>>>>> Edmund GRIMLEY EVANS writes: >> There is no sparc port of the Hurd. The first thing required is a >> sparc port of GNUmach.
EGE> What about the old plan of porting the Hurd to Linux as a EGE> microkernel? What are the relative costs and benefits of EGE> porting the Hurd to Linux, and porting GNUmach to other EGE> architectures? Thomas and Roland thought this was nearly completely insane, but I did look at this more seriously. After a couple of days of thinking and investigation, I have to say that I agree with their conclusions. The work that it would take to add the necessary features to Linux would be much better spent porting existing Linux drivers to GNU Mach, and working on libc binary compatibility. If you really want to work on another kernel besides Mach, investigating Fiasco (free L4, http://os.inf.tu-dresden.de/fiasco/) would be a better idea. If we're going to make a serious effort to port to a new microkernel, why not pick one that will give us better performance? Grabbing existing Linux features sounds like a nice idea, but frankly, I'm shocked at how quickly people like OKUJI Yoshinori and friends are improving GNU Mach. As Thomas put it, Mach and the Hurd is basically reusing most of the Linux kernel code, it's just organized a bit differently. ;) So, that's my opinion: I used to be for the idea of using Linux as a microkernel, but now I have a much greater love for Mach. Kinda like how I used to be for using Red Hat as a distribution, but now I can't live without Debian. :) [BTW, part of my reasoning stems from a discussion I had with a certain top-ranked Linux kernel developer who made it clear that, though he/she wouldn't wish death on anybody, microkernel features would probably never make it into the official Linux distribution, especially if they were only useful to the Hurd, so we'd have to develop our own fork.] -- Gordon Matzigkeit <[EMAIL PROTECTED]> //\ I'm a FIG (http://www.fig.org/) Committed to freedom and diversity \// I use GNU (http://www.gnu.org/)
