[KirinDave]

What's the point of having a chain at all? It's not a unqiue or prigiledged BFT solution. It has some nice but not essential properties.

If you end up scaling off-chain, then one asks why you have the PoW chain and it's phenomenal cost at all.

[zodiac]

> What's the point of having a chain at all?

Different people answer differently, but many would give censorship-resistant currency as a use case. My personal answer is more complicated.

> If you end up scaling off-chain, then one asks why you have the PoW chain

Let me generalise slightly to include PoS, which I greatly prefer but which is also pretty expensive, and suffers from the same fundamental inefficiency problems. Obviously, you need on chain consensus and guarantees so that the payoff matrix for the offchain game is setup correctly so that honest behaviour in the offchain game is a Nash equilibrium. So that when I "pay" you some money in a payment channel you really have received it in a certain sense, and offchain payments are basically as secure as on chain ones.

[KirinDave]

> Different people answer differently, but many would give censorship-resistant currency as a use case. My personal answer is more complicated.

Nothing about blockchains prevents censorship.

> Let me generalise slightly to include PoS, which I greatly prefer but which is also pretty expensive, and suffers from the same fundamental inefficiency problems.

Most literature I've read suggests potentially a hundredfold increase in throughput with Proof of Stake when executed correctly.

The only fundamental inefficiency problem PoS shares, by my reading, is the broadcast mode limitation where every node needs the full chain to be broadcast before it can move forward. However, PoW systems are nowhere near that rate limit.

I'm certainly happy to be shown this is wrong. Got resources?

> Obviously, you need on chain consensus and guarantees so that the payoff matrix for the offchain game is setup correctly so that honest behaviour in the offchain game is a Nash equilibrium.

You're right, the offchain game needs to be set up so that honest behavior is the optimal behavior. But this is where the cart is firmly planted in front of the horse. What you need is not a blockchain, what you need is trust that controls are in place to resolve a dispute equitably. The blockchain is a methodology not for eschewing trust, but contracting it out in a novel way. A blockchain, folks with guns, an escrow service with a good API, international banking: they all serve the same purpose here.

We should use these tools when they serve us. But what I've seen is that for an actual daily-use currency all that's going to happen is a million little token vendors which can negotiate lightning channels to one another will show up and we'll just have new banks that are like the old banks, but with the threat of a chain resolution which may or may not occur. The practical outcome will be indistinguishable from a healthy banking system, won't actually stop fraud, will be subject to the exact same controls that other physically rooted banks are subject to, and also be a lot more confusing.

I think lightning is solid; in that it will work. I also think it reduces a play at fundamentally restructuring commerce into a flashy buzzword pitch to reshuffle the existing players on the board somewhat.

Having been deeply in the (somewhat broken) American banking system, I can tell you that the system desperately needs to be reworked from the ground up and that can only happen with a truly distributed, truly transparent (an often undersold or ignored aspect of Bitcoin but a truly phenomenal property), truly and fundamentally new way of performing commerce.

We don't have that.

[zodiac]

> Nothing about blockchains prevents censorship.

The decentralization of block producers helps. For instance we saw a few years ago that a few payment processors could decide to not support wikileaks donations, meaning that the only non-cryptocurrency way to donate to them was to mail them cash in an envelope. A government seeking to do the analogous thing would need to effectively make the network soft-fork a rule in disallowing payments to wikileaks addresses, something much harder to do.

> I'm certainly happy to be shown this is wrong. Got resources?

I have never seen any legitimate public blockchain claim 100x improvement from PoS over PoW. I'd be surprised if it were more than 3x for ethereum.

One very powerful upper bound in any non-sharded blockchain without validator delegation is that every full validator must process all transactions, and we want validators to be able to do so with relatively weak hardware (e.g. a Macbook Pro). Certainly, a 100x increase in ethereum TPS would violate this upper bound; for instance, the current state size (2GB) is not designed to fit into RAM, and hence every transaction must do a random disk read and write onto a hard disk. The actual computation can be pretty expensive too, e.g. I've seen a benchmark that ECDSA verification takes 500ms on an i5 processor.

The "broadcast mode limitation" you mention imposes another upper bound, but I do not think it is sharper than then validation upper bound.

> The practical outcome will be indistinguishable from a healthy banking system, won't actually stop fraud, will be subject to the exact same controls that other physically rooted banks are subject to

So this is more subjective but I have to disagree here, depending on what you mean by "healthy banking system". We've learned from history that to have banks run on fractional reserve yet not be subject to frequent bank runs, we have to make them hold reserves in some central bank. And even if you just set up an "anti-escrow service" you have anti-fraud and AML laws to comply with. The end result is that there are huge capital requirements to create new banks most places in the world and fraud still occurs; the capital requirements lead to an oligopolistic industry that doesn't actually provide the consumer services it exists to provide very well. This seems pretty backwards to me. I think in the near-term, payment channel hubs will be much cheaper to stand up than banks and the trust model will be the same as on-chain payments: if you accidentally pay someone off-chain you have no recourse. It seems that building fraud protections on top of this base is more sensible.