Discussions around block size always include someone obsessed with storage, and that argument can go both ways. But storage is the least interesting aspect. Increased transactions per second means increased validations per second, which has a number of implications including increased latency in the peer to peer network. This affects mining protocols, orphan rates, mempool size, throttling algorithms etc. There are several interesting aspects, and storage is very far down the list.
A lot of work has been done for the past few years to improve that situation, look at the new crypto library for a few concrete examples. This kind of work is what led the developers behind segwit to provide for a quadrupled block size. I am not familiar enough with the details to have an opinion if that is somehow an optimal block size, but that's where we are now that segwit locks in.
I don't know what it is about block size that makes everyone and their mother so desperate for an opinion about it. People who have no clue about the underlying protocol or the various scaling properties of the involved algorithms pull a number out of the air and somehow that's their opinion of the matter. There are probably very interesting things to discuss here but it's completely drowned in noise of one single constant.
I completely agree, my point is that even with full 8MB nodes the barrier of entry is not much higher than it is today with 1MB node. Or to put it an other way I don't think we'd have 10 times more nodes if block size was at 100kB. I postulate that the main barrier of entry right not is simply bothering to actually set up a node.
At least in the "first world" the average joe has the resources available to run a full 8MB node without having to starve himself, both in terms of bandwidth[0] and storage[1]. So while there's some discussion to be had about what constitutes a safe amount of nodes online (Taek has some interesting points about that elsewhere in this thread) I still think it's pretty irrelevant in this blocksize debate.
>But storage is the least interesting aspect. Increased transactions per second means increased validations per second, which has a number of implications including increased latency in the peer to peer network. This affects mining protocols, orphan rates, mempool size, throttling algorithms etc. There are several interesting aspects, and storage is very far down the list.
That's a very interesting point and something I hadn't really thought about. I'm surprised though, given that nowadays the rule of thumb is "wait for 5 confirmations before accepting a transfer" it seems to me that the latency between the moment I issue a transaction to the network and the moment it's "mined" into a block is not that significant? I mean, at a rate of 8MB every 10 minutes you have the time to do a lot of crypto work on the nodes...
[0] https://www.statista.com/statistics/204952/average-internet-... [1] https://www.newegg.com/Desktop-Internal-Hard-Drives/SubCateg...
> As for bandwidth it amounts to less than 14kB/s on average.
Fourteen kilobytes a second is quite an understatement of the bandwidth needed to operate a bitcoin full node. Remember, nodes don't just download the chain: they also relay unconfirmed transactions and blocks to other nodes, which increases bandwidth requirements by a few orders of magnitude.
When I ran a full node a while back it was eating a couple of TB a month. Nowadays that blocks are full I suspect that it's quite a bit more than that.
Anyway, that's not even my main point. I just thing that I think this particular problem is overblown and used as a bogeyman by the opposing party. Not that "big blockers" are without blame, there are insane conspiracy theories on both side and even in between.
Hacker news is basically one of the only places where I've seen people being able to discuss these issues on a technical level without going full "the illuminati freemason bankers are trying to destroy bitcoin".
Not doing those things actually weakens the network. If most users operated a node like the one you described, there would be several negative effects on the overall network's health:
* It would take longer to relay transactions: full nodes would be less interconnected and there will be more "hops" between any two nodes. This in itself can cause many problems: it gives miners an incentive to centralize geographically (latency increases orphan rate) and can make a 51% attack slightly easier.
* The network will become more vulnerable to DoS attacks, because only nodes that relay transactions need to be targeted.
* Full nodes will have even higher bandwith requirements, because they will have to relay transactions to more nodes (that don't relay in turn), like a torrent with many leechers.
It's kind of like how the development of popcorn time is bad for the health of torrent networks (not the same though, different reasons).
> Anyway, that's not even my main point. I just thing that I think this particular problem is overblown and used as a bogeyman by the opposing party. Not that "big blockers" are without blame, there are insane conspiracy theories on both side and even in between.
I get it, but I don't think this problem is overblown.
It's true but it's not exactly in the same category. I was mainly considering attacks that would attempt to mess with the ledger, not a denial of service. DoSes are a problem of course, but it's transient and can be fixed "in hindsight" if it actually becomes an issue. Successfully tampering with the blockchain on the other hand would be a massive shitfest and non-trivial to fix.
