KDE/kdelibs

[Oswald Buddenhagen]

On Fri, Nov 27, 2009 at 03:45:24PM +1100, Ian Wadham wrote:
> On Friday 27 November 2009 2:53:02 am Friedrich W. H. Kossebau wrote:
> > Just that new features need testing and feedback before they get stable.
> >  But if all application developers just wait for the stable version it
> >  simply will never get stable.
> > 
> I hope not too many KDE library developers think this way.
>
they do.

> It has to be the responsibility of library developers to test their
> work to the greatest extent possible before releasing it, regardless
> of whether any application finds a use for it.
>
quite to the contrary, new major functionality (classes) is not
permitted into kdelibs unless there are at least two real-world users
(i'm pretty sure it was three in earlier times).

>  That is what makes library writing so difficult and challenging.
> 
> > So new features and code using these features need to be developed
> > in sync for best feedback. I think so far this has worked very well
> > for all the modules which are part of the SC.
> > 
> It is nice when synergy between library and application developers
> occurs.
> 
no, it's vital.

> It is not so nice when an application does not work properly and the
> writer is unexpectedly left having to track down a problem that may
> lie somewhere in his/her own code or in kdelibs or in Qt, especially
> if the writer's appeals for help fall on deaf ears or the response is
> "try and fix it yourself".
> 
welcome to the real world.

if you force a self-contained library development model onto kdelibs,
nobody will add any functionality any more. even qt relies on feedback
from the internal tool application developers and the broader community
- and that despite the requirement of having three example applications
for each major functionality.