[RFC] Input design

[Magnus Vigerlöf]

On lördag 21 april 2007, James Cloos wrote:
> >>>>> "Daniel" == Daniel Stone <daniel at fooishbar.org> writes: "Nicolas"
> >>>>> == Nicolas \"Ikke\" Trangez <eikke at eikke.com> writes:
>
> JimC> An option for using HAL is ok.  But the server must not rely on it.
>
> Daniel> Sure.  As I said before, it will build and run just fine without
> Daniel> it, but you're going to have to feed in the input devices
> Daniel> manually, as you do now.
>
> OK. But, ...
>
> Nicolas> I still think allowing something similar to the current DBus
> Nicolas> interface should be possible too, as I think on some platforms/
> Nicolas> in some environments being able to add devices to the server
> Nicolas> from client space should be possible.
>
> If this means that client apps (or other dbus clients?) would be able to
> instruct the server to add or remove devices, then that sounds like the
> optimal solution for non-hal systems.

Not only for non-hal systems. Certain devices (Wacom tablets at least) will 
need to have several input devices defined for one kernel device. The current 
driver can separate different tools (stylus, eraser, mouse, etc) and 
different parts of the active area into different input devices. I don't see 
how this could be done automatically by the server. Sure, setting up a simple 
basic mouse shouldn't be a problem, but the more advanced features?

We also have applications that are dependent on the name of each device and 
assign properties to specific devices by their name (gimp), so these names 
should be persistent and assigned to the specific hardware (I would expect 
this from a user pov).

Cheers
  Magnus