Re: Autoplacement hint

[Havoc Pennington <hp redhat com>]

On Mon, 2003-12-22 at 01:40, Mike McCormack wrote:
> I'd really like to have a toolkit that doesn't have to work around 
> broken window managers, libraries and applications, but then the users 
> of my toolkit will consider it to be broken.  My guess is that many 
> window manager authors will argue that autoplacement when PPosition is 
> set is not in violation of the ICCCM.  So what should I do?

The way X works is that the WM is given broad leeway to implement
different policies. In fact many of them are configurable (IIRC Sawfish
has an "honor PPosition" option, or used to).

Fighting this in the toolkit is broken and just makes things
unpredictable and flaky. What you should do is write code that works
with standard WM policies, and if people want to run a WM that does
things differently, they need to complain to their WM author or change
WM. After all, that's the _whole point_ of having replaceable WMs - the
WMs can be different.

Swing and some GTK apps and other stuff will also break when not
honoring PPosition, metacity originally tried to ignore PPosition and
had to give it up because it's just not practical.

The least-buggy steady state of the whole system is definitely for
everyone to follow the specs; if everyone starts trying to work around
everyone else you get a big mess. Adding a "but I really mean it this
time!" hint doesn't help IMO. ;-)

Havoc