The Human Cost of Desktop Linux Extensions

It's hard to make extensions for GNOME and KDE on Linux. Here's why. Spoilers: everybody is right.
Website: trafotin.com
Donate:
✨ Patreon: / trafotin
▶️ KZbin Membership: / @trafotin
Links:
📒 Odysee: odysee.com/@Trafotin:4?r=H3rc...
🐘 Mastodon: vt.social/@trafotin
𝕏 Twitter: / trafotin
📁 Gitlab: gitlab.com/trafotin
🪙 Crypto:
XMR: 84ZpcYxjfkT7uFGXgmi2jH2wyhUBMx8hGBJ3sAp478rKSShMAJHR3DhVVPSwCAskReRBPifzpA5Vu7HPpzAxHUux3SFS4bh
🎵 BGM: [フリーBGM DOVA-SYNDROME / FREE BGM DOVA-SYNDROME]
dova-s.jp/
👋 Outro: Khaim - Neon Lamp
khaimmusic.com
👇 Sauce:
trafotin.com/v/human-cost-des...
Chapters:
0:00 Intro
00:45 The Death of Bismuth
2:12 Polonium & Spaghetti Code
5:20 GNOME's Extension Transition
7:42 Don't Read Every Blog...
10:40 Extensions Are Complex, But So Are Desktops
12:56 GNOME Forge Needs You
14:51 Extension Makers Need You
15:42 Outro

[@fogbank]

More than anything, it's the cost of incompatible API changes. Linus, being the sensible guy he is, breaks user space APIs only in extreme cases. So do the maintainers of critical, strategic system components like glibc, curl, openssl, pam and such. They wouldn't have become critical and strategic otherwise, because you can't build foundations on quicksand. Developers involved in the desktop side of Linux, on the other hand, seem quite happy to introduce breaking changes even in minor releases and to generally chase the next shiny thing and tear down sizeable parts of the environment and rebuild them anew to "improve" it. I believe this is one of the reasons why Linux on the desktop has never been taken seriously by the mainstream: it's hard to make it stable and featureful if you're constantly rewriting it. That is not what the developers of anything that aspires to be a *platform* should do. I would never, ever write a KDE application or a GNOME application because not only their APIs are complex, but I also know that my app will stop working in a few years at most and then I'll have to learn _another_ set of APIs and rewrite parts of it. If it's an extension, I might have to rewrite all of it, or it might not even be possible anymore. Like, "hey guys, heads up, we're about to break your code again!" And I'd be like: "Hey guys, guess what? My code will stay broken, because I'm not being paid to play fetch with whatever stick you pull out of your butt." Boring is good. Boring wins in the long run.

[@Henry-sv3wv]

that leads me to the conclusion: "Why should someone make an app for linux if microsoft values your time more". but if you earn money with software then it's good that nothing is ever finished ^.^

[@fogbank]

@@Henry-sv3wv If you make money _maintaining_ software, yes‌; if you aspire to make money by _selling_ it, at some point you go flatpak yourself or good luck.

[@kuhluhOG]

there is one exception to the list of applications which don't break stuff: glibc they do that pretty much every version for A LOT of their (even non-experimental) stuff

[@mahtja1559]

I've never needed to use the caffeine extension in Gnome. I just disable the automatic suspend and screen blank in the Gnome settings. Do these not work on some environments?

[@JessicaFEREM]

it's more steps

[@lamename2010]

@@JessicaFEREM You do it once, and then it's set. Dunno, finding a random extension in a large list of other extensions, just for my purpose, instead of just looking in my settings, seems to be taking the long route.

[@CotyTernes]

Needs may change throughout the day or across days. Having an extension that you can just click to enable or disable. I'm not a Gnome user, but I would assume there's also quick options to set timeframe for the session of caffeine or bind to while an app is open.

[@shaunpatrick8345]

@@CotyTernesneeds are simple: screen on when you're there, screen off when you're not. Just use Super+L when you leave your desk.

