Re: [PATCH] Expose input device usages to userspace

From: Elias Naur
Date: Tue Mar 14 2006 - 06:44:21 EST


On Tuesday 14 March 2006 11:59, Arjan van de Ven wrote:
> On Tue, 2006-03-14 at 11:46 +0100, Elias Naur wrote:
> > On Tuesday 14 March 2006 09:22, Arjan van de Ven wrote:
> > > > > No, I don't think this is needed at all - users should be
> > > > > interested in what capabilities a particular device has, not what
> > > > > type it was assigned by soneone.
> > > >
> > > > I see your point that an application should not rely too much on
> > > > device usages. However, the main reason I want device usages is to
> > > > help applications and users identify and (visually) represent
> > > > devices. For example, games could show an appropriate icon graphic
> > > > representing each active device. The event interface already has a
> > > > few other ioctls for this kind of information:
> > >
> > > ok then you should consider to do it the other way around: make a way
> > > of asking
> > > "are you matching THIS profile".
> > > rather than
> > > "what profile are you"
> > >
> > > that way devices can present multiple faces etc; which is going to be
> > > needed as more and more weird devices come into existence.
> >
> > If by profile you mean a device usage like Mouse, Keyboard, Joystick etc.
> > is your proposal covered by the bit field ioctl exposed by my patch? For
> > example, a device can already expose itself as both a joystick and a
> > mouse (see the hid-input.c changes from the patch).
>
> no that's not what I meant; I really mean asking "can you do THIS
> profile". Example would be a device that could be either a joystick and
> a mouse, or a touchpad and a mouse, but not both a touchpad and a
> joystick. So the app should ask "can you do THESE", and the driver can
> then do anything complex it wants to come to an answer. (Of course a
> generic helper for the simple case is fine)

I think I'm beginning to understand what you mean, but it still seems way too
complicated for my taste. Devices with multiple profiles seems like a static
property more than a dynamic one, so simply splitting up these weirdo devices
into multiple logical devices at registration time seems like a better idea.
Each logical device would then have a separate set of axes, keys and, with my
patch, usages.

- elias
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/