Lightning Network: “Bitcoin Doesn’t Scale” [pdf]

http://lightning.network/lightning-network.pdf

by bootload • 11 years ago

138 106 11 years ago

DougN7• 11 years ago

I've always wondered how long until the blockchain would grow before it became unlikely new users/miners would want to hold a copy. At some point, it seems like it would grow faster than typical network download rates meaning a new miner, etc would never be able to get their initial copy.

desdiv• 11 years ago

You don't need the full blockchain to become a miner, you just need a database of all unspent transaction outputs. The last time I checked said database was only a couple hundred MB.

As for people who do need the full blockchain in the future scenario you describe, they could always pay someone to sneakernet it to them.

TeMPOraL• 11 years ago

As for people who do need the full blockchain in the future scenario you describe, they could always pay someone to sneakernet it to them.

I just imagined an armed convoy protecting a truck loaded with hard drives.

sickpig• 11 years ago

Users could use SPV clients in case of resource constraints, on the other hand there are incentives in place, that till now, have kept miners from shutting down their nodes.

Bitcoin core latest version (0.10.0) have reduced significantly the bootstrap time of a node (e.g. through "header-first" mechanism).

Future releases will introduce other improvements that will lead to a reduction in terms of network bandwidth/latency (O(1) block propagation, IBLT) and IO space (e.g. pruning)

That said it's true there's no economic incentives in place to "just" run a full node. Person that run full nodes does that for a lot of reasons modulo the economic one.

Hence the number of full nodes is steadily declining http://www.bitcoinpulse.com/#/chart/bitnodes/num_nodes

There are proposals (1) floating around to fix the situation through, guess what... economic incentives, but we are for from getting some sort of agreement, worse than that I dare say that in this case we're still struggling with problem awareness.

(1) https://bitcoinism.liberty.me/2015/02/09/economic-fallacies-...

oleganza• 11 years ago

That's an interesting idea, can't wait for the paper to come out. Compared to my scheme of Ripple-like IOU network with point-to-point joint escrow (https://www.mail-archive.com/bitcoin-development@lists.sourc...) here are pros/cons (to the best of my understanding so far):

Pros:

1. They already have a paper and I don't :-)

2. No need for extra capital lock up between nodes to cover bidirectional IOUs.

Cons:

1. Channels can be "closed" any time forcing everyone in the chain (could be 7-10 nodes) to make a BTC tx whether they really want it or not.

2. Necessary timeouts require channels to be closed after some time, while in theory some of them can be kept open indefinitely. In my scheme, interior pairs of nodes can pay each other accumulated transaction fees in separate transactions whenever they want without affecting anyone.

knocte• 11 years ago

Interesting! So I guess this means that thanks to the Lightning network, we can have off-chain transactions without the need to rely on centralization (i.e. ChangeTip).

sharpneli4• 11 years ago

Their assumption of 2 channel roll-overs per year per person is quite likely far too low. In their comparison they basically state that a person would be participant in a channel roll over per 3650 transactions made.

This seems to be overly optimistic considering that the acceptable credit card chargeback rate is 1%. 36 times per the rate of channel roll overs.

In addition even nicely working channel has to be created. So even without anything like chargebacks assuming 3650 transactions per year on a single chain seems to be extremely optimistic.

It might reduce spam on the network though. Bringing us from the 240G blocks into 24G blocks (Based on the well known Stetson-Harrison method). A fine improvement, but doesn't really seem to be a silver bullet for the scalability issues.

moe6• 11 years ago

This seems to be overly optimistic considering that the acceptable credit card chargeback rate is 1%. 36 times per the rate of channel roll overs.

That may be true. But even when multiplied with your proposed factor 36 it still looks very promising:

~4.7G blocks, ~108Mbit/s, 252T/yr for full blockchain archival.

In fact, that looks almost too good to be true for a blockchain with 7 billion participants. Did I make a mistake in my math?

Keep in mind that (if I understand it correctly) these are only the requirements for miners ("supernodes"). Mere mortals would largely operate off-chain (and SPV) with near-zero bandwidth and storage requirements.

sharpneli4• 11 years ago

Even if it would help that much the original issue with scalability still remains. It's not feasible for any individual to validate the blockchain. Maintaining a full node is going to be expensive.

