On Mon, Aug 11, 2003 at 10:47:13PM +0400, Yaroslav Rastrigin wrote: > Just imagine Linus has managed to include some higher-than-C language into the > kernel, and gave everybody an ability to code one's favourite kernel features > in, f.e., Lisp. Or Lua :-). Not a scheduler or VM subsys, since they are too > performance-critical, but device drivers, FS, some syscalls and such.
There's no point in allowing device drivers to be implemented outside the C code although it would be possible to call the driver code. Also Ion doesn't and will not allow classes/objects to be implemented outside the C core although much of the stuff can be accessed from Lua. > Useful ? Yes, certainly, lots of features could be tested quickly. But what > will kernel look like in two years after introduction of this addition ? Same as before if Linus doesn't start accepting all sorts of scripts into the kernel tree but insists on decent, optimal C code. The fact that people can test things and write their toy scripts shouldn't affect what goes into the kernel. And anyway, a kernel is entirely a different beast than a WM or almost any application for that matter. The kernel would be closer a library than an application. You don't really use the kernel, there's no need for users to be able to script the "feel" of the kernel unlike the WM. OTOH, /proc and /sys are already a sort of primitive scripting interface. > I'm thinking about WM as a kernel for GUI. That's the X server. Hmm... NeWS & postscript... -- Tuomo
