Slashdot Mirror
KDE 3.0 Alpha1 Available for Developers
Dre writes: "Just a few weeks after the release of the rock-solid KDE 2.2.1, the KDE Project today announced the release of KDE 3.0 Alpha1. Targeted at developers who want to get a head start on porting or writing applications to KDE 3, the release is pretty much a straight port of the KDE 2.2 branch to Qt 3. However, for developers this brings an impressive array of new features to KDE, including new database classes, new data-aware widgets, improved RAD development with a much-enhanced Qt Designer, a new powerful regular expression class (with full Unicode support), improved internationalization support (including the ability to mix different character sets in the same text), bi-directional language support (for languages such as Arabic and Hebrew), multi-monitor (Xinerama and multi-screen) support, better integration of pure Qt applications into KDE, and hardware-accelerated alpha blending. With the Qt port out of the way, the KDE developers can now focus on the planned
KDE improvements. Read the full announcement here, or go straight to the source
(alternative
link)."
That said, I expect that there will be far more new features in KDE 3.0 than what's described on that page. Most likely the developers just haven't bothered to tell anyone about all the new features they're going to add yet.
And with KDE's blazing release schedule, 3.1 will be upon us before we know it, with all sorts of goodies :-)
main(c,r){for(r=32;r;) printf(++c>31?c=!r--,"\n":c<r?" ":~c&r?" `":" #");}
icon server, Waldo Bastian bastian@kde.org
KIO/KHTML: pipelined HTTP requests, infrastructure, Waldo Bastian bastian@kde.org
I think the icon server in particular will help with startup times of KDE apps. The pipelined HTTP requests will make loading of webpages faster.
In addition, a lot of the speed problems actually lie with GCC and the GNU linker, which KDE can't help with. The GCC and ld developers have been made aware of the problem, and a lot of work is going on on their end to speed up the dynamic linking of C++ programs. Once these optimizations start making it into stable releases of the linker, KDE will be much more responsive.
main(c,r){for(r=32;r;) printf(++c>31?c=!r--,"\n":c<r?" ":~c&r?" `":" #");}
About the icon server: Currently each KDE program that wants an icon (every one) goes and checks each directory where that icon might be found (of which there are a lot, KDE has a very customizable icon system). The icon server would catalog all the icons available on startup and serve them to the programs that need them whenever they ask, preventing a lot of disk reads. I think that's the basic idea.
main(c,r){for(r=32;r;) printf(++c>31?c=!r--,"\n":c<r?" ":~c&r?" `":" #");}
Better (full) PNG transparency support in the browser
Alpha-blending for all icons everywhere to reduce jagged edges (especially for small icons)
Neat eyecandy effects
What I'm interested in is what happens when Render isn't available? Do all those effects go away cleanly, or do they stay there using slow software emulation?
main(c,r){for(r=32;r;) printf(++c>31?c=!r--,"\n":c<r?" ":~c&r?" `":" #");}
> KDE programs are ./configure --enable-final
> very slow to compile
use
this reduces compile times by more than half, in my experience
> and load.
Use objprelink.
> but the point is that KDE is a hell of a lot slower than GNOME.
From what? Load times? Look at other big applications written in C++ and compiled in g++, like Mozilla and OpenOffice. They tend to load slow too. If you actually look at speed of applications, KDE wins hands down. Konqueror versus Nautilus. Konqueror wins. KOffice versus StarOffice/OO, KOffice wins. Other components tend to be around the same speed.
> So, from my perspective, which is not that of a compiler designer, GNOME is a lot freaking faster than KDE.
Yeah, "ordinary users" don't even compile KDE or GNOME.
The pre-linking relies on the fact that once libraries are loaded, they never move in memory. That could be a false assumption, but the gcc team is going to great ends to make sure it isn't. The issue as demonstrated is that 'helloworld' will be much larger, and much slower to load when it links against the QT libraries (or any large set of libraries). Thus, similar performance is lost when starting KDE applications linked against the QT libraries simply because they are all loading the QT library linkages.
You sir, are one of three things: ignorant, a troll, or an idiot. KDE is entirely built on components, and I can swap them in and out and upgrade them and advance functionality without upgrading the whole.
Which brings up an important point - 3.0 should look just like the existing 2.2.1, with no new (user) features. The major number bumped up *because* of the component nature of KDE - the new version is the exact same (on the front end), but is binarily incompatable with 2.x. The back end gives advances that will allow the 3.x line to advance quickly on the front end.
--
Evan
"$30 for the One True Ring. $10 each additional ring!" -- JRR "Bob" Tolkien
Most of my expertise falls into software development in Java, but recently I have been pushing myself to do more C++ development on Linux. I don't want to rely on proprietary languages. Anyway, the modern C++ is far far different from the C++ of many years ago. Things have changed, and C++ is growing up. Sure its a very complex language, if you try to learn the legacy aspects of it, but if you stick to the core OO constructs in C++, then you have a nice efficient programming language.
I am coming to realize that Java has very little over C++. Garbage Collection is more of a buzz word than an actual worthwhile feature, and it should be noted that high-level memory leaks are still possible in Java. Sure they are memory leaks of a different kind, but unreleased yet unused memory is a big problem in many large Java software systems.
In addition, for even an intermediate software developer, how difficult is it to code your own destructors? I mean, really, at worst you have a few loops in a destructor. Anyway, most JVM garbage collectors are unpredictable and hog performance at the worst of times.
The most important thing that Java has over C++ is a comprehensive set of user-friendly yet powerful APIs. But in return, C++ as templates and STL, allowing for elegant generic software systems.
When it comes down to it, C/C++ are here to stay, until some real yet practical innovation in Functional Programming languages, mobile/concurrent languages, or Declarative Programming languages is made. I am all for newer better higher-level programming languages such as Haskell, Pict, Lolli, and Mozart-Oz, but Python, Ruby, C#, and other newer procedural/OO languages are not the revolution, they are not the future, and at best, they are nothing more than slight iterative improvements on an overused overdone, and over talked about paradigm.
Give me C, give me C++, and if you can, why don't you give me something new for once? I am tired of the same old Ford Tempo with a new paint job and a new name.
Have you ever right-clicked on the file name input widget? It says right there: "Completion: (none, manual, menu, automatic, short automatic)"
This goes for EVERY input widget that accepts URLs. Not only for file dialogs. KDE rocks. :-)
Home Page