I'm not really worried about the number of nodes becoming insufficient to relay the transactions because clearly people who are invested in bitcoin have all kinds of incentives to keep it running. That means that big exchanges and "whales" could easily invest a few dollars a month to run a full node on a high bandwidth VPS. Keeping Bitcoin usable would actually be a good incentive to run full nodes if that ever became a problem. If Bitcoin really keeps growing eventually banks, big companies and governments will want to run nodes and I presume they won't host them on a raspberry pi connected to a 56k landline.
On the other hand if your internet connection is too crappy to support 8MB blocks (14kB/s, remember?), I can't really imagine that it'll do a lot of good as a backbone for the bitcoin network. If anything you'll end up eating up bandwidth from external nodes as they forward you the blocks and transactions while not really contributing much yourself.
When the lightning network is up and running you can bet that I'll be using my full node for that too.
120GB of disk space is a big hit. My drive is a 500GB SSD. Bitcoin runs a lot slower on an HDD. But mostly people saying "oh it's only $5 a month" I feel aren't actually running nodes. It's inconvenient already, the sizes are already hurting adoption, and making the chain bigger is definitely going to make that situation worse.
And once it's actually synchronized HDD or SSD doesn't matter much in my experience.
I don't think it's really a factor in adoption, most users don't feel the need to run a node already. People running nodes are enthusiasts first and foremost.
That's a big part of the issue. Already people prefer to run lite clients for convenience. There are heavy systemic risks that get worse and worse as the number of full nodes decrease. Back room deals become a lot easier for those who do run full nodes, because there aren't enough people running full nodes to prevent unfavorable network changes.
One you have synced your node, an HDD isn't a big deal but if you only run your node every few weeks or months you have thousands of blocks to sync each time. Then it's a pain, and SSDs are nicer. (This is my personal situation). I typically only boot my node if I intend to spend or receive coins, the sync time is already frustrating to me on an SSD and i7.
You don't want an ecosystem where only enthusiasts are verifying transactions. That's abusable, and if it forms a significant part of the economy it absolutely will be abused.
It's true but as I pointed out in a sibling comment those risks are pretty slim as long as it's highly unlikely that all node runners could be colluding. It takes only one "uncorrupted" node to raise the alarm publicly, then anybody can spawn a node and check for themselves if they want.
>You don't want an ecosystem where only enthusiasts are verifying transactions. That's abusable, and if it forms a significant part of the economy it absolutely will be abused.
Okay but that's the way bitcoin works, with or without SegWit, with or without big blocks. There are not enough incentives to run a full node already. I don't think that will change unless you manage to create an incentive somehow.
For $20 a month someone could easily run a large-block full node and that price will only come down.
Also, would you pay $20/mo for a credit card? That's a big expense to participate in a financial system.
How many independent nodes do we need to ensure it keeps working?
I currently run a full node since I have a gigabit connection but I will definitely stop if I need to increase storage space by that much. The current growth rate is maybe a tenth of that.
You also need to keep in mind that most residential usage limits are around 400 gigabytes per month. After this change, within a year the vast majority of people will not be able to run nodes outside a datacenter.
Within a few years when the blockchain is multiple terabytes it will be nearly impossible to download on a residential connection due to size and especially usage limits.
Also disk usage arguments still apply to SegWit. For instance: see this 3.7MB block on the test network, containing 468 transactions.
https://testnet.smartbit.com.au/block/00000000000016a805a7c5...
so if we're extrapolating from worst case scenarios, that's 658GB per year!
Not everyone has access to such resources. You and I do, but many people run their lives on PCs with 32GB EMMC soldered to the motherboard, mobile devices, netbooks, public school computers with limited network-based storage, and internet connections that can barely allow a couple gigabytes per month. If a technology is not inclusive of the median hardware possessed by that (large) community, it will not gain traction beyond hackers with like us.
Also, the vast majority of people I know do not use desktop computers. Lugging around a 6TB external drive just to deal with payments, and then dealing with backing up that drive, is not exactly a good sell.
So the question is: amongst the people running full nodes today, how many wouldn't be able to cope with 8MB blocks?
Also note that 8MB is the max size and blocks could be smaller, I don't expect that 8MB blocks would get fill up in the short term.
