There's such a deep seeded, systemic bias against linux that it actually can never win, to any degree or magnitude, because the moment it starts winning we just move the goal-posts for the flimsiest of reasons to ensure it can't quite claim that victory.
Linux is obviously and clearly the most popular operating system kernel on the planet. Oh, no, that's no good a measure, servers are messy, let's refine it to most popular consumer operating system kernel? Oh... it, could also reasonably claim that title? No no, no Android, that doesn't count. Nope, No Chrome OS either, you can't have that, that's, well, that is linux, but its not. Just nice, pure, desktop linux, yes, perfect, arch linux, kde desktop, that'll never trend up and thus is the perfect new-new definition of desktop linu--wait hold up, I'm getting word this is, not possible, its actually SteamOS? Nope, kill it, that's not desktop linux either, kill it.
excuse my ignorance, but are there any major developer tools that don't support Windows? I can only imagine some internal enterprise tooling doing this.
>Ultimately compatibility layers are good because they reduce developer workload so that developers can focus on what really matters
I don't mind them as a concept, but I personally want as few points of failure between me and my software as possible. Some software is already either overly bloated or buggy (or both) as is without wondering if there's now compatibility layer issues on top of it.
Git on Windows is only supported by installing a whole suite of Unix tools and a shell.
Tools like ccache/sccache treat windows (well msvc) as a second class citizen.
Go, the poster child for cross-compilation shatters that illusion when you need to use CGO.
Python, I believe things have gotten better but the last time I tried getting tensorflow up and running on Windows it was a long and painful path involving third party python distributions, native toolchains and changing drivers.