Hybrid kernels are mostly marketing.
Linux could be called hybrid, because of FUSE, Xorg ...
Linux is more microkernel-like than Windows.
Porting all the current bug report stuff in Ubuntu to work on top of ABRT probably doesn't make sense, and I bet it would also be a lot more work.
Last edited by JanC; 06-12-2012 at 06:15 PM.
Btw, printer driver example is very wrong. You should have installed printer at least once before you claim something like that.
BBtw, maybe you should learn linux better... before trying to improve it? Just suggesting.
Common desktop Linux machines nowadays have 2-16 GiB of RAM and at least 400GiB harddrives, big majority being multicore.
The correct answer would be "depends on specific linux application segment"
Also, good stable kernel ABI is actually practical. But has seen little use, since that would mean all increase in maintenance cost. But this is actually arguable, as even within current kernel there are low quality drivers, unstable drivers etc.
Last edited by crazycheese; 06-12-2012 at 06:19 PM.
Linux is a monolithic kernel and just cause it has modules doesn't make it hybrid or a micro-kernel.
http://en.wikipedia.org/wiki/Monolithic_kernelLinux is a monolithic kernel. Device drivers and kernel extensions run in kernel space (ring 0 in many CPU architectures), with full access to the hardware
Last edited by linux5850; 06-12-2012 at 06:34 PM.
On hardware, microsoft makes sure drivers are developed exclusively for them. For companies that can get away from that, they (ms) demand at least that release clock is syncronised with windows driver releases - which would guarantee that driver is never more outdated for windows than for any other OS.
On software, microsoft makes sure you not only use only its proprietary technologies, but you improve it for them for free (you mentioned crappy .net, good example of good bullsh!t; also directx).
On user / selling side, microsoft has long history of addiction (where is my start button?) and also Windows is preinstalled on 95% machines via secret agreements with OEMs. Install only windows and get ~70% discount, since everyone does that those who denies have their preconfigured systems at much higher prices solely due to MS software and risk going bankrupt.
You have triangle here. Whoever controls 2 sides of this triangle, controls the third side and makes the monopoly run.
So you run in squirrel cages, microsoft has set up. True developers always take risks, however. They try new things, they never lock up the path and they appreciate independence aka neutrality.
Kernel modules are actually lighter forms of modules that make true microkernel. Linux has much lighter inter-process communication, hence no "server" parts and gets advantages in speed and lower complexity. These advantages however, theoretically, end when kernel grows large enough to maintain, so it just NEEDS to be fragmented in smaller pieces.
This is how I understand it, and it is not thing of "worse"/"better", but more thing of "worse IF USED IN" / "better IF USED IN". Kernel developers seems to be ok with current stand of things, so why bother.
Support for it in LibO would be nice of course.
Last edited by JanC; 06-12-2012 at 06:40 PM.