On Thu, 6 Nov 2003, Brian Buhrow wrote:

> 	The question is, how hard will it be to write a device driver
> environment that would actually work?  I imagine certain classes of drivers
> would need to be attacked first.  I.e. USB drivers and then network
> drivers, etc.  Unfortunately, network drivers are probably the easiest, but
> the least interesting because there are already a lot of native network
> drivers for many network cards under NetBSD.  More interesting, to me, at
> least, would be Windows drivers for exotic USB devices.
> -Brian

Yes, the USB is quite straightforward to implement.  One nice thing
is that a typical USB driver does not use any functionality that
need kernel mode calls, so the compatibility layer can be done
completely in user space.

The problem is that an USB driver does not have any value unless
you can make calls into it, so you need to implement e.g. the
network stuff too, in order to us a network USB device.

Most of the exotic USB devices does export proprietary IOCTLs
only, so running that driver is not of any use unless you can
figure out the IOCTLs (although running the driver may be useful
if you can run the corresponding application in Wine...)


I actually did some of this work two years ago, but I abandoned
that project (mostly due to me getting other things to do at work,
so I nuked my windows driver development environment...)  Maybe
it would be time to dig up my old stuff, and give it a try again?

What USB devices are interesting to support using this method?

   /Krister