So even in the best case it would be "Instead of scaling ridiculously badly bitcoin scales only horrendously badly".

A natural solution for this, if we really really want to get cryptocurrencies to work, would be to basically have a cryptocurrency per nation or so, with different PoW function to avoid shifting asics around. The only reason Bitcoiners want a person in Argentina to keep a Finnish potato sack purchase in their books is just to pump up the value of their 'investment'. Every solution made in relation to Bitcoin is made from the viewpoint that if it would make any current Bitcoin holders potentially miss out on global domination it's ignored. Even if it would actually work better.

pash2• 11 years ago

If I'm reading the slides right, longer chains of transactions on a side channel do not increase risk, so it's innappropriate to imply that a more heavily used side channel introduces risks comparable to credit-card chargebacks.

Each party to the side channel has a multisig transaction signed by the other parties involved, so if one party cheats (or fails to comply for innocuous reasons), then any party can shift the entire chain of transactions to the blockchain without risk. So the only reason you don't want side channels to persist for too long is that each party has to keep around more and more information as the number of transactions over the channel increases; that's burdensome in its own right, but it also likely means that a side channel naturally decays over time, as the probability increases that one of the parties involved drops out or loses track of information. So you don't want to roll over side channels too many times, but it's not a matter of risk of trust.

This really is a very simple and elegant solution. Basically, it recognizes (1) that cryptographic verification of intermediate transactions only needs to be shared among the parties involved, rather than among the entire network, and (2) that any sequence of transactions among a sub-network of parties can be considered as a string of intermediate transactions. And (3) it requires no trust because the entire sequence of transactions can be moved onto the blockchain by any party at any time.

sharpneli4• 11 years ago

Sure. I simply meant that the assumption that an average person participates only in 2 chains per year and performs 3650 transactions in one chain is really far fetched.

Consider the amount of transactions you do. Basically every party becomes a single side channel which requires some activity in the actual blockchain. It would reduce the amount of transactions that are the style of "My local supermarket which I visit every other day" but not anything else.

late2part• 11 years ago

Sorry, you lost me at:

The SQL Database Model ○ Very scalable, very fast

rnbrady• 11 years ago

lol

imaginenore• 11 years ago

Pretty much every modern DB is capable of thousands of transactions per second on moderate hardware.

bhouston3• 11 years ago

Interesting parallel to the growth and development of decentralized P2P networks around 1999/2000.

First there was Gnutella (as Napster had a centralized directory) - but it didn't scale, although it proved popular.

Then they added support to Gnutella for spoke/hub models, which had superpeers that reduced the load on most individuals. But in the end Gnutella never really took off.

What replaced Gnutella was bittorrent, which actually only did transfer via P2P but the searching and discovery was centralized again PirateBay and Trackers.

But for all bittorrent's success, Netflix (and others), which are centralized and highly controlled alternatives, are far more popular and commercially successful for watching online video -- it is just less work to use Netflix, even though it doesn't have the benefits of being free and open.

So for all the attraction of P2P decentralization, often centralized, commercial focused easy solutions win out. There are costs to be paid by being decentralized and often general people who are not ideological do not want to pay these costs.

But I am mostly an outsider with regards to Bitcoin so who knows.

flatline• 11 years ago

While there are certain advantages to netflix's distribution model, such as my ability to stream directly to a tablet or AppleTV, the main advantage in my mind is that it is a legal means of acquiring content with ease. If BitTorrent allowed me to pay for and legally obtain content, I'd probably use it quite a bit.

The inefficiencies in the Bitcoin protocol are pretty big by comparison; every full node has to maintain a copy of the entire blockchain, the proof of work scheme is terribly redundant and wasteful. Some partial solutions (peercoin, ripple/stellar) have arisen in the crypto space but they just don't have the popularity of Bitcoin, it'll be interesting to see where the field stands in even five years. The real world demand for decentralized digital currency is still questionable in my mind.

bhouston3• 11 years ago

The real world demand for decentralized digital currency is still questionable in my mind.

This is a key point. I believe that general people do not care about decentralized currency, what they want are easy payment solutions. If bitcoin doesn't provide easy payment solutions that are cheap, convenient, works everywhere, then it will be beat by something that does provide that.

