Hi, On Thu, Sep 06, 2007 at 10:47:54AM -0600, Michael Heath wrote:
> Phase 3 and 4 are logical, but I'm confused about your distinction > between Linux Drivers and GNU drivers. GNU Mach already uses a lot of > (old) Linux kernel driver code; this doesn't mean that its 'compatible > with linux drivers', and it would be impossible to make it completely > compatible with all Linux drivers. > > If we do spend the time to port driver code from Linux to > Hurd+microkernel, why do we need to replace that, assuming it works > well? > > I don't think theres a big distinction between Linux Drivers and GNU > drivers in this sense. Yes, drivers need to be created, but the goal > should be to make drivers that work WELL, whether they do that by > using Linux code or not, and not to make an exclusively GNU driver. Well, I'm not so sure about it. Of course, reusing existing drivers would save the Hurd an *enormous* amount of maintenance. Yet we might want native drivers in the long run. Of course, as long as drivers run in a boring old-fashioned framework like they do in gnumach, there is no point in rewriting them. But when we have full control over the drivers, we can do much more interesting things with them: We can create a framework that integrates the drivers better with the rest of the system, running them as more or less ordinary translators. In fact, I believe half of the potential power of the Hurd architecture is wasted without such a well-integrated approach to drivers... -antrik-
