Re: xinerama and struts

[Havoc Pennington <hp redhat com>]

Ben Jansens <xor orodu net> writes:

   Another option.
   +-------+-------+
   |       |   2   |
   |   1   |.......|
   |.......|-------+
   +-------+

This one is clearly the right behavior IMO. If you make the struts
relative to a given monitor, then it works out.

This may show that we may need a way to make a strut relative to a
given monitor, even if we also have partial-width struts.
 
> In the case of openbox/blackbox, the Slit (dock app tray) uses the
> strut, but is only as long as the apps it contains. So this strut
> weirdness is almost a permanent 'feature'. What we've done is to use
> the strut for maximizing and window placement, and ignored it, using
> actual windows for snapping. I think that the preferable behavior is
> different depending on what the strut is being used for.

Right so with GNOME we have the classic "explosion of interacting
features" problem where you make whether the panel spans the whole
screen or not a setting, and suddenly it's more complicated to make
everything else work as we can't just assume the panel does or doesn't
span the screen.

One interesting point is that the WM could do heursitics based on the 
nature of the window:
 - if the window is fully on one monitor, interpret strut relative
   to that monitor
 - if the window is not spanning the whole screen, ignore it for 
   position constraints
etc.

I think these heuristics work really well with current GNOME/KDE
anyway. But I don't really like relying on them.

Havoc