- Create a desktop kernel fork. Linus & co. are not in the business of writing/maintaining a desktop kernel. Their goals are larger (and smaller) than that. The desktop kernel can track the mainline kernel, but shouldn't adopt every latest ABI or other changeâ"just do a major update every 3-5 years.
What on earth would that achieve? And what is the difference between a "desktop kernel" and a "server kernel" or whatever.
One glaring example of this is the kernel scheduler. If you ever used the OSX / IOS notice how smooth the graphics is and well responding it is to touch events even under heavy load compared to Linux desktops / Android. Although there have been much improvements done in Linux recently especially Android stack that smoothes the UI experience, occasionally you will still encounter random hiccups in the UI (even with a quad-core CPU). Same is true with other Linux-based such as the Nokia N9 Harmattan, where although the graphics is pretty much smooth, the user sometimes encounters stuttering when swiping or scrolling windows.
As the GP pointed out, part of the problem here is that the Linux kernel have vested interests towards server infrastructures. The current scheduler, CFS (tries to) balances CPU utilization and user interactivity fairly. If Linux were to achieve a more smoother response we need a scheduler for extremely low latencies instead of fair prioritization.
There is a project to change the scheduler towards this low latency approach, but even the project author is not expecting it to be accepted in mainline soon - the reason why we need a fork.