Release schedule clarifications

[Dan Meltzer]

On 10/24/07, Frerich Raabe <frerich at hex.athame.co.uk> wrote:
> On Wed, Oct 24, 2007 at 08:08:30PM +0200 or thereabouts, Torsten Rahn wrote:
> > I'd also like to point out that so far KDE 2.0 was certainly our most
> > problematic release. However even for that one the participation of
> > enthusiasts was a whole lot better. KDE 2 RC1 still had rough edges (more so
> > than any other release until KDE 4). KDE 4 on the other hand still has got a
> > lot of holes (calling it "rough edges" would be euphemism) in addition to the
> > rough edges.
> >
> > While many apps work very well the workspace (plasma) still has serious
> > problems in terms of performance as well as quite some gotcha's in terms of
> > basic usability.
>
> [..]
>
> > That all being said the late progress especially with regard to plasma has
> > been totally amazing. It's clearly visible that we are getting nearer to the
> > release each day and that KDE 4 has got a lot of potential to be an awesome
> > release.
>
> I'd like to add a few thoughts and experiences which evolved while porting
> my little 'KNewsTicker' pet project to Plasma:
>
> 1.) The technical possibilities of having a graphics view as the desktop are
> amazing and probably allow great freedom when it comes to designing your
> GUI. I had great fun brainstorming about KNewSticker TNG should look like.
> Unfortunately I had no good ideas, but still :-)
>
> 2.) The technical performance of QGraphicsView (and thus, Plasma) on X11 and
> MacOSX is still very unsatisfying. In particular, QPaintDeviceX11 (that's
> what I looked at) apparently performs rather poorly. For instance, the
> NewsTicker (which basically has ten QGraphicsView Items and moves them
> along) eats 15% CPU - on a 2GHz Intel Core 2 Duo using NVidia hardware
> accelerated graphics. I was wondering whether maybe I did a stupid mistake
> and thus checked what other QGraphicsView demos perform like. However,
> seeing how e.g. the 'collidingmice' demo (ships with Qt) eats about 50%
> on my box, 50% on the Mac Mini at work and <1% on the Windows XP box was
> more than sobering. And the 'collidingmice' demo basically moves seven
> sprites over a pixmap background image (so you'd think that'd run like
> *snap* this - which it does, but only on Windows. I surely hope that
> the required optimizations can be done for Qt 4.4 or Qt 4.5 latest and
> that they don't reveal a design weakness.
>
> Two sub-paragraphs dealing with the 'this is not an issue' responses I
> thought of:
>
> a) "Today's machines certainly have the power" - they do. However, leaving
> aside that I'd rather prefer to keep at least 95% of the processing power
> to my applications, the CPU usage is also annoying when it comes to laptop
> usage (more CPU usage == more power consumption == less laptop lifetime).
>
> b) "The API is what counts, and it rocks. The performance fixing should be
> left to Trolltech". I agree that the API is *sweet*, however I'd rather not
> embarrass myself by recommending people a desktop of which I cannot even say
> when (if at all) it will perform reasonably well. I'd rather not sit there
> and twiddle my thumbs hoping that with The Next Release(tm) it's going to be
> all happy happy joy joiy.
>
> All in all, despite the fascination when it comes to visual possibilities,
> I'm not convinced you can build a panel & desktop & applets on QGV on X11
> with satisfactory performance.
>
> 3.) While doing the KNewsTicker port, I noticed that (due to lack of
> corresponding uspport in Qt), the Plasma people had to invent their own
> Widget and Layout system (so Widget inherits LayoutItem, Plasma::Applet
> inherits Widget and so on). This is necessary so that you can have e.g.
> little input fields, buttons, checkboxes etc. in your Plasma applets. I
> feel this is a gross duplication of concepts and probably a maintenance
> nightmare. Luckily, it seems the Plasma people think the same and they're
> planning to ditch the whole Widget/Layout stuff (except maybe for esoteric
> Layout algorithms) in favor if a Qt 4.4 feature (namely, having QWidget
> objects in a QGraphicsView). However, this means that the 'Plasma library'
> (which people like me who intend to develop Plasma applets use) will
> change in a source and binary incompatible way. This in turn makes the
> Plasma 'platform' very uninteresting except for a few little experiments
> for KDE 4.0. FWIW, hearsay suggests that Qt 4.4 is going to be out in
> Q1 2008 (so between KDE 4.0 and KDE 4.1 we will have SC and BC breakage
> in the Plasma library).
>
> These three things written down, I'd like to express that playing with
> Plasma applets (which are QGraphicsItem objects) is a lot of fun. However,
> I can't shake the sensation that Plasma will not be as mature as I'd
> expect it to be for KDE 4.0. At least, at htis point, I'm not too motivated
> to continue developing with it. Maybe it would be interesting to introduce
> Plasma with KDE 4.1 (after all, then it has more time to mature and it can
> use Qt 4.4 - so real widgets and real layouts).

Hi,

Question for  all really, but this seemed the best place to reply.

What about simply not releasing workspace with 4.0?  What if we just
cut off problems at the pass and released only kdelibs,
kdebase/runtime, kdebase/apps (maybe?) and required that users run it
within kde3 or whatever their desktop of choice was?

This would eliminate one of the showstopping release issues, and allow
for people to work on getting kde4 ready for release instead of
getting basic functionality working.

Thoughts?

>
> (please don't flame me to death, those are just some notes I took while doing
>  the porting work).
>
> - Frerich
>
>