Linux is a fish, stop trying to make it a bird. The fact that a significant portion of the Linux-user population thinks/believes/hopes/expects that it will someday be a bird won't make it so, or do anything to unblock the technical, legal, and organizational roadblocks.
If you want a FOSS desktop OS that can win the "right battles", here's what you do:
1. Come up with a name and a logo. Trademark them. Make a basic set of rules that people have to adhere to if they want to use your logo. Obviously, get a lawyer to look it over to ensure it's ironclad.
2. Fork FreeBSD (or any other open-source-but-not-copyleft-licensed kernel)
3. Pick a GUI layer. GTK, Qt, WxWidgets, doesn't really matter as long as you keep the API stable so programs written in 2028 will run in 2038 (good luck doing that on Linux).
4. Create a driver API interface so someone can write a Realtek Wifi driver once and it'll never need recompiled or updated for a newer kernel. The driver file will work in 2028 and 2038 (of course, excepting the case where there's a new CPU architecture, or a security vulnerability).
5. Stabilize the application-level API as well. That means, probably pick a version of glibc and stick with it forever. Patch vulnerabilities, but maintain backward compatibility as much as possible. Application binaries should work forever.
If your instinctive reaction to these bullet points is to think "who's gonna do all that" then yeah, I agree with you. It's not going to happen.
If you want a FOSS desktop OS that can win the "right battles", here's what you do:
1. Come up with a name and a logo. Trademark them. Make a basic set of rules that people have to adhere to if they want to use your logo. Obviously, get a lawyer to look it over to ensure it's ironclad.
2. Fork FreeBSD (or any other open-source-but-not-copyleft-licensed kernel)
3. Pick a GUI layer. GTK, Qt, WxWidgets, doesn't really matter as long as you keep the API stable so programs written in 2028 will run in 2038 (good luck doing that on Linux).
4. Create a driver API interface so someone can write a Realtek Wifi driver once and it'll never need recompiled or updated for a newer kernel. The driver file will work in 2028 and 2038 (of course, excepting the case where there's a new CPU architecture, or a security vulnerability).
5. Stabilize the application-level API as well. That means, probably pick a version of glibc and stick with it forever. Patch vulnerabilities, but maintain backward compatibility as much as possible. Application binaries should work forever.
If your instinctive reaction to these bullet points is to think "who's gonna do all that" then yeah, I agree with you. It's not going to happen.