Re: _NET_WM_[GET_|TAKE_|REQUEST_]FOCUS & urgency

[Elijah Newren <newren gmail com>]

On 5/26/05, Luke Schierer <lschiere users sf net> wrote:
> On Thu, May 26, 2005 at 10:44:25AM +0200, Lubos Lunak wrote:
> >
> >  Actually you cannot completely rely only on WM_CLASS even for non-broken
> > apps. You can have several instances of Konqueror running which you should
> > consider separate, you can have one instance of Konqueror running but showing
> > several windows (which you again should consider separate), and you can have
> > dialogs shown for Konqueror from other processes (kwallet can show a dialog
> > asking for password that more or less belongs to the Konqueror window, the
> > same way kcookiejar can show a dialog about cookies). KWin implements rather
> > non-trivial (and hackish in places) logic that tries to guess which windows
> > actually are related.
> 
> It seems you are complaining not that WM_CLASS does not work, but
> that it works too well!  afterall, in your example, they are all
> Konqueror windows, you just want them treated differently.

Hehe, my understanding of this conversation is basically "Applications
implement the ICCCM spec extremely accurately--too bad the information
it provides is useless/misleading in many cases."  Is that correct? 
E.g. if a user launches another Mozilla window from the taskbar, they
think of it as a separate instance of the application.  As an
implementational detail, the Mozilla developers have the second
instance tell the first to open a new window, in order to save on
resources.  According to the spec, both windows _are_ from the same
instance of the app, so the WM_CLASS doesn't differentiate the two. 
Thus, the spec exposes implementational details that are
counter-intuitive to user expectation.