why kdelibs?

Alexander Neundorf neundorf at kde.org
Wed Nov 3 18:32:01 GMT 2010

On Tuesday 02 November 2010, Aaron J. Seigo wrote:
> On Monday, November 1, 2010, Diederik van der Boor wrote:
> > I can't stress enough it needs a good plan to limit impact on apps.
> this would be absolutely critical indeed.
> something that is completely opaque to me in this discussion so far is
> whether the idea is to:
> a) grind kdelibs into a fine dust of classes, rename some from KFoo to QFoo
> (and all the Qtification that implies) and reorganize the remainder (which
> should be a minimal set) into a new set of K* libs
> or
> b) define a new set of granularities for libraries in kdelibs (non-ui,
> desktop, framework, etc.) and split up our existing classes along those new
> lines, keeping names largely in place
> or
> c) do a bit of (a) and (b), and if so what the guidelines for deciding when
> to do (a) and when to do (b) are
> probably because it isn't a concrete plan yet, but rather a general
> proposal. these are the kinds of details that are needed before we can move
> on.

Just my personal view: if we start something like that relatively soon, i.e. 
this or next year, keeping source compatiblity should be the top or at least 
the second top priority. Even if it means doing things in a not-ideal way
(like adding wrapper classes or leaving stuff just as it is although it could 
be nicer).
Which would imply that renamings/removing of stuff would be a no-go.
I guess this matches mostly b).

My own view would be that it would be nice to have a set of libs which are 
independent from the "platform" and a set which uses/relies on 
the "platform".
What does "platform" mean ?
when looking at this with my 3rd party developer hat on, I would say it means 
that a lib depends on the platform if it needs some daemon (kded etc.) 
running (would dbus itself count here ?) or if it relies on that some KDE 
tool was executed before (e.g. kdesycoca, or systemsettings).
Everything else should be fine.
If we (at work) would have to install shared-mime-info, ok. It's a small 
package, no big deal.  The same for shared-desktop-ontologies.
Running mysql or another database ? No-go.


More information about the kde-core-devel mailing list