Here's my obligatory take on "why the Semantic Web failed", a time-honored tradition for those hoping to replace it, before inevitably failing themselves. Why this? Why now? What's different?

Domain topology

I'm not sure what to title this section; please help me think of a concise description.

The semantic web envisions data being "consumed at its source", where The People Who Have Data run some HTTP endpoints that serve more data about that data, and that's how you get your data.

But there's a fundamental mismatch between the shape and distribution of data as it originates (people who publish data) and data as it is used (people who use data). Most people who use data also publish data, and the mismatch is there, between each other, as well.

It's sort of in the definition of "use"


The Web changed

Web 3.0 didn't happen because something else became Web 3.0 first

Things are in the cloud now

When the Semantic Vision first manifested its destiny, people still ran their own email servers. If you had data and wanted to publish it, you had a natural place to do so: on your own computer. Ironically, hosting a server in 2019 is a bigger commitment than hosting a server in 1989. You could get hacked! You could get DDOSed! You could go viral and see your AWS bill skyrocket!

Part of this shift is a sobering realization of the impermanence of places and services on the internet. We (correctly) have more faith that Amazon or Google will be around in another thirty years, serving our content, than some random IP address. The web is infrastructure. Of course we want to drive on highways, but fewer people are willing to maintain them.

It's not that the idea of cloud services is incompatible with the semantic web - it's that the specific services that have emerged aren't very semantic. We don't want to use the URLs of files on Google Drive as identifiers because they look gross and we're a little ashamed of using Google Drive in the first place. AWS S3 isn't going to respond

and maintaining that level of organization is a monumental, team-of-librarians-sized continuous project.

Data was never abstracted from its host

Data is still dependent on its host for identity in addition to resolution. Data could only be retrieved if given an identity from some naming authority.

"Hashing" is a great decentralized naming authority for files, but the web runs on "resources" (1990-2010) and "services" (2010-).

The Semantic Web failed because it tied data to its host.

Prescriptivism vs descriptivism: the URL edition

The URL as a concept collapsed under its own success. By catching on as the one true way to address and navigate the web, it fated itself to evolve in sync with it. As the web shifts from a network of pages to a network of applications, the URL's usage-definition is drifting further and further from its dictionary-definition: the URL now means "an opaque application-state container", whether the W3C likes it or not.