Re: [Libusb-devel] [libusb] #33: Mac OS X libusb 1.0 native HIDbackend

[Segher B. <se...@ke...>]

>> That's simply not true.  A user-mode program cannot just use WinUSB.
>> Instead, I have to create a driver package with an INF file, then
>> install that driver package.  Only then can I use WinUSB.
>
> The point is that the "driver package" doesn't (have to) contain any
> kernel
> drivers, it (can) use the ones that come with the system.
>
>> Segher is exactly right.  Windows simply does not have the fallback
>> no-install-required generic USB support that Linux has.
>
>> Segher Boessenkool wrote:
>>> The point is that on OSX and Linux, you need *no* special kernel
>>> driver;
>>> the base USB support allows you to do a driver in user space.
>>> Mswindows
>>> apparently allows no such thing.
>
> Segher is claiming that MSWin devices always have to come with
> a kernel driver that needs installing. This is not true.

I am not claiming anything, I am trying to understand the situation.
There's a whole lot of conflicting information now.

On Linux and OSX, you need some configuration to allow the user access
to the device; on Linux you typically need to add a udev rule for your
vid:pid, on OSX you typically don't need to do anything.

Other than such permission checks, the kernel gives userspace direct
access to any USB devices it doesn't have claimed.

Are you saying it works like this on mswindows?  Why then do lots of
device developers on mswindows find it necessary to use HID class
devices to avoid this all?  It isn't considered a problem on Linux
or OSX.

Something must be missing here...


Segher