners, or this will cost him important hash rate.  No way a miner of any importance is going to wait multiple delays in the P2P network for the new block on which to start mining, while his competitors started already.  All the seconds lost between the availability of the new block to other miners, and to him, is wasted hash rate.

Given the fact that "having a good link to another miner" is a mutually beneficial element for miners, there is in my mind no doubt that there is a direct backbone network of miners (at least, the important ones), and that thinking that the P2P network can "filter" miner blocks is delusional.

[unamis76]

There's no way to scale bitcoin to mainstream levels on-chain, there's just no possible way, because it would take insane levels of block size

Insane levels? Why? Hasn't SegWit been developed? Won't it tidy up transactions on the block? You have no confidence SegWit or similar can work in the future or that it can't be further developed or improved?

and no, the hard disk capacity and network speed will not catch up on time with the block size needed to deal with all the transactions a global mainstream scenario would need.

Blocksizes won't increase drastically overnight.

Big blocks can be decentralized if internet bandwidth, storage and CPU/RAM costs (needed to validate the tons of transactions in big blocks) go down significantly. At this very moment, a 100-MB-block Bitcoin could be enough to serve 1 billion users (the maximum potential I see today: first-world households and the upper-middle class of third world countries), but that would be definitively too much for consumer hardware, mid-scale businesses in first-world countries could probably handle up to 8-10 MB.

A decentralized LN depends mostly on availability of 24/7 stable internet connections, less on bandwidth. Because LN is not 100% trustless, clients must be online 24/7 to avoid being scammed and to enable a decentralized routing. That's not a problem in big cities and well-connected countries, but it could be one in rural areas, especially in third-world-countries.

This is pretty much what came to my mind while reading the first post. This makes big block decentralization on par with LN decentralization.

New hardware will always appear and get cheaper, so consumers will eventually switch when their hardware becomes slow, unsupported or obsolete, so that's a plus for Big Blocks. Availability is pretty high (many people have slow, but stable connections), so that's a plus for LN.

However bandwidth and investment in connections by ISP's is very relative to areas and willingness of the ISP to invest... and that takes a lot of time, so that's a less for both.. I think things even out.

[Lauda]

There's no way to scale bitcoin to mainstream levels on-chain, there's just no possible way, because it would take insane levels of block size

Insane levels? Why?

That's a very weird question. Are you familiar with basic math, i.e. one being done in primary school? 7 TPS in theory, 3 TPS in practice at 1 MB. That's ~300k TXs per day (current statistic). To serve 1 billion people, you'd need at least a capacity of 1 TX per person per day (which can be used as some average, as some will do a lot more while others won't transact each day). 1000 MB blocks brings us to 300 000k TXs per day or 300 Million per day (~3000 TPS). In comparison, Visa does 2000 TPS on average and is capable of doing at least 10x that. It's just nonsensical to even attempt this after a certain size if you want to retain decentralization.

Hasn't SegWit been developed? Won't it tidy up transactions on the block? You have no confidence SegWit or similar can work in the future or that it can't be further developed or improved?

Do you even understand what SegWit does? It doesn't seem like you do. SegWit is equal to big blocks essentially (if used and ranges from 1 to 4 MB depending on said usage). SegWit scales down the complexity of validation time from quadratic to linear. That's the only significant improvement related to bandwidth, propagation, storage, RAM or processing which are the constraints for big blocks.

Blocksizes won't increase drastically overnight.

Nobody says that they are. You need to be prepared for the worst case though, even if it means someone filling up the maximum block size from day 1.

New hardware will always appear and get cheaper, so consumers will eventually switch when their hardware becomes slow, unsupported or obsolete, so that's a plus for Big Blocks. Availability is pretty high (many people have slow, but stable connections), so that's a plus for LN.

Incorrect generalization. RAM and SSD prices are currently rising due to shortages. This is of course, temporary, but invalidates your statement.

[dinofelis]

There's no way to scale bitcoin to mainstream levels on-chain, there's just no possible way, because it would take insane levels of block size

Insane levels? Why?