Even if you could with huge engineering effort slowly move Bitcoin to cheap, easy, everywhere, it is likely going to be beat by non-decentralized solutions because those non-decentralized solutions are cheaper to create, deploy and maintain -- also easier to monetize.

I think that the future is to companies like Stripe, ApplePay -- technologies which focus on customer capture via ease of use and many options and being agnostic towards the underlying technologies.

Bitcoin is neat, just like Gnutella was, but being neat doesn't correlate with commercial success.

AnthonyMouse2• 11 years ago

The real world demand for decentralized digital currency is still questionable in my mind.

My feeling is that the interesting thing about Bitcoin as a currency is the pressure it puts on centralized payment systems to compete with it. The problem with banking is that it's so hard to innovate, primarily because of the regulatory burden both making it very expensive for anyone to enter the market and severely constraining what can be improved if they do.

For example, with credit cards the burden of fraud is put on the credit card company. That turns out to be one of their major costs, even though the credit card company is hardly the lowest cost avoider for credit card fraud (they're just a convenient deep pocket), which requires them to charge high transaction fees to cover the liability. Hence no micropayments; fees too high.

Now suppose that Bitcoin disabuses people of the notion that the payment processor should be the one who takes the hit for fraud. If you don't secure your secrets then you lose your money; you can buy insurance if you like. Suppose regulations get passed to allow this for Bitcoin. Now non-Bitcoin payment processors can make the argument that they should be able to offer a centralized service with the same rules, and we can finally have micropayments.

lectrick• 11 years ago

behold, bittorrent bundles

https://bundles.bittorrent.com/

woah• 11 years ago

I think the reason that Netflix is winning out is not a technical reason, but a business reason. Since they are centralized, they are able to capture revenue from the people using the service. This allows them to hire developers and designers to make the service as good as possible and available on many devices. Also the fact that their founders aren't constantly being jailed is helpful too.

dualogy• 11 years ago

Funnily enough bittorrent still "wins out" in the handful of rogue countries where it isn't criminalized

danbruc2• 11 years ago

BitTorrent is not criminalized, it is perfectly good and legal. On the other hand using it to obtain copyrighted content like movies, music and software obviously is a crime.

jMyles• 11 years ago

I use bittorrent for legal purposes every day. It's the best way, for example, to obtain daily episodes of DemocracyNow!

celticninja6• 11 years ago

or where netflix is unavailable (i.e a lot of countries)

willejs• 11 years ago

"But for all bittorrent's success, Netflix (and others), which are centralized and highly controlled alternatives, are far more popular and commercially successful for watching online video -- it is just less work to use Netflix, even though it doesn't have the benefits of being free and open."

What about skype, that was built upon supernodes, and was decentralised?

bhouston3• 11 years ago

Skype was never radically decentralized -- it was a managed centralized/decentralized network. Its decentralization was to offload costs onto users to provide a cost advantage compared to running a lot of centralized hardware, but otherwise it was centralized.

Skype accounts were centrally registered and a single company was responsible for network health, security, etc.

celticninja6• 11 years ago

Im not sure that Netflix is the preferred option. You say

(Netflix et al) are far more popular and commercially successful for watching online video

File sharing versions are not aimed at being commercially successful so i think that is an irrelevant comparison.

Also you I expect are from the USA, netflix/hulu etc arenot available everywhere so they are certainly not more popular than the alternatives in countries where they are not avaiable.

Popcorn Time was a free, decentralized Netflix alternative and it was hugely popular, only that the developers pulled the plug on it did it not become the defacto method of watching movies for free online.

Anyway not sure how these compare to bitcoin but I dont think your analogy holds up even when not compared to bitcoin.

dublinben• 11 years ago

It's worth mentioning that PopcornTime[0] is still available at a new domain, and has been forked by at least one group of developers.

[0]http://popcorntime.io/

timdaub• 11 years ago

Good point.

I think the market will always decide which media consumption method it wants to use depending on how frictionless the experience is.

In the case of Bittorrent, as a mainstream consumer you'd have to know what program to download, where to find a torrent file yada, yada yada, while on Netflix you can just enter your credit card information and watch as many movies as you like.

I think recently, Facebook also realized this, which is why they are currently aiming towards bringing a great search engine to their customers.

Once that is in place sufficiently, homepages for small businesses will become obsolete, as people rather search for a company/product on Facebook than on the internet as it is a much easier to do.

