Hacker News Reader

Comment by senko | Hacker News Reader

What I find curious in that web3 scenario is this:

1. Today's internet is decentralized because people can't run their own web, smtp, nntp, xmpp, servers, WP blogs. Which is true, we can't expect people to run this the same way we can't expect every car owner to have a mechanic shop.

2. The proposals for "web3", to the best of my understanding, require people to run their own blockchain nodes, manage wallets, etc.

Based on our collective experience in (1), how does anyone believe that (2) is a reasonable proposal? Handling these over to someone else will just result in another new set of platforms that web3 proponents so desperately want to run away from.

Not to mention someone still needs to actually run the servers, reintroducing the same complexity from (1).

View on HN

Exactly! And, in addition to that complexity, there are scalability issues that are easier to solve with some amount of centralization. Would you rather build on a platform where you're paying a well known company or a platform comprised of a bunch of anonymous nodes trying to minimize operating costs?

That Filecoin thing is a great example. Yeah, I want to store my mission critical data on a bunch of EoL, consumer grade hard disks that someone is using to participate in the blockchain. /s

And no regulation isn't the utopia the crypto enthusiasts are selling. Don't completely unregulated systems typically end up dominated by the richest, most ruthless, least ethical participants?

It is not some amount of centralization. Our experimental university work sees all startups cheat with servers. They still sell it as Web3. If you run the servers, you need to police the service! You simply become Web2 again. It should be self-organising and pure, like Bitcoin.

Scientifically Web3 consists of: peer-to-peer technology (e.g. Bittorrent) with a distributed ledger (e.g. Bitcoin) and the social practice of sharing (e.g. Github). This will take 10-15 years more to develop further. The startups are very noisy and are now discovering that doing real "distributed systems" is very hard. They consistently cheat by using central servers. Adversarial Byzantine attacks are unsolved (or require wasteful proof-of-work to yield no performance).

> peer-to-peer technology (e.g. Bittorrent) with a distributed ledger (e.g. Bitcoin) and the social practice of sharing (e.g. Github)

2 out of 3 of those things (minus distributed ledger) has been around for about 20 years now. Y'know, The Pirate Bay and others?

Some of this just requires the right wrappers/UX and greasing from hardware manufacturers.

There was another thread in the last day or two about someone asking where to get a dumb TV. In that discussion it became clear to me that there's already mesh networks all over the place without people realizing it, or if there's not, there could be. The only difference is it's being used by Amazon, Samsung et al to connect their devices and talk about you than to enable you to do something on a peer-to-peer service. I'm not saying it's easy, or that it could happen, but it seems like if peoples' routers or IoT hubs or whatever came with P2P stuff in the box with an easy UX it could move things along?

I generally think IoT/home security hubs/smarthome hubs etc are plagued by vendor lock-in, and I wouldn't be surprised if someday it got dropped in favor of something more standard. At that point I could see P2P maybe coming baked in? Not sure.

Twitter is basically RSS for the masses, cloud is just remote servers for the masses, I suspect decentralized/P2P for the masses is just a UX problem too.

Lots of things happen in steps too, from more hardcore tech crowds to the general population. Email was once the realm of college students and defense people only, and the web was a curiosity.

Others can run the nodes. The point is the network of nodes are distributed and no node is more important than any other. They are all just peers.

You can make the same arguments but with servers instead of nodes. That's how things are today. But the web isn't decentralized.

No, today we have servers with some form of leader or centralised point of failure.

With a properly decentralised network any node can be taken down without taking down the entire network.

Not all public DLTs are born equal however. Leaderless networks which are not vulnerable to DDOS attacks are the dream.

What is this single point of failure? You can take any server down on the internet and it will be fine. The DNS may be a more vulnerable point, but still, lots of different organisations have DNS servers on lots of different places.

>What is this single point of failure? You can take any server down on the internet and it will be fine. The DNS may be a more vulnerable point, but still, lots of different organisations have DNS servers on lots of different places.

Not single points of failure, but a few major (AWS, GCP, Azure, Cloudflare) points of failure.

Take any one or more of these down and there are major disruptions in commerce, connectivity and communications.

Sure, there are other hosting providers, but there's far too much centralization of Internet resources on those four platforms.

Any real decentralization solution will require ubiquitous (multi-)gigabit symmetric links to/from the Internet on commercial and consumer connections.

Because if you don't have enough bandwidth, you'll need to host your content/resources in "the cloud" (read: someone else's servers). And that locks all of us into centralized platforms.

tl;dr: If you want real decentralization, you need to provide the resources to do so at all levels of the networking/application stack.

Edit: Fixed usage: Symmetric vs. synchronous.

According to wikipedia:

> A single point of failure (SPOF) is a part of a system that, if it fails, will stop the entire system from working.

Internet doesn't have a single point of failure. People that depends on X with X being the major clouds may be considered a point of failure for your specific application, but that's just your application, not the internet itself. Internet is already decentralized by nature.

> Sure, there are other hosting providers, but there's far too much centralization of Internet resources on those four platforms.

Because it's convenient. Decentralization has a constant cost. Centralization benefits from economics of scale.

> Because if you don't have enough bandwidth, you'll need to host your content/resources in "the cloud" (read: someone else's servers). And that locks all of us into centralized platforms.

Peer-to-peer solutions work fine, and you don't need multi gigabit synchronous links for that.

What Web3 seems to be trying to do is to have the convinience of centralization with the resilience of decentralization. For example, distributed authentity and payments. That's a great and noble goal! But I wish people would come forward about that and the tradeoffs involved instead of just talking about decentralization. Internet is decentralized by nature.