That's a very weird question. Are you familiar with basic math, i.e. one being done in primary school? 7 TPS in theory, 3 TPS in practice at 1 MB. That's ~300k TXs per day (current statistic). To serve 1 billion people, you'd need at least a capacity of 1 TX per person per day (which can be used as some average, as some will do a lot more while others won't transact each day). 1000 MB blocks brings us to 300 000k TXs per day or 300 Million per day (~3000 TPS). In comparison, Visa does 2000 TPS on average and is capable of doing at least 10x that. It's just nonsensical to even attempt this after a certain size if you want to retain decentralization.

Satoshi's initial vision was blocks of about 1GB, and he didn't think of 100 GB a day as a problem, as he was only thinking of "specialized data centers serving the block chain to most users using light wallets", and that was in 2008. 

However, that was under the hypothesis that bitcoin was to be a daily currency for the mainstream, which it clearly isn't for the moment, is not going to become in the foreseeable future, and most probably, never.   Although there is a modest bitcoin adoption for some transactions where one pays for goods and services, this is a tiny part of its market cap, which is essentially about trading and speculation, and there's not much indication that bitcoin will leave the speculative sphere.  Bitcoin is not money, but is a speculative financial asset (like derivatives and stock).

As such, the idea that bitcoin is going to be "mainstream fiat" in the next few years is, today, most probably more remote than it was in 2008, and if the technological challenges of 1 GB blocks didn't scare Satoshi in 2008 whenever bitcoin were hypothetically going to reach VISA style levels of transactions in the not too remote future, they should even less keep us awake at night right now.

But between the (in my mind totally delusional, at least in the near future) hypothetical needs of VISA like transactions with current technology levels and their eventual problems today, and the actual problems induced by 1 MB blocks, there's a factor of 1000.  Even though bitcoin's adoption is far below the VISA like needs (and most probably will never need that), it does suffer from a 1000 times more severe constraint of 1 MB blocks.

As I indicated above, given that the only nodes that play a significant role in the current bitcoin system are about 20, the miner nodes, which have all the infrastructure that is needed to keep bigger block chains, the argument that there would be less proxy copies of them around without much power if ever one would relax somewhat the severity of the 1 MB constraint and that that would lead to "centralisation" (as if a network that is running on 20 deciding nodes, well, 5 deciding nodes with majority, and a few thousand proxy servers can be called a decentralized network today) sounds somewhat crazy in my ears ; especially because one doesn't really know what is going to be the realistic future need of the bitcoin system.  It might very well be that bitcoin never needs more than, say, 4 MB or 8 MB of transaction room for real, because there's no indication that it is going to replace mainstream payments in any foreseeable future.  However, bitcoin thriving on the needed illusion of "replacing fiat in the far future", something needs to be told in the story that bitcoin could, eventually, if that totally hypothetical possibility were, against all odds, to arrive, handle that.  On the other hand, as block rewards are going to dry up in the not so distant future, miners would like to have a fee market that guarantees them a sufficient possibility to make transactions scarce enough so that they can squeeze out enough money from people in the need of transactions of their so valued bitcoins.  Miners are not going to project for the "VISA like volumes" that most probably will never happen, and would like rather to take money from the speculators transacting bitcoin from exchange to wallet and back.

In other words, the scaling problem is most probably a non-existing problem when thinking of VISA level transaction volumes, because there's no indication that bitcoin will reach that.  On the other hand, the 1 MB limit is showing its limits, and even for a speculative asset, more transactions seem to be necessary - be it because for the speculative asset to remain credible, it has to have in it the potential to realize what it will most probably never do: become a mainstream currency, because that's the story behind bitcoin.

[franky1]

last few posts are of the mindset "gigabytes by midnight to beat visa".

wake up think logically.
natural progressive and node capable growth over next couple decades not midnight.

secondly halting blocks at 1mb for native keys with the dooms days of "it cant get to gigabytes by midnight to beat visa". is just delaying the chance of natural growth over the nxt few years

EG imagine we had 2mb in 2011.. meaning the 2015-2017 debate didnt happen
looking at mempool now average 3mb all time.
this year we would have been discussing 4mb

now imagine we had a 8mb consensus rule from 2011.. but then had a second rule where nodes suggest a preference below that of
0.50 2011-2013
0.75 2013-2015
1.00 2015-2016
2.00 2016-2017
3.00 2017-2018
4.00 2018-2019
5.00 2019-2020

where it does not break consensus or split the chain, just like moving from 0.75 to 0.999 did a couple years ago. but allows nodes to be in a little more control as a bit of risk management to stop pools jumping straight to 8mb instantly

[hv_]

/sigh

Bitcointalk has really become a shitshow lately.

