3 ms·
> Isn't the major problem of the current internet is inherent centralization of services because the initial promise of 100% decentralized network is simply too
by imiric 2y ago
> Isn't the major problem of the current internet is inherent centralization of services because the initial promise of 100% decentralized network is simply too complex to realistically manage?
Not quite. The internet _is_ decentralized. What made the web so centralized from the start could partially be the result of lacking tools that made publishing as easy as consuming content. I.e. had we had a publishing equivalent to the web browser, perhaps the web landscape would've been different today. You can see that this was planned as phase 2 of the original WWW proposal[2] ("universal authorship"), but it never came to pass AFAIK.
So you could say that the problem is partly technical. But it's uncertain how much this would've changed how people use the web, and if companies would've still stepped in to fill the authorship void, as many did and still do today. Once the web started gaining global traction in the early 90s, that ship had sailed. People learned that they had to use GeoCities to publish content, and later MySpace, Facebook and Twitter. These services gained popularity because they were popular.
There have been many attempts over the years to decentralize the web, but now the problem is largely social. As you say, we've had the fediverse for over a decade now. How is that going? Are technical issues still a hurdle to achieve mass adoption, or are people not joining because of other reasons? I'd say it's the latter.
Most people simply don't care about decentralization. They don't care about privacy, or that some corporation is getting rich off their data. They do care about using a service with interesting content where most of their contacts are. So it's a social and traction issue, much more than a technical one. The only people who use decentralized services today are those who care more about the technology than following the herd. Until you can either get the average web user interested in the technology, or achieve sufficient traction so that people don't care about the technology, decentralized services will remain a niche.
There is another technical aspect to this, though. Even if we could get everyone to use decentralized services today, the internet infrastructure is not ready for it. Most ISPs still offer asymmetrical connections, and residential networks simply aren't built for serving data. Many things will need to change on the operational side before your decentralized dream can become reality. I think this landscape would've also been different had the web started with decentralized principles, but alas, here we are.
[1]: https://info.cern.ch/hypertext/WWW/Proposal.html https://info.cern.ch/hypertext/WWW/Proposal.html
- pdimitar 2y ago> As you say, we've had the fediverse for over a decade now. How is that going? Convenience trumps everything. All the parts of the iPhone existed for a few years before it -- especially PDAs with touch pens -- but what made the iPhone succeed was putting everything into convenient and easier package. The amount of time worked on thing X has almost zero correlation with its adoption, as I think all of us the techies know. > Even if we could get everyone to use decentralized services today, the internet infrastructure is not ready for it. Most ISPs still offer asymmetrical connections, and residential networks simply aren't built for serving data. While that is true, let's not forget half-solutions like TeamViewer's relay servers, Tailscale / ZeroTier coordinators, and many others. They are not a 100% solution but then again nothing is nowadays; we have to start somewhere. I agree that many ISPs would be very unhappy with a truly decentralized architecture but the market will make them fall in line. I have no sympathy for some local businessmen who figured they will run with tens of millions with $50K investment. Nope, they'll have to invest more or be left out. So there would be a market reshuffling and I'm very OK with it. --- But how do we start off the entire process? I'd beet on automated negotiation between nodes + making sure those nodes are installed on much more machines. I envision a Linux kernel module that transparently keeps connections to a small but important subset of this future decentralized network and the rest becomes just API calls that would be almost as simple as the current ones (barring some more retry logic because f.ex. "we couldn't find the peer in one full minute"). I believe many devs would be able to handle it.