Note that I said:

   Not single points of failure, but a few major 
   (AWS, GCP, Azure, Cloudflare) points of failure.

And I most certainly didn't say or imply that such failures would "take down the internet." Rather, I said:

   Take any one or more of these down and there are 
   major disruptions in commerce, connectivity and 
   communications.

That's very different from what you appear to think I said.

>Because it's convenient. Decentralization has a constant cost. Centralization benefits from economics of scale.

But there are costs to that centralization too. Not measured in dollars, but in freedom, creativity and choice.

>Peer-to-peer solutions work fine, and you don't need multi gigabit synchronous links for that.

Do they? If that were the case, we wouldn't have these behemoth tech corporations.

Decentralized solutions exist, but aren't currently viable for a variety of reasons. I chose to pick on asymmetric network links (the asymmetric part is the important bit, not necessarily total bandwidth, although that's important too) as a blocker to decentralization.

Feel free to disagree, but decentralization needs to be about more than just distributed block chains and services hosted on centralized server plantations.

Edit: Fixed usage: Asymmetric, not asynchronous. Sorry, still drinking coffee here.

You mean like e-mail? Sadly, most people use gmail now and there are whitelists so you don't get to write your own email server.

Similarly, basically everyone using Ethereum just uses Infura nodes [1]. Web3 is no more decentralized than anything it purports to improve upon.

[1]: https://infura.io/

I feel email could really do with disruption.

Least of all: require escrow-held-micro-payment to send an email to some location. If the item is marked as spam, the sender loses their money (perhaps to the spam-filter maintainer).

decentralization of mechanical infrastructure is irrelevant. decentralization of human control over said infrastructure is all that matters.

one does not correlate with the other.

if anything, the strongest incentive is to mask centralized ownership behind a screen of decentralized infrastructure.

a culture of anonymity helps discourage inconvenient questions about transparency and thwarts attempts to formally analyze the relationship between agents and principals in a blockchain network.

all cryptocurrencies are thus fundamentally based on deception, as the blockchain is designed to expose activity statistics that demonstrate 'decentralization' of agency, but anonymizes the linkage between agent and principal, making those statistics worse than meaningless.

> Based on our collective experience in (1), how does anyone believe that (2) is a reasonable proposal?

The people who are invested in crypto and think it's the future. They do the same reasoning all the time: crypto can be used for X, so X will be done now! Thinking about why X wasn't done before is not something they do.

> The proposals for "web3", to the best of my understanding, require people to run their own blockchain nodes, manage wallets, etc.

AFAIK, the proposal is to use federation, i.e., only those who can/want run servers, whereas normal users use those federated servers.

It doesn't change things. There were federated alternatives to most centralized sites you see today. I remember Mastodon for Twitter, I'm sure there are more. The point is that web3 is not adding anything new here. We could always have gone to federated environments, but turns out it's more complicated than it seems and people don't care too much about what's inside as long as it works.

Did you read the article?

The lesson is that when you compare centralized and decentralized systems you need to consider them dynamically, as processes, instead of statically, as rigid products. Centralized systems often start out fully baked, but only get better at the rate at which employees at the sponsoring company improve them.

Centralized platforms often come bundled at launch with compelling apps: Facebook had its core socializing features and the iPhone had a number of key apps. Decentralized platforms, by contrast, often launch half-baked and without clear use cases. As a result, they need to go through two phases of product-market fit: 1) product-market fit between the platform and the developers/entrepreneurs who will finish the platform and build out the ecosystem, and 2) product-market fit between the platform/ecosystem and end users. This two-stage process is what causes many people — including sophisticated technologists — to consistently underestimate the potential of decentralized platforms.

I do not agree too much with that part. The fact that there are compelling use cases at launch is independent of centralized/decentralized only depends on who's launching. Google Wave was a centralized platform and the use cases were a mess, and blogs with RSS were a decentralized platform with a pretty clear use case. I don't think the percentage of half-baked use cases is too different between centralized/distributed. It also conveniently forgets several issues:

- Building a decentralized platform is technically more complex than a centralized one. So it requires more developer resources for a feature that users don't care too much about. It's also harder to iterate, because you're spending much more time in building a design consensus for new features instead of just rolling it.

- Centralized platforms also have to work out the product-market fit between platform and developers. Mobile? Twitter apps (the Twitter API was a big reason that drove adoption at the beginning)? WP blogs?

- Decentralized platforms will have technical limitations and hurdles that will probably hurt user experience. For example, harder discovery, diverging features, delays, data consistency...

So when he says "The question of whether decentralized or centralized systems will win the next era of the internet reduces to who will build the most compelling products", that's true, but avoids mentioning the fact that, resources being equal, compelling decentralized systems will be considerably harder and slower to build. And this is not new, which is the point I made in my comment. Nothing about crypto changes these facts.

people running their own X is a solution (or means) to a problem, not the problem (or ends) itself. The issue is control.

The problem is lock in. Running proprietary software still has the problem, even on your own hardware. The idea is to have everything run to spec/protocol, with mechanisms preventing customised behaviour - the issue is similar to the problems within bit-torrent wrt custom clients not behaving correct i.e. not maintaining ratio, which meant there needed to be a decentralised mechanism for managing this across-network.

The idea is it doesn't matter so much who runs the software or what they run - the network protocol maintains the peace.

the new generation of middlemen surrounding this have already been funded by VCs and have started putting themselves in the midst of all crypto stuff, it already happened

Reply on news.ycombinator.com