There is one thing even Jon forgot: even if you purge some parts owned by me from the kernel, you would have to use filtration-abstraction-comparison tests to determine if other people's copyrights are not derived from my work... If they are, independently of the will of the copyright owner, you cannot re-license it, too.
Suppose, as an example and illustration only, that I was the guy who, in 1992 (this is just an example), wrote the original mutex implementation. Suppose, furthermore, I fell out of the surface of the world yesterday -- I went to live in the Tibet, up the Himalaya, and no-one can contact me. Suppose there is not one line of kernel code that I wrote in the current kernel, but suppose, too, that all implementations of mutex are derived from one another, adding things, removing things -- ie. making TRANSFORMATIONS to my code --, till you get to the current status quo.
The current author of the mutex code has no right to relicense it. Period. It's a derived work, and I (as a copyright owner of the original code) have a MONOPOLY on what can be done with ALL derived works. And the only thing I have authorized before getting out of touch is to license them under the GPL.
IMHO, the thing stretches even more: as lots of things in the rest of the kernel are derived works from my work, those can't be relicensed too. More: imagine, graphically, if you "yank" the current mutex implementation from the kernel. You would have to write (clean-room) something that fits in the "void/hole" left behind. But to do this, you would have probably to use other code that is derived on my code... gotcha! Your code now is derived from mine and you have no other option to distribute it than doing it under the terms of the GPL.
Point being: the linux kernel is, for all purposes, un-relicensable in practice. This is a GOOD THING. It's even safer than the FSF social contract in assuring the code will remain GPL'd and won't ever be closed. I explain myself: in the case of a liquidation (if the FSF went bankrupt), the successor could close-source all of the GNU code, because it all belongs to the FSF. Obviously, forks could be made, but the damage would have been done.
To prove me wrong, you would have to make a directed graph of copyrightable chunks of the kernel and/or kernel patches, where the chunks are the nodes and the derivation relation are the edges. Then, if you find any clicks (and I don't believe it for a moment), you can in theory remove that click and rewrite it.
Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds