[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
NetBSD Next Logical Step: Microkernel...?
Ok this may seem like a relatively stupid question with easy answer. As you can gather from the title, my question is:
Why hasn't NetBSD decided to change over from a Monolithic Kernel to a Microkernel Design?
I know some people may have the incentive to answer this question with the response that transferring from one Kernel architecture to another is not an overnight job that will just take a few days and one developer, I understand that changing a kernel design takes alot of manpower and in region of a years depending on the number of developers.
The question is more directed towards well the tools and design of the NetBSD operating system. In my view NetBSD's design and tools and what it aims for are nearly a perfect match for the designed modularity of a Microkernel. Both provide something that will probably benefit the other in more ways than one over the long term.
Take NetBSD's build.sh toolchain that allows for NetBSD to simply be built from one CPU architecture to another with very little to no code variation. Something that the Microkernel tries to implement by only having the core drivers etc necessary in kernel space. Then take the RUMP anykernel design which spells microkernel in a nutshell. The RUMP anykernel is the biggest asset of all in my opinion, you've already got the whole IPC, hypervisor etc created and probably wouldn't take alot of work to take it one step further (welcome to be corrected if i am wrong).
The only reasons i can hypothesize as to why one wouldn't change is to compare both mono and micro kernel designs. Which if you were to compare with what we know now about microkernels and what we knew 10 years ago the answer is probably very different as to why you would or wouldn't choose that design, although i don't wont to go into it in case it starts a flame war or whatever.
From my perspective it just seem that moving to the microkernel design would be and probably has been the next logical move. But maybe i am wrong which is why i have posted this...
Main Index |
Thread Index |