What's the point of tiers if we bend the definition of them?

[Albert Astals Cid]

El Dimarts, 17 de març de 2015, a les 09:16:50, Kevin Ottens va escriure:
> On Tuesday 17 March 2015 08:21:22 Martin Gräßlin wrote:
> > On Monday 16 March 2015 23:16:51 Albert Astals Cid wrote:
> > > We have KPackage depending on kdoctools but since it's optional we
> > > pretend
> > > it's not a real dependency and call KPackage tier2
> > > 
> > > What's in it for us other than lying to ourselves? What's so bad about
> > > KPackage being tier 3?
> > > 
> > > It'd still have only 3 dependencies and be totally usable, no?
> > 
> > Maybe it needs to document both:
> > * without optional packages it's tier 2
> > * with optional packages it's tier 3
> > 
> > I can imagine that this will be interesting for more frameworks where it
> > could be that depending on which platform we run on it's a different tier
> > (kglobalaccel is currently tier 1 for non-X11, but tier3 on X11).
> 
> I'd be concerned about the complexity of that though. It needs to be
> communicable to the outside.
> 
> Now perhaps it's a question of "what's needed to get it to build and be
> usable on all platforms". I'm not sure that kdoctools is part of those
> requirements. It's "just" to get manpages generated.

The problem is that "usable" is an ambiguous by definition, while following 
the dependency chain is quite clear.

Albert

> 
> IOW, not having kdoctools doesn't prevent you to build something on top of
> kpackage and run it while not having KWindowsSystem breaks kglobalaccel on
> some platforms.
> 
> Regards.