This same principle again will probably apply to Bitcoin and digital payments in general. Right now, Apple is doing a great job reducing friction in payments, at least for mainstream consumers.

sp3324• 11 years ago

Commercial success depends on a solid technical infrastructure. Netflix evaluated Bittorrent vs centralized distribution and went centralized. YouTube did the same. This indicates that BT is not technically better than a centralized model.

But I think the only reason Netflix doesn't use BT is that the content owners are super-skittish about people having a copy of the data on their computers instead of streaming it and throwing it away. So even if BT is a better technical solution, it's politically unworkable.

aminok6• 11 years ago

This is not like Gnutella's supernode solution. The Lightning Network works on top of the Bitcoin network, which would continue being flat, with all nodes being equal.

IkmoIkmo3• 11 years ago

The comparison to Netflix is ridiculous. Netflix being awesome is not a function of centralized being better than decentralized, it's a function of torrented content being illegal 99/100 times and thus it's very tricky to commercialize it, apart from (shady) ad networks that accept the mere indexing of torrent links.

But consider for a moment the popularity of Popcorn time or torrents in general. They're massively popular despite the fact everyone knows it's illegal and knows many countries have histories of punishing illegal downloaders, let alone the big uploaders. And it's not just that they're popular, they work really, really well for something completely free.

Does that mean we prefer using commercial solutions most of the time? It depends. Torrenting tech is monetized through ads but kept free, just like things like gmail or facebook. The difference is that gmail isn't illegal and thus can get venture capital investments, partnerships, build elaborate products (like gmail for business) that interface with online identities (like businesses, phone security) and infrastructures (banking & payment processing infrastructure), have customer service etc. While most torrenting services have to operate in the dark, anonymously, outside of infrastructure, with few partnerships, little investment etc.

In short, I don't think we should try to reason from analogy here. Centralized/decentralized doesn't inherently predict anything.

I think at the end of the day we have to look at the 'per person' scale. The average US citizen makes 2 transactions a day (one of the metrics I saw used in this slideshow, too.) The average bitcoin transaction is 500 bytes. For a global citizen it's likely slightly less than 2 on average considering the US one of the most consumerist societies, but let's say 2 for everyone.

Alright so 1 kilobyte of data per day for everyone in the world.

Is that a big deal? No. It's not. It's still 14 TB of data, daily, but it's manageable.

It requires an 82 mb/s connection to handle that much data. That's for the entire world populations (1 month old babies included!) to average as many transactions as the average US citizen.

Just put that into perspective. We're not talking about this kind of technology today, bitcoin has only a few million users. It's unlikely bitcoin will ever become popular enough to service 7 billion people in the first place, the internet's been around for decades and it's not even close, hell electricity or clean water doesn't even do that. But hypothetically if it happened, we'd be talking about some date that's at LEAST 20 years out from now to get to 7 billion and that in itself is already extremely optimistic. For perspective, it took the internet nearly 2 decades to get to 2 billion internet users.

So to say a 82 mb/s internet connection in 2035 is too big of a deal for what is likely going to be a node run by a professional business like Coinbase. (i.e. I believe bitcoin's infrastructure will be run more and more by professional businesses rather than home desktop PCs, but these businesses will exist in every country on the planet and be decentralized a bit like how ISPs are decentralized only much more because running a node isn't nearly as big a deal obviously, as it's one of many things running on the ISP's network).

Average home internet is about 5 megabytes per second today in OECD countries. And we know bandwidth, like CPU and storage, has grown roughly 50-60% in the past decades and it's likely to average something close to that (like 40-50%) in the next few decades. 50% yoy would be 3300x in 20 years. 5 MB/S would be 15 gb/s by 2035. Optimistic? Absolutely, it's the most optimistic metric based on past data. But let's just say it's extremely likely that reaching 1 gigabyte per second will easily happen by the time a bitcoin node needs 0.082 gigabytes per second. And that's for a standard, average, consumer, home connection. NOT the much better industrial connections that you expect bitcoin nodes will be hooked into by 2035 (if it indeed is so important that 7 billion people use it for 2 transactions a day).

