Re: screen/display switching

[Havoc Pennington <hp redhat com>]

Hi,

Suggest moving thread to wm-spec-list, so setting Reply-To and quoting
your whole message.

On Thu, Mar 13, 2003 at 09:18:41PM -0500, James Willcox wrote:
> Hi,
> 
> In gtk+ 2.2, multi-head support was added, giving developers the ability
> to move windows to different screens or displays on-the-fly.  I think it
> would be great if we could push this ability up to users.  For me, the
> most common use for this would be to move windows between screens since
> I use a multi-head setup w/o xinerama.  Currently, there is no way to do
> this (unless the app has something specifically built-in to allow it).
> 
> If a window supports this spec, it puts "_NET_ALLOW_SCREEN_CHANGE" in
> its WM_PROTOCOLS.  The actual screen switching uses the same "X Message"
> protocol that the startup-notification stuff uses (outlined in
> startup-notification/doc/startup-notification.txt), with
> "_NET_SCREEN_CHANGE_BEGIN" and "_NET_SCREEN_CHANGE" as the message types
> used.  The body of the message would contain either a full-fledged
> display name, like "foo.com:0.0", or just a screen identifier like ":1"
> (referring to screen 1 on the local display).

I would add:

 - the X message stuff would be sent to the client window to be moved,
   not the root window as in startup-notification (probably obvious, 
   but good to clarify)

 - Perhaps more explicitly separate screen change from display change;
   I don't think apps can reliably determine from a display name 
   string like ":0.1" that it's another screen on the same 
   display they're currently using. Maybe I'm wrong.

 - Maybe there should be a protocol that asks the entire application
   (i.e. all windows with the same group leader) to change
   screens/displays. Yes the WM could just move every window that
   shares a group leader individually, but it might be more efficient
   or just nicer if the app did it. This protocol could simply be
   "send the screen change message to the group leader" perhaps.

> If this sounds reasonable to people, I can write up a more comprehensive
> description of the spec.  AFAIK, Qt doesn't have the ability to do
> screen/display switching, so I guess this is mostly aimed at a time when
> other toolkits get this.

Havoc