It's one of those situations where "close enough" isn't. The fine details matter.
Windows isn't the way it is because of some purposeful design or anything. No, it's decades of poor decisions after poor decisions. Nothing, and I do mean nothing, is intuitive on Windows. It's familiar! But it is not intuitive.
If you conform to what these commercial offerings do, you are actively making your software worse. On purpose. You're actively programming in baggage from 25 years ago... in your greenfield project.
The advantage to Free Software is that you don't have to change everything with Windows, Apple, Adobe, or Google demand you do (unless they grab control of a FOSS project, like in Firefox's case.) There are a number of writers who recommend Linux and Free software only for that reason - that once you get a workflow going, you don't want to change it according to corporate whims.
> practically never requires its user to fire up a terminal window
This can be a problem. But it will be less of a problem with LLMs. We need to encourage amateur (and proficient) Linux adopters and users to lean on AI to deal with anything giving them problems. I had an LLM walk me through updating a .deb package in MATE to match HEAD upstream, and to do it in a way that would be replaced when Debian updated the package itself. This is something I've been carefully avoiding learning for a decade, and if I had taken the effort to try to learn, it would be weeks of research and I'd have messed up the system multiple times along the way. Instead, after a few false starts, I did it and gained the knowledge to do it again.
So for example, a hypothetical Windows DE could offer XP, 7, and 10 modes which the user can freely switch between which would never change. This delivers on two fronts: first, it presents a familiar, comfortable UI for the user, and second, it offers a promise that most of the popular Linux desktops do not which is that significant changes will not occur, even over long time scales.
I disagree on LLMs/terminal use. Too many things can go wrong in too many different ways for LLMs to be of much use to users for troubleshooting in many cases, and there's also the issue of the user even knowing what to ask for in the first place (even many moderately technical users aren't going to have the foggiest clue what a Debian package, MATE, HEAD, or upstream are).
The system really just needs to be engineered to 1) be extremely robust and not break in the first place 2) when it does break, have the ability to silently self-heal 99% of the time. A non-essential but excellent bonus would be 3) to be able to express what's wrong and what needs to be done to the user in that last 1%. This won't be easy to accomplish, but the first distro that does will be richly rewarded with user loyalty.
Totally agree. My first distro was Elementary because it was sold to me as Mac-like. It’s…sort of that, but it was enough for me to stick with it and now I’ve tried 3 other distros! Elementary is still in place in my n150 server. Bazzite for my big gaming machine. Messed with Mint briefly, wasn’t for me but I appreciated what it was.
Familiarity is so important.
The DE needs to be as close to a drop-in replacement as possible while remaining legally distinct. The less the user needs to relearn the better.
