Sven Slootweg (soft-deprecated) @joepie91@pixie.town

(This has been true for at least 7 years by this point, probably longer)

Since I realized that probably not a lot of people know this: Google does network-level (ie. TCP/IP-level) fingerprinting to detect automated requests, in addition to HTTP fingerprinting and more 'traditional' methods such as cookies and IP tracking.

I only know with *certainty* that this is used for scraping prevention on the search engine, but I would not be surprised if this were used for user tracking as well, given Google's history.

There's a lot of common advice about things like API stability and upgrade paths and whatnot, but that sort of advice tends to be very "draw the rest of the fucking owl" - the recommendations above are how you actually *get to* those kinds of design properties without burning yourself out.

(You can go further and have a fully-decoupled 'constellation' of unopinionated libraries, or take a standards-first approach to every component, but those sorts of approaches require a lot more upfront work and practice than the things described above - I definitely recommend working toward them, but start with the above or you'll never get anything out the door!)

How to make a framework or library that doesn't end up producing applications that are miserable to maintain or extend:

1. Make every component optional to use and easy to bypass.

2. But make sure that the lower-level mechanisms that those components use are also directly addressable/usable.

3. Leave deliberate gaps in your API design for bespoke application-specific things to go.

4. Do not make assumptions or structural demands of user code unless you're also willing to design abstractions that ensure them on behalf of the user.

5. Decouple the parts of the framework as much as possible, and treat them as mostly-separate libraries that just happen to work well together, preferring standardized data representations between parts where possible.

That's really all you need to allow users of your framework/library to do their own thing and swap out the bits that don't work for them, but so so few projects actually do these things...

"And it means we have to be intentional about where we invest our time and resources so we can make the biggest impact. [...] With that in mind, we’ve made the difficult decision to phase out two products: Pocket, our read-it-later and content discovery app, and Fakespot, our browser extension that analyzes the authenticity of online product reviews. [...] As we wind them down, we’re looking ahead to focusing on new Firefox features that people need most. This shift allows us to shape the next era of the internet – with tools like vertical tabs, smart search and more AI-powered features on the way."

What is *wrong* with you people???

This is not the first time I'm saying this but it probably bears repeating: the foundational technologies underlying the web are about as decentralized as they can be, with the current state of technology.

The reason the web is heavily centralized has absolutely nothing to do with the technology, and everything with politics, capitalist ideology, and the social dynamics that result from those.

If you want to "decentralize the web again" then you need to acknowledge this, or you will just be working into the void.

Also, what happened in operating system land in December of 2023? There's a sudden spike in marketshare then...

A little over a year ago, the global Linux 'marketshare' according to StatCounter surpassed 4%. Those with an activist background will see where I'm going with this, I think.