Re: Application modal transients windows

[Paul Warren <pdw ferret lmh ox ac uk>]

On Fri, 5 Nov 1999, Michael ROGERS wrote:

> >The 1.9b draft suggests that we should use the GTK method of application
> >modal dialog boxes as standard, where WM_TRANSIENT_FOR is set to the root
> >window, and the dialog is kept above all other windows in this group.  
> >
> >John Harper has pointed out that this is not inline with the ICCCM, which

[...]

> IIRC GTK has a function to set the WM_TRANSIENT_FOR hint - something like
> gtk_set_transient_for :) . If we put code in our window managers to print a
> warning on stderr when a window has WM_TRANSIENT_FOR set to the root window,
> application writers will be pestered to change their applications to use the
> ICCCM-compliant form.  =)

:-)

Thing is, set WM_TRANSIENT_FOR to the root window is quite a neat and not
entirely unintuitive way of implementing application modal windows.  Our
choice is between:

(a)  Coming up with a new scheme for indicating an application modal
window.  GTK (and possibly other toolkits/apps) will need rewriting.
Window managers will need to recognise this new hint, whilst still
supporting the GTK way (possibly with a warning).  The advantage being
that we comply with the ICCCM to the letter.

(b)  Making the GTK method the standard.  We break nothing, and most WMs
support this already.  We save ourselves a hint.  We bend the ICCCM very
slightly.

Whilst in general I think standards should be stuck to rigidly, this is
such a minor infringement of the ICCCM that IMHO we can let it pass.  Do
we take what is written in the ICCCM as gospel, or do we instead work from
what the current interpretation of it is?

I don't really care either way, but I think that (b) offers the least
hassle all round.

cheers,

Paul