Hi! > > > >> Often, in the embedded world, the person defining the operating > > > >> states will not be a kernel developer, and may not be comfortable > > > >> with, or capable of, creating a kernel module. (There are > > > > > > > > That's okay, we can create HOWTO or something. 'System designed is > > too > > > > stupid to hack kernel' should not be an argument. > > > > > > It's not a matter of stupidity. The example I gave of not being > > > able to create a kernel module was the situation where loadable > > > module support is not available. A HOWTO wouldn't alter this. > > > > So they need to recompile the kernel. Not a big deal when you are > > developing new machine. > > > 1 time yes, several maybe, thousands no. > > think about it as finding minimums in a n-dimensional discrete function: > you really don't want to recompile a kernel each time Okay... for that kind of development you probably want custom kernel with sysfs interface. And if that sysfs interface turns out to be nice enough, maybe we even want to merge it. But for now please submit patch with only basic functionality, and keep sysfs interface outside. Getting interfaces right is hard and you do not want to block core functionality because of that. Pavel -- (english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
