Re: Linux Window Managers and Accessibility

[cwitty newtonlabs com (Carl R. Witty)]

Bill Haneman <Bill Haneman Sun COM> writes:

> Carl said:
> 
> >Bill Haneman <Bill Haneman Sun COM> writes:
> >
> >> Sasha said:
> >> 
> >> >Host name might be a better solution, and by the way, fits into 
> >> existing
> >> >standards - ICCCM WM_CLIENT_MACHINE property could be used.
> >> 
> >> 
> >> Right, thus the suggestion of char* for the parameter type in the method 
> >> mapping application ids (including hostname) to Desktops.
> >
> >Hostname is also not guaranteed to be unique.
> 
> It should be unique within the context of the accessibility client-server 
> relationship, or else completely transparent (e.g. app must migrate seamlessly 
> if physical host is transient).

Well, let me give a hypothetical example.  Suppose I'm a system
administrator, and I'm responsible for running two semi-private
networks (both networks use 192.168.*.* IP addresses).  I have a
machine of my own in each network.  Both machines happen to be
192.168.3.7, and both machines happen to be named "cwitty".  This
doesn't cause problems, because neither the name nor the IP address is
exposed to the global Internet.  I have X clients on both machines
running on my X server (over NAT gateways).

Obviously, this is an extremely contrived example.

> >> The goal here is to map an accessible application's unique 
> >> identification info (provided by the app when it registers with the 
> >> AccessibilityBroker) onto the appropriate desktop/desktops.
> >
> >Why not have the AccessibilityBroker give a unique identification
> >token to the application when it registers?  The application can pass
> >that on to the window manager.
> 
> I agree with the first part, but what (existing) API for the window manager 
> would one use to pass this info to the window manager?  The key issue is that 
> the window manager must be able to map application identifiers to virtual 
> desktops.  As you point out if the application is not using the window manager 
> (in other words, maps no windows) then the window manager can report that, and 
> the broker can choose to map it to desktops according to its own heuristics.

Why do you care if it's an existing API?  You're going to have to
change the window managers and the applications (or at least the
toolkits) anyway.

I think you're going to have to come up with a new protocol to give
this information to the window manager.  (Something as simple as
setting an ACCESSIBILITY_ID property on top-level windows would be
fine.)  There just is no definitively identifying information which
applications currently pass to the window manager (I'm pretty sure).
And if you're coming up with something new, it seems like you might as
well use something truly unique (an ID from the AccessibilityBroker)
as your identifier, instead of using things like IP addresses,
hostnames, and PIDs.

Carl Witty