Firstly, we need both on-chain and off-chain scaling. What we don’t need is unlimited size blocks being controlled solely by the miners or any 1 group for that matter. On-chain scaling should be done as safely as possible with a super-majority of community support.

Secondly, LN is only 1 off-chain solution to scaling being worked on. The whole shill story of LN = centralization is bullshit.

Yeah.

Off chain scaling is live and started with litecoin (and fellow alts) and exchanges = 'economic user community'

On chain scaling is blocked mostly by these community above.

LN is a nice niche for unsecure micro payments. I d put this into alt scaling.

[d5000]

Yes, that is true: as a kind of "collective indelible memory" non-mining nodes do have the function of "keeping the memory".  However, one can wonder what that memory is worth, without miners.  After all, we don't care much about a static block chain ; this is not something like, say, the "memory of the Holocaust" which has to be transmitted from generation to generation, and kept intact even though evil forces would like to modify it.  No, a block chain only has a meaning as a dynamic entity that can grow.

I disagree a bit here - in a catastrophic event a "static blockchain" could be useful to enable other miners to  mine than those disconnected or affected by a power failure. It hasn't even to be a catastrophic event - normal fluctiations in pool activity (one strong one shutting down, another one emerges) could lead to severe security holes if only the mining pools had stored the blockchain.

 

Although of course, the miners in China may have a backup of the chain too... unless they have been bombed or something.

Yes, but until the power is back, there would be other miners (or pools) taking the Chinese's place. In a mature ecosystem, that should happen relatively quickly to avoid the block time going up too much: Bitcoin businesses not engaged in mining would set up "emergency capacities" and independent miners would work with non-Chinese pools. All these activities benefit from the massive redundancy.

But let us imagine another scenario: a big *internet* failure in China's outside link.   Suppose that the rest of the world is cut off from China's network for, say, a week.  Suppose that nodes have copies of the block chain, outside, but almost nobody is mining.

That's pretty improbable ... first, non-Chinese independent miners would simply use non-Chinese pools, and second, like in the first scenario, such a blockchain split would incentive all big stakeholders of the Bitcoin ecosystem to set up emergency mining capacities to keep the network running.

If the majority of the economic nodes (exchanges, payment processors ...) are also located in China, then your scenario becomes more probable but then we would have already a big centralization problem and everyone outside should try to not do anything (not accept payments, principally) until the link is re-established.

My argument has always been that there is OF COURSE an "intranet of important miners" because every second counts.  A miner wants to receive *as quickly as he can* the blocks from other miners, because otherwise, he's wasting hash rate mining on an old block.  So every miner has all the reasons to be connected with a very good network connection to the most important other miners, or this will cost him important hash rate.
[...]
Given the fact that "having a good link to another miner" is a mutually beneficial element for miners, there is in my mind no doubt that there is a direct backbone network of miners (at least, the important ones), and that thinking that the P2P network can "filter" miner blocks is delusional.

I've also thought about that scenario. But is it that easy as you describe? Mining pools, after all, in theory are competing entities and no pool would like to give a "faster connection to other miner"  for free - and if yes, then they are colluding and we are effectively 51%ed already.

What I doubt is if it is really "mutually beneficial", because if there is no direct link between both miners the benefit is the same than if both have a direct connection. That is only mutually beneficial if e.g. two pools unite against the other pools.

That doesn't mean that your scenario isn't possible or even probable, but here I would like to see academic investigations about that topic. (If you have access to some, then I'll be grateful for a link).

[leopard2]

Let's imagine we get LN working nicely with segwit, we get people finally being able to make instant, cheap transactions to buy stuff in a viable way in real-life shops and so on, without having to wait for a single confirmation in a reasonable way (no Roger, 0-conf may have been reasonable back then when nobody gave a fuck about bitcoin, so not anymore)

What would be the fees for on-chain transactions at this point?

Ah, finally an intelligent question.

