KDE3 - the last Real KDE Desktop, KDE4 never got there, Plasma is still a mix
The comical part of the article was the comment that KDE3 was considered bloated and heavy (cough... maybe for that 486 with 512K of RAM). KDE3 is a blistering fast full featured desktop environment where (with SSD), you can boot Linux from power-off to full desktop is 11 sec. The author is right about the quote that current KDE is built on cars and coffee machines, because just to start current KDE5 (same SSD), you will have time to stroll down to the coffee machine and get a cup before the desktop is ready....
I have used KDE since (KDE2?) though its pinnacle of 3.510, then starting with KDE4-4.0.4a released May 2008 as the default desktop on openSUSE 11.0, though the current 5.2 built from git sources. From the time 4.0.4a was release to present I have authored over 400 bug reports for things that simply don't work or don't work right or for missing functionality that didn't survive the port from Qt3 on. Everything from systray and clock font control to the wrong physics model used for sphere and cylinder rotation. And then to find the same problem bugged in 2008 against KDE4 still not fixed in KDE5. Sigh...
The problem being that neither KDE4 or KDE5 were developed for a specific Qt version. Most of the apps and interface were developed for Qt3 and then have been ported (more or less successfully) to later versions of the toolkit. There have been new applications developed to replace some of the older ones (Ocular for Kpdf, Spectacle for Ksnapshot, etc..) and in each case the tools remained broken for years as they tried to duplicate functionality available in the older versions. In the case of Spectacle ending up with a dumbed down version that was incapable of taking a screenshot of itself due to the single process limitation)
Now to be fair, Plasma (KDE5 or whatever they choose to call it this week), does work much better now than it did when it was launched, as did KDE4 before it was scrapped for KDE5. But at what cost? The new additions to kate/kwrite are neat, but don't provide any additional functionality. Konqueror, the premier file manager in KDE3 (--profile filemanagement), remains a crippled awkward shadow of its former self, and Dolphin (a/k/a the dumbed down version of konqueror meant to be easier for novice users), simply lacks any real file management features. So Plasma is now usable, but clumsy compared to KDE3. This after two+ years of dealing with some applications that were still KDE4, some Plasma and a myriad of layered dialogs that are difficult if not impossible to navigate in system settings (the new Kcontrol). To this day you still can't put a different wallpaper on each of your multiple desktops (something trivially doable KDE3) without awkward hacks to the "activities" feature or otherwise.
Gone are many efficiencies and finely tuned human-factor considerations present in KDE3 designed to minimize the number of keystrokes or mouse-clicks needed to accomplish any one task. Gone are the considerations that minimized the number of times you would need to move your hands between the keyboard and mouse and then reposition on return to the keyboard. Not something that was intentionally lost, but the unintended consequence of porting a finely-tuned desktop and applications from Qt version to Qt version and basically having to duct-tape and bailing-wire the underlying class hierarchy code to "get things working" with the new toolkit.
Now don't take this as throwing rocks at the KDE team. The KDE3 team was/is an incredibly talented group that developed a desktop from the ground up arriving at what was arguably the most complete and efficient Linux desktop ever made. The sheer amount of human-factors efficiencies integrated into KDE3 is still impressive today. Many of those same talented developers are still there.
The problem comes when decisions are made to run off chasing new gotta-have "widgets" provided by Qt4 (the KDE summit in 2007), compounded by putting out a release before it is ready (alpha-quality at best... the 'a' at the end of 4.0.4), and then the ensuing decade of trying to make it all work. What about the end user while all these gee-whiz new things are being tried? The user-base suffers -- partially working apps, crashes or lost functionality and lost time and productivity trying to chase down why X doesn't work anymore, or you can't do Y with it. And..., then to throw it all away in 2018 to run of chasing Plasma -- restarting the process of patching and porting and broken apps and lost time and functionality which we are now 2+ years into (and hoping it doesn't take another 8 to get right) before... new "widgets" in Qt6 become the next latest rage.... Yes, we do fail to learn from history, and yes, we are destined to repeat it.
So I wish the KDE team good luck with KDE5, and I hope that KDE5 nears actual refinement and completion (something KDE4 never really achieved) but the experience since 2008 and the new KDE4-4.0.4a on until today epitomizes why it is almost impossible for anyone but hobbyists and developers to use a Linux desktop. (don't think Gnome3 is a panacea either -- it has stumbled mightily since Gnome2 trying to hit a moving GTK+3 target (which is now scrapped for the new .... wait for it ... GTK+4). There is a reason GIMP is still built on GTK+2...
So with the thousands of hours spent collecting data, strace'ing misbehaving apps, picking though logs to author bug reports, ... would I do it again? You bet, so long as my daily driver desktop allows me to just get work done when needed, and it does, e.g. "konqueror --verison, Qt: 3.3.8c, KDE: 3.5.10, Konqueror: 3.5.10" :) Hope springs eternal.