Decentralizing the web

Tim, you made the web fundamentally client-server. Servers have urls, can be indexed, have centralising power. Clients do all the work: run the code, render the page, be subservient. Clients do not have urls, cannot be indexed and cannot communicate among each other (gRPC excepted). Don’t you agree this design has massive centralisation as a side-effect, and how can we fix the design?

He’s actually been working on solving that exact problem:…

Facebooks was a terrible idea. Why Centralize?

decasteve 3 days ago [-]

The concept of Facebook was always terrible. Why do we need a single platform for people to congregate socially? We have a distributed platform built with common protocols and services that we can all use.

Let the web be distributed physically, organizationally, and socially. Let people congregate for purpose and not for the sake of congregation itself.

We are here on Hacker News because of a common interest. I belong to a number of sites fulfilling a way of connecting to people who share the same niche interests. I don’t want everything under one roof or in a single walled garden. AOL failed and the Internet succeeded—stop trying to bring it back because it’s more profitable for those that run it.

.. Facebook’s downfall as a service won’t be another social service that has the same features but is distributed/blockchain/FOSS/quantum/whatever. Facebook has an enormous amount of resources and talent that will allow them to keep their moat in the social networking space.
No, its downfall will come from a paradigm shift in technology that renders the feed-style social network obsolete. This is how most behemoth companies fall, they don’t get outcompeted at their own game, the game just changes around them. Maybe it’ll be VR, maybe it’ll be a virtual social assistant that has no browsable UI but helps you keep in real-time contact with the people you care about, maybe it’ll be a new kind of device or screen or wearable that completely changes the game. But it won’t be another social networking service.

Why decentralized social services fail

>The “why” is not technical.

This is one of thr few essays about decentralization that understands that the problem isn’t technical.

However, I’d go further than his explanations of incentives and say that the fundamental problem is that decentralized technical protocols do not solve the centralization of how money is spent.

Examples of that misunderstanding:

– SMTP the protocol is decentralized (technical) and yet we have giant email providers GMail/Hotmail/Yahoo which is centralized (money). The big providers spent $$$ on 1 gigabyte mail storage + backups + convenience. SMTP specifies how fields are laid out but it doesn’t put money in everyone’s bank account so they can run residential SMTP servers so the email ecosystem stays decentralized.

– Git the protocol is decentralized (technical) but Github the service is centralized (money). Why? Because Git the technical protocol is not a bank fund that gives every programmer a free $10 VPS account to host their own git repo. The centralization of money spent (Github invests in a datacenter but individual programmers do not) results in centralization.

– Bitcoin protocol is decentralized (technical) and yet the phenomenon of giant China “mining pools” emerges which is centralization (money). The ability to spend money on liquid cooled ASIC chips in a datacenter located near the Artic Circle is “centralized” to the entities that can spend that vast amount of money. The exceeds the ability for the home enthusiasts to compute hashes on a spare computer in their bedroom.

The common theme: technical protocols can be decentralized but the real-world implementation of those protocols end up centralized because physical things like cpus, harddrives, network bandwidth, etc cost money.

This pattern of decentralized technical protocols vs centralized economic behavior is ignored by virtually all decentralization enthusiasts.

So the real puzzle to decentralization is, “How do we _decentralize economic behavior_ when everybody doesn’t have the same amount of money to spend?” Nobody I’ve read about so far has figured that out . That includes Sandstorm/IPFS/Filecoin/Mastadon/Diaspora/Ethereum etc.

> The common theme: technical protocols can be decentralized but the real-world implementation of those protocols end up centralized because physical things like cpus, harddrives, network bandwidth, etc cost money.I agreed with you up until here. Not convinced that centralization is due to the economics of running a computer on a network. It’s also a due to reputation and trust.

People use gmail and hotmail, not just because they don’t want to pay for the computer, it’s because they trust these companies will do it properly and reliably.

Github isn’t the best example either. It didn’t spring up because of the economics of running a git server. It got big largely because of its social features, and also because it was a pretty trustworthy git repository. I feel far more confident pushing to a git repo than to some physical server somewhere. There’s a trust aspect.

I’ll skip Bitcoin and got to another mostly decentralized system: paper currency. Exchanging paper currency is largely decentralized, but most people deposit it in a bank, in part due to interest, but largely due to the trust the bank provides.

The common theme to centralized services isn’t the cost of computing hardware, but rather raising trust/reliability in the service, outsourcing expertise, and risk mitigation. Big central service providers are good at exactly this.