Theming was HUGE in Linux back in 2005-2010, consistency was paramount, adherence to toolkit was extremely important just like Apple claimed.
For Apple it was about delivering a consistent branded experience, for the GNOME and KDE communities it was about being able to consistently customize the experience.
In Windows? Always a free-for-all of such heavy customization on the side of the third-party applications that a Windows look and feel was never really a thing, nor was theming because of the same thing.
Then GNOME 3.0 came and they went for the consistently branded experience approach, and you were left with a black and gray desktop environment that at best tolerated customization.
And it works alright, but Desktop Linux at most you'll get to work alright, the third party app ecosystem is small, without the feeling of ownership and freedom that an end-user can have with their system, customizing not the kernel but the things they're actually interacting with, what's left? An unremarkable desktop environment with few applications, that you'll want to replace with macOS as soon as you have the money.
I agree that theming was much more consistent a decade ago in linux-land. GTK2 + QT with GTK2-style would get you an almost seamless experience.
The "dark or light" theme we have today is an absolute joke by comparison.
Many Gtk-engines were also much faster in terms of pure rendering speed.
Another issue as mentioned is that per-control CSS skinning breaks easily with custom themes. Instead of using system colors devs often hard-code a custom look.
Finally, some programs go for a fully custom theme that doesn't even work properly without. Darktable is one example. Darktable looks cool, but a long time I had a hard time reading it's controls since it didn't respect system's font sizes nor it allowed to change it. The contrast was poor. It now comes with several themes, but nothing matches my system theme, which is more accessible.
As much as I love darktable, I'd skip the custom UI any day.
> Another issue as mentioned is that per-control CSS skinning breaks easily with custom themes. Instead of using system colors devs often hard-code a custom look.
After years of using MacOS, I made a commitment this year to use, support, and develop for Linux on a regular basis.
Starting off with little knowledge of GTK, I progressed from "hello world" to working on my first app, but end user customization has always been close to my heart, so naturally I started looking at what it takes to bring a GTK desktop app from "stock system UI" to "developer and user themable".
All this to say, it could be my inexperience, but I'm finding that GTK seems to be very much "all or nothing" here. I can use all the default widgets and be 100% native, and I can "* { background-color: pink; }" my way into a blank canvas, but if I want to make custom controls that build on the user's system theme and whatever accessibility he/she has set up for him/herself, I'm on my own to make my best guesses.
There's no reliable way to determine whether the user is scaling text, using a dark or light theme, or something super high contrast for accessibility. I can try to query some built in widgets and make decisions from there, but I've found that quite flaky as well.
Moreover, even finding which classes to assign to widgets to "piggy back" off the common system colors when building my own widgets is a chore of hunting through themes like Adwaita to find the piece of the system widget I'm trying to utilize. It's not quite WPF "copy the entire widget's XML and re-implement it from scratch to customize it" bad, but for as powerful as the CSS support seems to be in GTK, it feels like there's a layer in between "full system UI" and "total rebrand" that's missing.
I would suggest against trying to piggy back on built-in CSS styles; usually with the default widgets you want to favor composition over inheritance. But if you really need to you can use the GTK inspector to look at the styles on any given system widget, that should be a bit easier than grepping through the CSS.
> Another issue as mentioned is that per-control CSS skinning breaks easily with custom themes. Instead of using system colors devs often hard-code a custom look.
Ironically, you used to be able to use system colors via CSS. Like it usually happens on the web, that feature was also removed because of security reasons - apparently it made it easier for scammers to render fake system popups.
I still use gtk2 style for Qt to this day, as it's still supported in Qt 5. I run i3 as my window manager, and most of my applications are either Qt or still using GTK2 (like pcmanfm).
Reading this discussion got me thinking. I've been using KDE now for years specifically because every part of it can be customized, but for the most part, i've used the same theme and configurations now for pretty much the entire time i've been using kde.
I still think it comes down to branding though, even with those totally customizable environments. The difference is, it's not someone else's brand. I get to turn my environment into my own 'brand'.
That's always kind of been the appeal of theming and customization to me, the ability to remove someone else's brand from my computer and replace it with my own.
the ability to remove someone else's brand from my computer and replace it with my own.
In Windows 2000 and earlier, before system file protection, it was possible to use a resource editor and/or a specially crafted BMP file to change system icons, boot animations, start menu imagery, login screen, etc. With a hex editor you could change any UI text, too. I had a heavily customized W2K alongside my heavily customized Linux (even started working on my own graphical boot animations and graphical login for the text console as part of a media-focused distro project I had joined).
