Desktop UIs are hardly mature, but for the last 15 years UI innovation has been about putting lipstick on a pig rather than deep-diving subsystems and figuring out how to graphically represent and control them.
Designers aren't doing subsystem deep-dives, they're just implementing subsets of existing wholly inadequate GUIs using new widgets. Conversely, systems people who understand the un-GUIed subsystems aren't doing design work to better expose them, or expose them at all. It's an unproductive and frustrating stalemate.
Desktop UIs are mature, and were such by the end of '00s, in the sense that they've made the best use of the keyboard and mouse. Sure, it's okay to add subtle new things, like window snapping on Windows (the thing where you drag it to the edge of the screen) or inertial scrolling on macOS. It's absolutely not okay to take a perfectly functional and polished UI and "refresh" it with disproportional controls and huge fonts because people now carry phones in their pockets. The only good reason for a complete UI redesign is change in the way it's interacted with, for example because some novel kind of input device came about. But then again, the mere existence of touchscreens shouldn't be detrimental to the UX of those who use keyboard and mouse.
You're talking about cross-cutting UI concerns (themes, widgets, window system), which I agree are largely mature, while I'm talking about vertical combinations of specific UIs and subsystems, which aren't mature by a long shot.
Example: "delete file" on Windows. Windows has relatively aggressive file locking, but if you try to delete a file that's open, it just errors out, it doesn't tell you what's using the file and give you the option to (force) quit it.
See also: networking, disk management, permissions, sound, preferences, ipc, etc. There are a hundred of these "just needs a bit of work" UI verticals, but designers keep trying to solve these problems with themes and functionality subsets, which are doomed to fail because the underlying issue is a lack of expressiveness in the existing UI, not an excess of it. Meanwhile, systems people live with "just use CLI / sysinternals / wireshark / nmap" like it isn't an issue that bog-standard tasks still require arcane tools in 2020 (arcane by the standards of typical users).
Yeah, in my Windows days, I remember having a lot of small utilities that filled the gaps of the OS. There was one for deleting files too, I don't remember the name, probably unlocker something — it showed you all currently open descriptors for a file and allowed you to close them. As opposed to macOS, which straight up tells you "this file is being used by X.app, close it to try again" when you try to delete an open file.
But this is how it works, which is UX. UI is more about how it looks and how it's interacted with.
They're actually still pretty archaic but no one's cared to do anything about it for decades so we've all been trained to accept the shortcomings. Stuff like background processes grabbing the focus while I'm typing my password, lame modals with an "OK" button, mysterious delays all the time, etc etc etc.
If you mean pixel density, then sure it can be made scalable by using vector graphics or using several different dpi variants of assets. All without changing visual appearance even. Apple managed to do this exact thing in 2012.
> Desktop UIs are hardly mature, but for the last 15 years UI innovation has been about putting lipstick on a pig rather than deep-diving subsystems and figuring out how to graphically represent and control them.
I disagree. To various degrees, the last "15 years of [desktop] UI innovation" have been trying to push trendy touch UI concepts into places where they don't belong.
Designers aren't doing subsystem deep-dives, they're just implementing subsets of existing wholly inadequate GUIs using new widgets. Conversely, systems people who understand the un-GUIed subsystems aren't doing design work to better expose them, or expose them at all. It's an unproductive and frustrating stalemate.