An analysis of Bitcoin's throughput bottlenecks

Mind you, I don't mind centralized systems, but the entire marketing of cryptocurrencies for the past decade has been how it is decentralized and outside of the the reach of any entity. We knew this wasn't the case in 2014, when Mt Gox folded and all signs pointed to fraud. We knew this wasn't the case in 2016, when the Dao smart contract was reversed by general appeal. We know it now as mining pools continue to rise. The decentralization is the main selling point of cryptocurrencies, and yet at every turn it is proven false.

Civilization works because of laws that are interpreted by courts and norms which are broadly accepted but not often checked/enforced. Crypto is trying to create a "trustless" future where you do not need society to coordinate and regulate. Our modern financial system needs that squishiness/trust which you get from the legal system to operate at full capacity.

Picture the suspension of a car. In our modern financial world, the legal system and regulatory institutions serve as the shocks and struts of your car. It allows you to drive fast while not having to worry about every imperfection on the road. Cryptocurency is like driving a car where your wheels are attached directly to the axel - yes it's perfectly efficient and there's nothing between you and the road, but that also means you have to be exceptionally careful where you are driving, and you can't do it fast.

There's a reason all cars come with suspensions.

First of all, thank you for being pretty much the only person who actually wants to talk about the paper, and not something tangential : )

> assumptions that are entirely arbitrary and biased toward the argument of not scaling Bitcoin

I do mention in the paper that the assumptions are up for debate, but I did try pretty hard to justify those assumptions so they are certainly not arbitrary. If by "biased" you mean that my conclusions are a factual result of my assumptions, then thank you. If by "biased" you mean that I decided what results I wanted to get, and then chose the assumptions based on that, I really would love a bit more benefit of the doubt. You don't know me, so its a bit strange for you to start accusing me of cooking the books on something I spend literally dozens if not hundreds of hours on.

In any case, I agree that there are other assumptions that may make sense. There are other bottlenecks to growth than blockchain space, and so it should be considered how much blockchain space will allow bitcoin to grow as fast as we need it to grow (for whatever reason we'd want that). This will change over time. The more people adopt it, the more likely blockchain space will become a bottleneck that substantially slows adoption.

But certainly it could be argued that right now we don't need 90% of the earth to be able to run a bitcoin node. I can absolutely see the argument for increasing the minimimum requirements in the assumptions to be much higher in the next 5 years, maybe even 10 years. But in 10 years, I think we'll want to make using bitcoin to its fullest capacity a lot more accessible.

> What you don't want is people in poorer countries having to trust third parties with their private keys

You absolutely have a point! Most people using custodial solutions that do batch things under the hood to get around scaling issues is certainly not a state we want to be in. It may happen anyway, even if we magically found infinite scalability, but we don't want it regardless.

However at the same time, we don't want the rugged pulled out from under us: we don't want bitcoin's rules to be changed out from under us. Enough full nodes need to be out there to ensure that various attacks aren't possible on the network and that blocks that propagate are valid blocks. I think there is some sweet spot on both of those amounts. And I do think more discussion is needed around the assumptions we're making. I wrote this paper in part to start exactly such a discussion about the assumptions and what concrete targets we have for bitcoin in terms of these network demographics.

What would you say is an acceptable percentage of the population using custodial systems like you're talking about? 10%? Less? What about: what's the percentage of the population that would be acceptable to run a full node? 10%? 50%? More? Maybe at some point I'll write another paper just on the different reasonable assumptions.

> Using a dedicated node, with computing resources that the user purchases and utilizes 100% for Bitcoin node operation, is ruled out, again arbitrarily for no reason that would be relevant to a global network attack scenario.

Again, not arbitrary. The reason I did that is to include normal people. Normal people aren't going to buy and set up a dedicated server for bitcoin. And what about a remove bitocin service? Well, that's just the service running bitcoin, its not the individual - so its no better than SPV. So yes, if we're ok with just say 5% or less of the population running bitcoin, we can remove the 10% capacity limit, but if its normal people - people don't like their machine running slowly all the time because bitcoin is chugging through the world's transactions. Most people just would switch to an SPV node.

> tens of millions of people could easily afford to purchase tens of TB of storage for running dedicated Bitcoin nodes, and tens of millions of people is easily enough to make shutting down or controlling the Bitcoin network impossible for any nation-state actor.

Well, yes, they could afford to. That isn't the question I'm concerned with. Its not about cost. Its about convenience. Will 10s of millions of people buy extra hardware just to run bitcoin? I would argue no, they won't.

But yes, you're probably right that 10s of millions of people running PUBLIC full nodes is probably enough. My understanding is that we're at less than 1/1000th of that goal at the moment. Were we to make it easier for people to run public nodes (either via IPv6 or Tor or NAT hole punching), it would make it a lot easier to reach that goal.

And I'm not saying we couldn't bring these requirements up eventually. Once we blast past 10 million public nodes, maybe we can re-evaluate. But as of right now, Bitcoin has a serious vulnerability to sybil attacks on public nodes.

> all Bitcoin needs to be shutdown and censorship proof is that it be possible to run a full node using something on order of 500% of the typical personal computing resources of the richest 1% of the world population

I certainly don't agree with that statement. It means 99% of people would need to run SPV nodes - which is currently not good. First of all, the World's 1% owns less than 50% of the wealth, so it would be a minority of the economic activity. Putting the majority of economic activity in SPV nodes would leave the system quite vulnerable in situations where a majority of hashpower hardforks to something dangerous or malicious. This isn't necessarily a 51% attack scenario. It could be a situation where miners misread the sentiment of users and choose badly, or where mistakes were made, or it could be a situation where it favors miners but not users, or a number of things like that that aren't a single colluding group. So I would say at very least, the majoritiy of economic activity should be using their own full nodes, and a vast majority is far safer. Perhaps this is 5% of the world.

> The political attack, of countries banning Bitcoin internally, and the best defenses against that, are not considered.

I'm curious what your analysis would be of that.

> increase the proportion of the population that is economically dependent on the network

I agree this would be a very helpful mitigation of political attacks in general.

> much more likely

This I'm not sure about. Yes political attacks are likely (and of course we've seen many of them already). However, political attacks that could succeed seem already quite unlikely. I would love to see your analysis of how likely one might be to succeed and what the network requirements would be to sufficiently mitigate that. And then the tradeoffs vs these other kinds of attacks could be compared.