I remember doing the same for Windows 2003. Got a little more difficult with the theme engine and having to use cmd shells running as SYSTEM but the end results were so satisfying. Sigh
> Then GNOME 3.0 came and they went for the consistently branded experience approach
This happened way before GNOME 3 was a thing. GNOME 2, Xfce, and others stopped their obsession with Windows and started thinking about OS X, which was eating the desktop UNIX lunch.
> Eh? Theming was fizzling out by 2005, which starts to correspond to the rise of OS X and Macbooks.
I don't know man, it's what I saw as more and more people came to Ubuntu with Windows Vista flopping, and they came with expectations of a flashy UX and a dynamic growth of the platform. Lots of theming, lots of hopeful mockups and GTK+ struggling to support transparency as theming engines hacked it in, the race to incorporate compositing and having everyone's graphics cards actually work.
I get it that there was a lull during the transition between GNOME 2 and GNOME 1, many of the myriad themes you found on repos in 2005 were ports of GTK 1.X themes, no more. But the influx of Windows users injected a lot of vitality into the theming scene.
Precisely, long-since retired technology. The Windows XP era lasted 8 long years, as people for the most part skipped Vista. It just wasn't very prominent and crazy custom UIs were rampant.
Enlightenment was awesome in the early 2000s, even having a translucent terminal feature. Then the Gnome project started giving the giant middle finger to their users and began removing all the features to use for customization. I remember the day when a Gnome update decided to change key bindings unannounced. Xchat flipped from Unix X11 key bindings to Windows bindings. The argument was that they were going after "Aunt Tilly", the mythical desktop Linux user who never actually appeared. This drove those of us who were developers away. I can't stand Gnome these days and use Openbox / xfce instead, 2 desktops that don't try to drive their users away. Sadly, Red Hat still funds Gnome primarily rather than the vibrant situation we had back then.
Just saying, if you're happy with a *box window manager that hasn't changed much in the last 15 years, you were probably not ever the target audience of gnome. I also don't see why those type of projects would need much funding, compared to ones that are actively seeking to expand their userbase beyond developers. (and I would agree that gnome is not perfect and is not the only contender in this category)
It's the loss of features that drove me away from Gnome. First configuration options were hidden, then removed. Window borders became larger, and bloated transitions slowed everything down. Menus were removed and replaced with incomprehensible drawings with no meaning. Keyboard shortcuts made non-obvious. Not at all an improvement over the state of the platform 20 years ago in my opinion when things were much more lightweight and interfaces were more discoverable. A lot has been lost for desktops in the push for an interface that "works" on tablets.
But none of that was configured in one place, with every other program taking its lead from that central configuration. Instead, effort must be made to work through every program individually, customizing its color scheme.
that's how it has always been. There isn't a system out there that does theming where some apps won't be left out. It's practically impossible. Even dark mode on iOS doesn't work on every app. And then there is Android... which is just unspeakably bad at all things UX.
I'm not sure when the last time you've used a Linux desktop is, but today that is mostly not the case. The third party app ecosystem is relatively strong today on Linux, with multiple implementations of Spotify, Discord and other "must have" apps. Since many of them are simply wrapped in Electron and shipped out to the end user, theming these apps are a cinch. If user empowerment is the topic of conversation, you shouldn't be ushering users to replace their systems with Macs "as soon as {they} have the money."
> consistency was paramount, adherence to toolkit was extremely important just like Apple claimed.
That is not my memory. If you were careful to use just programs from KDE/Qt or GNOME/GTK, then this was about right, but every graphics toolkit had its own way of handling theming, and if you combined programs built with multiple graphics toolkits (additionally: Motif, Java, raw X) that would not respect your configuration.
This has been bugging me for years. I was having a much better user experience ~20 years ago with Enlightenment desktop on top of Gnome or a lightweight window manager.
No, it is currently barely sufficient, certainly not great.
The scrollbar in System Settings is impossible to customise (through the GUI). Tray icons and log-off dialogue icons and the task bar appearance cannot quickly or easily customised because due to "theme"ing changing one thing also changes the other.
For Apple it was about delivering a consistent branded experience, for the GNOME and KDE communities it was about being able to consistently customize the experience.
In Windows? Always a free-for-all of such heavy customization on the side of the third-party applications that a Windows look and feel was never really a thing, nor was theming because of the same thing.
Then GNOME 3.0 came and they went for the consistently branded experience approach, and you were left with a black and gray desktop environment that at best tolerated customization.
And it works alright, but Desktop Linux at most you'll get to work alright, the third party app ecosystem is small, without the feeling of ownership and freedom that an end-user can have with their system, customizing not the kernel but the things they're actually interacting with, what's left? An unremarkable desktop environment with few applications, that you'll want to replace with macOS as soon as you have the money.