Steffen Evers said: > In particular, Alan Cox has picked out the Nvidea kernel modules as > a good example how a module "taints" the kernel and explained that > such tainted kernels will be not debugged in the future. > > So, what do you know about graphic cards regarding driver's > licenses?
the linux kernel crew doesn't want to support someone else's code if its not available. not many people contact the kernel team directly for such matters anyway(at least i hope they don't ..). you will still be able to get help from lists like debian-user and others that have users which use such modules(i do in 1 system). its a use-at-your-own-risk deal though. i held off for about a year while the nvidia drivers were maturing, they are pretty solid now(under 2.2.19 anyways). theres talk about putting code into modules/kernel that can detect when a tainted kernel is running so developers know more information. i think its a good idea. i wait for the day when the kernel API is stable, same for glibc. i was at a sun conference recently and was shocked to hear that solaris 2.x is fully upwards/downwards binary compadible. a binary compiled on solaris 2.5.1 ~5 years ago will still run on solaris 9 when it comes out next year(or whenever its set to be released). until that time i don't expect linux to be a complete replacment for users on the desktop. its just too much work for companies to compile 50 different modules for their drivers for their hardware, since most end users don't have the knowledge/ability to compile them from source. im just glad that nvidia releases enough source that it can compile on generic kernels, rather then binary-only modules like the ones from companies like adaptec(and others who i cant think of off the top of my head) which in many cases only work with 1 or 2 specific kernels. nate