[@CotyTernes]

@@shaunpatrick8345 different people different needs.

[@knightmarSPZM]

I like coding my own Desktop Environment, I'm not sharing it with anyone so I avoid a lot of "what if someone does that" so I can do it in such "fixed way" that I'm able to maintain it by myself and have it exactly to the single pixel and package how I want it ^.^

[@themadoneplays7842]

Yeah, the constant extension breaks is why I hate gnome so much. Sure it does happen on KDE too but not nearly as often.

[@mskiptr]

Extensions - no matter where - are basically always an uncoordinated mess. Unless they have a clearly delineated scope, they will be unsustainable no matter what. Extensions are just not the way to go.

[@kuhluhOG]

it's only a mess when the interface between the extension and whatever the extension extends is not well designed (which tbf, is quite hard, but as we can see in a lot of other projects, can be done; and then again, designing APIs is hard no matter what) and for a lot of systems, you kinda HAVE to design it via extensions, because if you don't, you will end up with such a behemoth that you think the Linux kernel is small in size

[@jamesdim]

The music makes me feel like I'm in a side quest

[@HUEHUEUHEPony]

I wonder if gpt-5 can be used to port api changes, Have a script reading a RSS feed and changelogs, loop over all files that use the changed API, then have gpt-5 generate the code back.

[@Trafotin]

The warning with using LLMs to generate code is the LLM won't respect conflicting licenses. There is a list on the GNOME engineering blog and they seem pretty straight forward and could be done in a few days. Faster with a couple sed commands!

[@CompuB1t]

One lf thing that I want to use Hyprland is because of tiling. I'm KDE user but I like to use tiling and bismuth and polonium, I hope KDE and devs of tiling fix that. And I hope standardization comes to Linux desktop.

[@fogbank]

What do you mean? The "Linux desktop" is chock full of standards. Do you mean "stabilization"? If so, HAHAHAHAHAHA! No.

[@CompuB1t]

@@fogbank This video is basically because devs and maintainers don't follow real standards or same standards. The APIs is something that breaks extensions and needs to be standardized.

[@fogbank]

@@CompuB1t No, it doesn't, because there's no "Linux desktop". Desktop environments are interoperable to a point thanks to t the various standardization efforts with dbus, portals and such, but extensions are incompatible by definition, since they're highly specific to their environments.

[@mohamad20zx34]

I hope pop is gets updated this year

[@user-oz9tn2td3q]

I used to play around a lot with tiling extensions in GNOME and KDE becausebI wanted the stability and ease of use of a DE with the benefits of a TWM...when really I was getting neither. These days I'll just use vanilla GNOME or Sway, which is actually really stable once you have it configured properly. Hell, even Hyprland is pretty stable, it just isn't packaged on Debian yet so it a no-go for my main PC.

[@kuhluhOG]

"Hell, even Hyprland is pretty stable, it just isn't packaged on Debian yet so it a no-go for my main PC." Considering that Hyprland uses a vendored wlroots with custom patches (which makes it incompatible with the normal version and practically is a fork at that point) and most distros have a policy against vendoring (some distros apply exceptions when a software becomes popular enough, Debian isn't one of them), I doubt Hyprland will ever come to Debian.

[@pamus6242]

I'm a wise man. I use what KDE offers and look no where else and after that I get on with my life.

[@sotecluxan4221]

[@Enraiful]

You don't know how refreshing it is to hear level headed takes.

[@Aoitori365]

gnome wishes they were apple so they emulate their anti-consumer polices

[@Trafotin]

Stop treating GNOME like they are a consumer company (I mentioned this in the video). They need time and talent and do everything for very little gain. Some of their devs are opinionated and bad at communicating because most developers don't know how to communicate with people. At the end of the day, there's a person behind the screen and on those git servers, and making comments like this only dehumanizes them more.