Simplifying the specs

[Matthias Ettrich <ettrich troll no>]

Struts:

     Is the array overkill? Shouldn't one strut that is valid for all desktops
     the defining window is visible on be sufficient?

     I know I requested the hint that complicated, but I'm a bit worried. The
     array makes that simply hint much more complicated and is bascially only
     useful for KDE-1.x's panel..... Probably we will not even make use of this
     in KDE-2.0.

     I suggest to simplify it to _NET_WM_STRUT, CARDINAL[4]/32


NET_WM_STATE_MODAL

     While I honestly believed the hint was a good idea when I asked for it, I'm
     not sure any longer. I did some testing in Qt and draw the conclusion that
     modality really should be handled by the application. Since we already have
     hints to notify the WM that a window shouldn't show up in the taskbar,
     there's not much functionatliy WM_STATE_MODAL can offer.

     Maybe we should instead encourage toolkit vendors to deal with modal
     windows according to ICCCM? This basically means setting wm_hints.input
     to false for main windows and requesting the WM_TAKE_FOCUS protocol.
     When receiving a take-focus method, the toolkit should simply set the focus
     to the modal dialog ( using the right timestamp). That's the "Globally
     Active" mode described in section 4.1.7. of the ICCCM and it's the only
     mode that works sufficient and user-friendly in combination with modal 
     dialogs. 

     WM_STATE_MODAL on the other hand might be a bit easier. And it would
     allow the window manager to define a beep-sound or a visual bell or
     anything if the user clicks fail. If each application does this, it will
     be pretty inconsistent (some playing sounds, some just beeping, some
     flashing....). Also I have no idea how good the Globally Active modal works
     with existing window managers. KWin does it fine, kwm probably suboptimal.

     Dunno. I just wanted to share my thoughts. STATE_MODAL may be userful,
     but it's a little bit duplicating. If we keep it, however, we should also
     emphasize on the fact that the application may request a state stange
     while the window is mapped (some dialogs like to become modal/unmodal
     wile being visibile).
   

Matthias