Quote:
Originally Posted by XavierP
|
Thank you for moving this question to a more appropriate location.
Minix does seem to have languished over the last few years, but Andrew Tanenbaum just received a substantial grant only a few months ago to continue his research, and the Minix3 project is looking for volunteer programmers to help continue the work.
http://www.minix3.org/news/
I am sure that your are sophisticated enough to understand that Minix is NOT just another knock-off linux distribution, out of the HUNDREDS that are out there.
Minix not only holds a unique position in regard to the development of Linux (because Doctor Tanenbaum's original Minix was instrumental in inspiring Linus Torvalds to create the first version of the Linux Kernel), but also because it has a unique more robust kernel architecture that incorporates features that have not yet found their way into mainstream Linux.
Andy's spirited debates with Linus about the merits of various kernel architectures have contributed greatly to the community, and I have to confess that, more often than not, I have found myself coming down on Tanenbaums side of the debate, and I think the Jury is still out on this one.
Personally, I think that arguing about a single extra layer of kernel abstraction (which would cause at most a few percent performance hit), is really STUPID in the context of our typical UNIX boiler-plate bloat-ware applications with libraries, calling libraries, calling libraries, calling yet more libraries, ad nauseam (not that Windowz is any better).
Consider a Linux GUI App on the desktop running under GNOME. GNOME CALLS GTK, GTK works through X windows, the Xwindows system uses one of several graphics hardware driver layers, and those layers intern make Linux system calls. If the app needs OpenGL then add another layer.
Moving the video drivers, and file system out of the kernel and up the food chain would indeed move them one layer farther from the lowest level kernel code, but, at the same time, also move them one layer CLOSER to the actual application code that is running them in the systems user space.
I think this could be made to be pretty much of a wash, as far as a performance hit goes.
Now, let's contrast this to the fact that the Linux Kernel NOW includes the functionality to dynamically load and unload drivers, which is also known to cause a hit on performance.
This hit can be quite severe when caching static data to the hard drive is necessary, but this arrangement provides NO ADDITIONAL STABILITY WHATSOEVER (actually quite the opposite).
This performance hit for the Linux Kernel was considered ok, but taking a couple percent performance knock (on kernel calls only, most of which could be made up elsewhere) for the worthy purpose of building a MUCH more stable Kernel was NOT considered worthwhile.
Sorry, I’m just not buying that.
I just installed Ubuntu 8.1 on 3 different hardware platforms, and got stupid FLASHY-FLASHY-FLASHY Scroll-Lock/Cap-Locks Kernel Panic!!! on two of the three (Crashing the kernel every time on entry into X-windows on one system, and intermittently during network access on the other).
In fairness, on the same three systems, Both Fedora and SUSE come up clean, with no kernel panics even in several hours of use, but I suspect that this is only because they don't load the same drivers as Ubuntu.
I’m pretty sure that there will be more pain before I get all the needed drivers working on these boxes, and having a kernel that can stay alive long enough to tell me a little more than ‘FLASHY-FLASHY-FLASHY’ would be nice.
Suffice it to say, that I for one would like to see the “great kernel debate” constructively continue, because I think Tanenbaum’s Minix is on the right track, and their is DEFINITELY room for further development, which will hopefully lead to cross-fertilization improvements that can be branched from Minix into the mainstream Linux Kernel.
I do see your point about not supporting a specific distribution unless those that created it want it that way, and provide support for this forum in turn.
If Andy Tanenbaum (or someone from his team) weighs in on this, would that change your mind?
- Delphin