Originally written in February 2024. This edition removes references to services that are no longer available and clarifies their historical status.
In 2015, I began to truly discover the Internet. Soon afterward, I already wanted to build part of it myself.
I was ten and had no clear idea what this digital space could contain. A search—“how do I create a website?”—led me to platforms that made publishing possible before I understood how the network worked. I thought these sites existed almost by magic, as tools available for the common good. I could see neither the companies, the servers, nor the economy behind their interfaces.
Programming came next. Computers capable of retrieving data stored elsewhere in the world within seconds seemed extraordinary. Understanding the web gradually meant removing layers of abstraction: a domain name points to a destination, a browser communicates with a remote machine, and every apparently immaterial service depends on very physical infrastructure.
Building my own services#
That curiosity eventually produced HomeAPIs, a collection of web and cloud services assembled for my own use. The project included email, storage, documentation, a file distribution system, and several tools for servers I operated.
HomeAPIs was not a commercial public cloud. Access remained limited, and some components never moved beyond the experimental stage. Above all, the project let me study what operating a service involves: identity, quotas, domains, storage, distribution, maintenance, and abuse risk.
I also considered custom-built NAS systems. That direction made the cloud less abstract: before selling a service, someone has to buy, power, connect, cool, monitor, and replace the machines that make it possible.
Dependency and privacy#
Building this infrastructure changed how I viewed ordinary services. A web page can load code from many different companies. Analytics, advertising, and authentication tools simplify publishers’ work, but they also multiply data exchanges and dependencies that remain invisible to users.
Not every cookie is advertising-related: some are necessary for sessions, preferences, or security. The problem begins when third-party mechanisms are introduced without a clear explanation, or when refusing them is harder than accepting them.
The same tension exists at the infrastructure level. Self-hosting gives you more control, but also more responsibility. Outsourcing reduces some immediate costs while creating dependency on providers, their prices, their policies, and their incidents.
What remains#
After eight years, the web feels less magical, but no less interesting. Every application reveals a compromise between simplicity and control, openness and security, speed and cost, centralization and autonomy.
HomeAPIs is no longer presented as an available service. It remains an important milestone: the point when creating pages was no longer enough and I had to understand the systems behind them.