Ok...
I had not kept up with the kernel source that much since most of my boxes run a patched openMosix 2.4.22.(
http://openmosix.sourceforge.net/) My net server run 2.4.26. My "daily driver cluster" (currently 6 boxes) has, well my own homebrewed kernels hack patched from 2.4.22-2.6.6 with only one interesting problem.. When they are all running as one they tend to over-write a harddrive (they together run faster than the drives can sometimes handle) fortunately this is rare and appears on one drive but all its parts...
The formerly available kernel HOWTO which was on the release of 9.1 but has been removed from the present HOWTO package stated the very first thing to do prior to building the kernel was to `make mr.proper` EVEN if the kernel was freshly downloaded and unpacked having never been used before thus being "clean". That is the way I learned kernel building and using that HOWTO I never had problems with 2.4 level kernels.
Which brings me to my wosre issue with 2.6 level kernels. IF you know what IS required to have a working kernel and you know your hardware then you should be capable of compiling a working kernel. With 2.4.# that tends to work fine. The kernel can be so small and packed it can be put on a flash chip for a LinuxBIOS or a etherboot.
But I found that was not possible with a 2.6 level kernel. IT requires MORE compiled into it to work than previous kernels. The 2.6 level kernel is heading toward the total "plug+play" thing. The kernel may soon find itself like the MS Kernel (bloated and far from user friendly) so it will work without user input on any hardware with and GNU OS software. Strangely in an article about a speech from a MS wheel he stated this very thing about the linux kernel and the direction of the 2.6 level was headed.
If and when the muck the kernel that bad (if they haven't already) I will go back to FreeBSD.