However, in my experience any node accepting incoming connections will have many more than the additional 8 incoming connections we might expect. I throttle total connections at 64, but unthrottled I've seen it go over 100, admittedly after a few days of uptime.
Simply spawning 10 thousand nodes would not allow someone to take over the network in any sense.
In practice as long as you have at least a few thousand nodes run by a variety of uncooperating users around the globe it should be enough to handle all the transactions and raise an alarm if anybody attempts something foul. The amount of coordination required to carry an attack by taking over the node network seems rather implausible.
For one thing you'd have to convince all the big exchanges and the miners to play along and then you'd have to hope that none of the people running the thousand of nodes would see the fork and raise the alarm publicly.
It might however make it easier to run "transaction malleability" attacks however but that's an other issue altogether and ought to be fixed in the short term.
What can you make against bitcoin with that?
Bitcoin users are important economic entities.
There is no distinction between a "consumer" and a "business" in Bitcoin. Everyone is a user. Everyone sends and receives payments.
What coinbase and other exchanges consider to be the "true" bitcoin matters a whole lot more than what some guy who just installed a bitcoin node on his raspberry pi thinks is the true bitcoin.
In the end bitcoin is worth what people are willing to pay for it. Even if you have the majority of the nodes if nobody accepts your forked bitcoin then it's worthless.
That way you still get to keep control of your keys, and you get a little control over the consensus rules you think are valid, but you don't have to sync the entire chain.
This is what most users are likely doing.
If that's not the case, then, well... that's what they should be doing.
Just because Coinbase has X amount of customers doesn't mean that many people are using Coinbase as a wallet. I'm a Coinbase customer but I don't use Coinbase as a wallet.
And I say this as someone that dislikes Coinbase and was burned by them (one day they seized all my coins until I verified ID on login, then erased my bank account connections and refused to delete my PII that I only provided to withdraw my coins in the first place.)
The CPU and bandwidth resources are trivial. Torrent traffic from a decade ago is many orders of magnitude more bandwidth than what bitcoin or even all currencies combined are. It is literally a non-issue.
I think that's a valid point, but we won't know how it will play out until we actually get there. It seems to me that the "large block" crowd is of the "let's not go there" opinion, but I think they're resisting the inevitable.
That seems to be the consensus and that's my main issue with the entire ordeal. We're talking about a $40 billion market cap currency, not a bittorent tracker. Maybe it would be worth taking a bit more time to run some simulations and figure out the long term plan?
IMO looking mostly from the outside (I'm not heavily invested in any cryptocurrency) it would've been more reasonable to bump the block size a little bit in the short term (maybe 2 or 4MB) to let the current architecture carry its weight for a few more months or even years and take that time to come up with something better than "back of the envelope calculations", to cite what I read in some LN papers.
That sounds like a terrible plan. Let it play out on Litecoin or some other toy blockchain.
The large block crowd wants the easiest most straight forward way of scaling now. Which isn't to say other solutions aren't welcomed. But it's strange that the Core team has made it an either or.
It's weird how in certain forums there seems to be overwhelming support for blockstream - but everyone I meet in real life doesn't have that same enthusiasm.
Not that it will do them a lot of good, what matters most is what the exchanges and other important economic entities handling bitcoins consider "the real bitcoin", having a majority of nodes won't do much.
There's simply not enough incentives for Bitcoin users to spawn a node so most people prefer to use a light wallet.
Furthermore ~400GB per year is not particularly insane, you can get a 6TB drive for less than $200 and that'll be enough to store more than 15 years worth of blockchain.
As for bandwidth it amounts to less than 14kB/s on average.
>there could be millions per second if micropayments do take off, and that will most likely happen via the channels as described in the lightning docs
Maybe, but I'm a bit surprised that so many people believe that LN will be successful when we have so little data on how it'll actually work. But I'm going to repeat myself, I already explained my concerns here: https://news.ycombinator.com/item?id=14759965 (with some insightful replies).
In particular I can't really imagine how LN could reach a Visa-level number of transactions without having a few mega "nexus nodes" concentrating most of the open LN channels. That would give those nodes some power as they could decide who they pair with (which channels they open) which would be a big blow to the decentralized nature of Bitcoin.
The idea that every bitcoin user would each end up having a handful of channels open and we'd keep routing around the graph (with all the balancing issues it involves) to pay everything seems rather optimistic and aligns poorly with real-world incentives.