Hacker News Reader

Comment by keymone | Hacker News Reader

> And don't give me the L2 lightning network bullshit. Both you and I know that it doesn't work as an actually decentralized payments system, and it plausible deniability and makeup for a centralized workaround of BTC's fundamental flaws

seems like you've made up your mind so doesn't make sense to engage, but if anybody else comes across this comment - it's pretty much tribalistic bs. lightning network works, it's a system that preserves Bitcoin's security guarantees while decoupling value transfers from onchain settlements. this is how you scale global distributed systems - you localize them into lots of smaller systems that can change state independently in a consistent manner.

View on HN

Correct me if i'm wrong, but to open a payment channel with another party on the lightninng network (for example Starbucks to buy coffees) you need an onchain transaction. Then when you finally want to settle all of the lightning network transactions (pay your tab) you need another transaction. So you need at least two onchain transactions for every merchant you want to trade with.

The reason this doesnt scale is because for 100 million people to open one payment channel, it will take over a year to process all of those transactions. That wouldn't happen though as the transaction costs would skyrocket to the point people wouldn't bother.

Am I wrong? Genuinely, this is how I see it and why it cannot work.

> you need at least two onchain transactions for every merchant you want to trade with

you absolutely do not. you open a channel once with some node and that channel can be used to send payments to all other nodes.

it's also simplistic view that you need 2 transactions per settlement of a channel because with channel factories you can batch-open-and-close-and-top-off many different channels with single (albeit fairly large) transaction.

systems like that grow organically and people get onboarded via many different methods, so it's not like tomorrow we will have a queue of 100 million people waiting for channels to be opened.

i do expect bitcoin network to hit capacity limit by transactions in future and we will see what it is going to look like, but i rather prefer bitcoin's glacial pace of becoming more and more efficient at ground layer, forcing all sorts of experimentation to happen in upper layers and in sidechains without ever compromising the foundation.

Ok so I've learnt something there, just one transaction to open a channel that everyone uses. But then what happens when you've transacted with say 30 different merchants, and the tab is called? Is that one bitcoin transaction to settle, or 30 since you're presumably sending bitcoin to 30 different wallets?

Personally I don't like the glacial pace of ground layer dev, as it makes me think this stuff can't be solved. Each way you approach it is a compromise.

It’s one transaction to settle any number of in-channel transactions. That’s why it doesn’t make sense to talk about 7 transactions per second because every one of them can represent a billion value exchanges.

It is infinitely more important to preserve stability of foundation than it is to experiment with new features. Bitcoin runs in its current form for more than a decade and that stability is worth a lot.

It's called lightning network for a reason. It's a network of Bitcoin channels. You can open a channel with Burger King and use it at McDonald's. The latter will route the payment instantly to the former, and all channel states will be updated. Read the lightning channel technical papers if you're interested, it's really worth it.

I'm concerned that looks a lot like a more conventional system where you have centralized payment processors. The best-connected ones are most successful, incentivizing acquisitions until there are very few of them or even a duopoly.

What discourages that?

No centralized processors, all channels are permissionless and lightning nodes cannot monitor or censor any traffic.

Right, but decentralized permission less L2 solutions are also technically useless because they generate an unsolved Canadian Traveller Problem

To which the only solution is basically opening a new channel - eg. A new L1 transaction

https://news.ycombinator.com/item?id=28941210

Lightning network style L2 models (where you preallocate channels) simply don't work as I see it.

You either have to centralize the system, or you run into an unsolvable Canadian traveler problem making the thing fairly useless.

Somebody should tell DHL and USPS that they should shut down their businesses because TSP and CTP are unsolvable.

Nah, you’re just wrong because good enough solutions are easy to find, nobody cares about 100% optimal.

It's not even NP-Hard, it's PSPACE-hard.

Try finding the shortest path to a network where you only discover the edges once you reach a node. You only get hilariously suboptimal solutions

Which is why in practice such a system has to be centralized (eg. El Salvador) or it's effectively useless (you resort to opening a new channel rather than passing the money along an unknowable list of channels)

you're overthinking it. and in the meantime LN works great. your analysis is obviously wrong somewhere, but your ideological position prevents you from accepting that fact and re-evaluating your assumptions.

again: LN doesn't need to optimally solve np-hard or worse problems to work, good-enough approximations are easy to find.

It works great at tiny scales between small cliques of people who know each other. And it has for years.

But it can't and never will work at large network sizes without being centralized. And no amount of development work will fix that, because it's a core design flaw.

Which is why the meme with LN was "its 18 months away from wherever you are in time".

It's always in early stages, because it can only achieve early stage levels of usability without devolving into a centralized system. By design.

> Which is why the meme with LN was "its 18 months away from wherever you are in time".

yeah, if memes are the way you keep yourself informed on the topic then i totally understand your point of view.

the reality is that LN works and what you call "small cliques" are just subnets that route payments internally efficiently but are also able to route externally when the need arises.

this is literally how distributed systems are scaled, internet being the largest one.

it would be centralized if nobody could use LN if some one or two largest nodes failed, but the protocol is open and everybody can open a channel with everybody else and nodes in the middle have no way to censor traffic because they don't know neither the sender nor the receiver.

you're simply misinformed.

The internet maintains the very centralized BGP to solve routability

Again, either you centralize or you keep having it be a toy

removing any big LN node will cause disruption only to the channels directly opened with that node, but will largely be unnoticed by the rest of the network.

so unless you point out the centers without which LN collapses, your claims are just bogus.

Reply on news.ycombinator.com