Re: RFC: frame size hints

[Dominik Vogt <dominik vogt gmx de>]

On Thu, Dec 11, 2003 at 03:12:02PM -0500, Thomas Fitzsimmons wrote:
> On Wed, 2003-12-10 at 07:03, Lubos Lunak wrote:
> > On Wednesday 10 of December 2003 01:10, Rob Adams wrote:
> > > Definitely should be "calculate".  Window managers implementing
> > > approximations should document this is their respective COMPLIANCE
> > > documents.
> > 
> >  I object. I hope the previous messages in the thread have shown that it can 
> > be a major pain to always get it right, if it's actually possible at all. 
> > KWin simply cannot calculate it, it can only estimate.
> 
> OK.  I'm fine with describing the hint as an estimate.  Having a
> standard way to get a decent estimate is way better than having nothing
> at all.  And describing the hint as an estimate will probably mean that
> more window manager authors will create best-effort implementations
> rather than just ignoring the hint and declaring non-compliance.  Rob,
> are you OK with this?

I think it is not right to ask for the decoration geometry only
once.  *If* the client is interested in this property, it must
subscribe to a client message informing it of changes.  With the
current draft, it will not work at all if the application is
started before the window manager, or if the config and/or WM
changes.

> > Even Havoc said in one 
> > message that the Metacity patch (which is the reference implementation for 
> > this) is not guaranteed to get it right.
> 
> Yes, but FWIW, I believe it is possible to fix the implementation so
> that the extents are always correct; the Metacity patch is just an
> initial implementation of an admittedly difficult-to-get-really-right
> feature.
> 
> > And I said already, I'd actually 
> > prefer if the wording stressed more the fact that it can be just a guess.
> 
> OK, should I add an implementation note or something?

Ciao

Dominik ^_^  ^_^