On Fri, 2006-11-10 at 10:38 -0600, Mike McCarty wrote: > > > >>It would be nice if everything worked out of the box, but the Linux > >>kernel folks have never felt bound to provide stable kernel interfaces > >>for drivers, leading to constant and routine need for users to source > >>newer versions of third party drivers (if they even exist) that have > >>been updated to the new interface change. > >> > >>Consider: what's changed? NVidia's driver? No. Your laptop hardware? No. > >>What's changed is the Linux kernel interface. Thanks Linus! > > > > > > Having a stable in-kernel API is complete nonsense. It would function > > May I ask how many years experience you have with kernel development > of any operating system? I wrote a small real time OS in 1984, and > have developed or supported four others[*] during the years 1986 through > 2002. I also have written device drivers for the same four, and led > a team of 15 engineers for 18 months doing device driver and hardware > interface/board support code. I have written the initial boot ROMs > for two boards, and helped develop same for five others. > > Oddly, I don't think that having a stable API for device drivers > is complete nonsense. I think you could find a successful real-world example or two to demonstrate that binary drivers are usable in practice. But this is a religious, not a practical issue other than the point that someone supplying binary drivers might not make them available for all the processor types where someone might want them. -- Les Mikesell lesmikesell@xxxxxxxxx
