Re: Proposal for ConfigureRequest handling

[Marko Macek <Marko Macek gmx net>]

> On Thu, Aug 08, 2002 at 04:00:52PM +0200, Marko Macek wrote:
> > No. What I'm saying that if a window manager supports NETWM hints, it
> should
> > MUST comply with this ICCCM rule about window positioning (at least for
> > NETWM compliant applications).
> 
> Okay, how does the WM find out if an app is NETWM compliant?  The
> application could set a property indicating this on its top level
> windows.  That may be the better solution since it takes the
> workload off the app programmers and lets the window manager
> handle it.  On one hand it's much easier to do that in the WM
> (because it knows the size of the decorations) and on the other
> hand it's less error prone in the applications.

I suggested (below) that NET_WM_PING is used for this. 
It is trivial to implement in the apps and must be registered in
WM_PROTOCOLS if supported. It also has no extra special semantics for window management.

Right now it seems there is no mandatory property that an app must set
unlike in the old gnome-wm hints spec.

> > We might use your proposed root window property instead of implicit
> > dependance on _NET_SUPPORTED...
> > 
> > We could also say that if application supports NETWM hints (TODO: how
> this
> > is determined? check for NET_WM_PING on WM_PROTOCOLS? AFAIK there is no
> hint
> > that an app must always set in current specification),  it must use the
> ICCCM
> > compliant window positioning when used with a NETWM compliant window
> manager.

> > > Hm, perhaps this could be done via the initial_state.  Set this to
> > > "withdrawn" and the WM sets up the window as usual, but does not
> > > map the frame itself.  The client would notice that it was
> > > reparented and could look up the frame size and its position.  It
> > > would then ask to map the window manually and the WM maps the
> > > frame.
> > 
> > Maybe. Is this a standard thing?
> 
> No.  There is no "Withdrawn" initial state yet.  It would be
> another extension to the ICCCM.  Perhaps "Hidden" is a better
> name.  Instead of listening for reparenting the app might want to
> wait for a synthetical Expose event to indicate the WM is done
> decorating the window.

Please, not yet another overload of "Hidden".

I think using WithdrawnState (on initial mapping only, the handling for
later handling if WithdrawnState would not change) might be acceptable, at least
for netwm apps, but probably for all...

MArk

-- 
GMX - Die Kommunikationsplattform im Internet.
http://www.gmx.net