In short, 1 kilobyte of data per person is so little that whether it's centralized or decentralized it doesn't really matter that much. It can be dealt with. Especially when you consider that many of these transactions can happen off-chain, on sidechains, treechains, less secure chains (e.g. to buy coffee or make a 1 cent payment to read a New York Times article automatically instead of being served a 10 second advertisement with a 1 cent revenue per view value, these transactions can happen on less secure channels).

Decentralized/centralized is an interesting discussion, sure, but at the 1 kilobyte per day per person scale, I don't think it's very relevant when it comes to technical scalability. It's relevant for other reasons definitely (like a decentralized digital global banking system being much less prone to anti-competitive banking monopolies and licensing than say a centralized one like Paypal) but I don't think technical scalability is the biggest problem.

Beyond that bitcoin is so nascent and small... I mean I quite like it and think it just might win, but I still feel it's a bit silly to talk about whether bitcoin can scale to a billion users all the time, when its growth rates are really not all that impressive. I mean don't get me wrong, scalability is important to talk about, but it commands such a disproportionate amount of talk. It's like wondering if 50% of the Mars debate was about whether Mars can support a city of a million humans, let's see if we can get 20 people living there first you know?

vamur2• 11 years ago

Best option would be a bitcoin blockchain connected to government run internal blockchain (or similar) systems. That would offload huge amount of trivial transactions, simplify current archaic internal systems and reduce effectiveness of using sanctions for economic wars - which can easily turn to hot wars otherwise.

vamur2• 11 years ago

Netflix is only winning in the U.S and some western Europe states. Given the current political issues and copyright enforcement it is unlikely to win in the East unless it is dirt cheap. And I'm not sure if Netflix has same day TV releases. If not, then it can't compete with same day releases.

EC1• 11 years ago

Netflix doesn't have a fraction of the available media out there. I can't find 90% of the movies I want to watch on Netflix. On any region. You can't use Netflix interchangeably with P2P, they aren't the same thing.

em3rgent0rdr• 11 years ago

But for all bittorrent's success, Netflix (and others), which are centralized and highly controlled alternatives, are far more popular and commercially successful for watching online video -- it is just less work to use Netflix, even though it doesn't have the benefits of being free and open. So for all the attraction of P2P decentralization, often centralized, commercial focused easy solutions win out. There are costs to be paid by being decentralized and often general people who are not ideological do not want to pay these costs.

I would disagree that the reason people use Netflix instead of bittorrent is not because Netflix is "less work" or that "there are costs to be paid by being decentralized". Your argument above is ignoring the legal issues with transmitting copyrighted data that deters ordinary citizens concerned about following the law. I would actually say that Netflix is more work than bittorret, espcially compared to the bittorrent clients nowadays that can stream bitorrented video. Plus in the past netflix always required a bloated plugin (silverlight).

Sami_Lehtinen• 11 years ago

You completely forgot eDonkey 2000 (ed2k) and eMule. eMule got perfectly working DHT implementation including serverless file search. Technically it was much more advanced than bittorrent, it didn't require .torrent files. It did fetch AHCI data from other peers. Only thing it lacked was efficient coordination between download and upload. Overnet (from ED2K developers) tried to fix that with Horde mode, but it failed because everyone was already using eMule. Hordre tried to pick 5 fast peers for mutual trading. Requirement for trackers and .torrent files felt really backwards stuff after using ed2k.

The Gnutella with super hubs was called G2 protocol. Gnutella also utilized GWebCaches for retrieving bootstrap information of other active nodes.

_cpancake2• 11 years ago

Bitcoin either needs tons of bandwidth and storage space, or centralization. You can't have your cake and eat it too.

sp3324• 11 years ago

It takes a lot of bandwidth to verify the blockchain initially, but you don't have to store it all. Once you see that an input has been spent, you can discard the earlier transactions without losing information about the current balances. So you'd only need a few hundred MB of storage space.

towelguy• 11 years ago

This is interesting, I thought in order to use a fully independent client you needed to store the whole blockchain. Which clients support this?

IkmoIkmo3• 11 years ago

Not really. US citizen makes 2 transactions per day. That's 1 kilobyte of data total.

That's really not a problem, centralized or decentralized, for bandwidth, storage or cpu to handle.

Fact of the matter is that a transaction costs some money, like 5 cents, which is in a way the price to transmit 500 bytes of data, verify it and store it. That's economical at any scale. And guess what, the marginal cost is only decreasing due to technology improvements. (moore's, kryder's and nielsen's laws)

The price of 1 gigabyte of storage is about 3 pennies today (retail), for example. And storage is not permanent either, you can prune old transactions (that's an innovation currently in development).