You have hit the nail on the spot, exactly that is the question. A good scaling solution might have LN or extension blocks, but the key question is, how can we ensure on-chain transactions are expensive enough to keep blocksize at bay, yet cheap enough to allow MACRO payments.

Micro payments we don't want (BU shills will of course tell you otherwise) onchain.

IMHO on-chain transactions will sooner or later become too expensive with Segwits current 2-4MB blocksize, so a different sizing algorithm is necessary. One that is smart enough to increase blocksize when transaction fees surpass a certain percentage could do the trick.  
There has to be a minimum as well, to discourage micropayments (which can go to LN, DOGE or hell....even the Romans did not use gold coins to pay for coffee.....)

[franky1]

As I indicated above, given that the only nodes that play a significant role in the current bitcoin system are about 20, the miner nodes, which have all the infrastructure that is needed to keep bigger block chains, the argument that there would be less proxy copies of them around without much power if ever one would relax somewhat the severity of the 1 MB constraint and that that would lead to "centralisation" (as if a network that is running on 20 deciding nodes, well, 5 deciding nodes with majority, and a few thousand proxy servers can be called a decentralized network today) sounds somewhat crazy in my ears ; especially because one doesn't really know what is going to be the realistic future need of the bitcoin system.  It might very well be that bitcoin never needs more than, say, 4 MB or 8 MB of transaction room for real, because there's no indication that it is going to replace mainstream payments in any foreseeable future.  However, bitcoin thriving on the needed illusion of "replacing fiat in the far future", something needs to be told in the story that bitcoin could, eventually, if that totally hypothetical possibility were, against all odds, to arrive, handle that.  On the other hand, as block rewards are going to dry up in the not so distant future, miners would like to have a fee market that guarantees them a sufficient possibility to make transactions scarce enough so that they can squeeze out enough money from people in the need of transactions of their so valued bitcoins.  Miners are not going to project for the "VISA like volumes" that most probably will never happen, and would like rather to take money from the speculators transacting bitcoin from exchange to wallet and back.

though there is a Fibre network of supernodes that propagate data between pools superfast and as the first tier outward to the node network, the node network does actually play a crucial part.

all them 20 pools could make their own rules and make blocks how they like just betwen each other, but the 800+ main merchants nodes and backed up by the other 6000 user nodes can reject blocks and literally not want to see blocks for good reasons.

meaning the pools end up wasting time making blocks that are seen as unspendable well before a pools reward matures...
making pools know they wont even get to spend it before they even get a chance.

we have already seen many many times that when pools try something that gets rejected in seconds they dont persist for 100 blocks and then try waving their rewards to merchants in an attempt to get merchants to accept the blocks.. its too late merchants threw them blocks aside hours earlier
so pools end up following the nodes rules within minutes .. not continuing to be funky for hours and then try swaying merchants to change their node to accept what pools have made

it seems somewhere along the line you have been sold by some pitch that pools dont have a symbiotic relationship with nodes.

maybe some altcoin works the way you pretend bitcoin works. but bitcoin does not work the way you try to presume it does. pools end up following what the node network find acceptable.

[The One]

Let's imagine we get LN working nicely with segwit, we get people finally being able to make instant, cheap transactions to buy stuff in a viable way in real-life shops and so on, without having to wait for a single confirmation in a reasonable way (no Roger, 0-conf may have been reasonable back then when nobody gave a fuck about bitcoin, so not anymore)

What would be the fees for on-chain transactions at this point?

Ah, finally an intelligent question.

You have hit the nail on the spot, exactly that is the question. A good scaling solution might have LN or extension blocks, but the key question is, how can we ensure on-chain transactions are expensive enough to keep blocksize at bay, yet cheap enough to allow MACRO payments.

Micro payments we don't want (BU shills will of course tell you otherwise) onchain.

IMHO on-chain transactions will sooner or later become too expensive with Segwits current 2-4MB blocksize, so a different sizing algorithm is necessary. One that is smart enough to increase blocksize when transaction fees surpass a certain percentage could do the trick.  
There has to be a minimum as well, to discourage micropayments (which can go to LN, DOGE or hell....even the Romans did not use gold coins to pay for coffee.....)

The question is, what business accept micro payments? Assuming you meant micro = less than a cent/penny