And again, that price is only dropping, in 1990 that price was a few thousand dollars by comparison.

_cpancake2• 11 years ago

But I need the entire blockchain for it to be decentralized, or someone else stores the entire blockchain, which is centralized.

inoop• 11 years ago

While I appreciate the effort, I wonder why people think Bitcoin will gain mainstream adoption in the first place. Besides buying drugs online, what is really the killer app here?

rplnt• 11 years ago

Getting rid of payment processors would benefit just about everyone. Only benefit they have is insurance.

TylerE2• 11 years ago

For consumers that's a MASSIVE benefit.

Perdition2• 11 years ago

You still have to buy Bitcoin. And given its volatility it is something you probably want to do on-demand, so now you have payment processors in the chain again.

s73v3r2• 11 years ago

I have protection from fraudulent purchases when using my credit card. They also give me the ability to reverse a transaction if I am not satisfied. Bitcoin gives me none of this. If I pay someone in BTC for something, and when I receive the item, it's just a box full of rocks, I have no real recourse or ability to get my money back.

mbrochh• 11 years ago

A few billion people who currently have no bank accounts could suddenly participate in global trade.

maxerickson2• 11 years ago

Is it a foregone conclusion that bitcoin would be cheaper and easier to use than a "Cash Before Delivery" service?

Bitcoin beats "Cash On Delivery" because it is trustless, but that's easy to address by having the delivery service allow a remit prior to the shipment.

Note that with CBD, a (cash using) recipient has to deal with 1 party, with bitcoin they have to deal with 2 or 3 (obtain, spend, receive).

Services would be somewhat different, but it still seems like quite an open question whether bitcoin would (in practice) be better than buying some credit from a payment provider (Paypal or Amazon or whatever).

Perdition2• 11 years ago

How are they buying Bitcoin in the first place?

I also doubt that people who don't have bank accounts have private computers. So really the pool is "people with computers and bank accounts in countries so replete with fraud that Paypal and CC companies don't service them".

aminok6• 11 years ago

You can transport value across the world instantly, with no intermediaries. This is a major efficiency gain.

With the Lightning Network, you could transport any amount of value, instantly, for very close to zero cost, allowing things like paying a random untrusted party for WiFi by the second, with incremental micropayments.

Someone12342• 11 years ago

You can transport value across the world instantly, with no intermediaries. This is a major efficiency gain.

Only in a fictional world where my employer/clients pay me in bitcoin.

In reality I need to go to an exchange, convert to bitcoin, then go back to either the same exchange or a different exchange and convert from bitcoin, all in all I've gone through two intermediaries and paid two fees.

For one example if I did that on CoinMill they would charge me $2.59 over market for 1 bitcoin ($252~254 in value) and the same again when I convert it back to dollars. $5.18, that's approx. a 2% transaction fee. TransferWise by contrast (who really are super hassle free) charge $2.88 for one end-to-end transaction, so less than bitcoin(!).

So yeah, your argument is completely broken in the real world. Only in a fictional world where everyone uses bitcoin does it work.

Artemis22• 11 years ago

This is a major efficiency gain.

Do you have an idea of the quantity of power Chinese bitcoin mining farms consume? That's only for the few transactions that the networks processes right now.

Here is a quick computation that shows that every transaction that goes through the Bitcoin network is equivalent to releasing 47kg of CO2 into the atmosphere: http://redd.it/2hiea4.

oafitupa3• 11 years ago

While I appreciate the criticism, I wonder why Bitcoin detractors' main argument is always "it will never go mainstream", while by their definition of "mainstream" (ie. "the average person"), gold isn't mainstream either, and yet it's such a great store of value that even central banks use it.

Animats• 11 years ago

That distributed-trust scheme is so complicated that there's probably a security hole in there somewhere. Ripple tries to do something similar. How's that working out? Are the fancier operations in Ripple actually used?

The Bitcoin transaction rate is currently only about 100K/day, or 1.1/sec. That's only twice what it was two years ago. There's no urgency in solving this problem.

Lightning Network: “Bitcoin Doesn’t Scale” [pdf]

Related Stories