If you were happy with the state of the Linux desktop four years ago, which I mostly was, then good news, it's still all there if you want it. You might need to abandon or downgrade Gnome, but it's all there. (Or in my case, go through the current KDE configuration and start unchecking a lot of boxes.) But absolutely yes, the latest attempts to chase Windows down have been a disaster. For how open source fundamentally works, there has been a shocking number of Architecture Astronauts running around lately. (And I do use that term in the fully original sense [1].) The pattern of:
1. Glorious design
2. Shitty first pass, released anyhow and made a standard
3. Laboriously fixing it over the period of years until it converges on "Slightly-but-only-slightly less sucky than what it replaced"
4. Goto 1
has become the standard operating procedure of the Linux desktop world.
I am not convinced that the Gnome and KDE projects are good organizational strategies. Replacing a lumbering behemoth accountable to its many customers with lumbering behemoths accountable to essentially nobody is not an improvement, no matter how "open" the latter may be.
(Oh, and I was happy only because I'm a professional programmer and I still prefer the command line over any "semantic desktop". I freely acknowledge this is not the common case.)
Linux was excellent on the desktop about 4 years ago. After that we had fedora turning to a big pile of broken packages, ubuntu adopting some seriously broken technologies (pulseaudio, compiz-beryl, tracker, unity, upstart, readahead, ... ).
My favourite anecdote about linux's degradation is this:
Earlier, applications like nautilus , cdburner etc would accurately predict and show the amount of time for operations like copy to complete. Moreover, copying stuff via the GUI was not significantly slower than command line.
With the latest versions of KDE (and even gnome), we've gone back to windows98 levels of inconsistencies (seeing the kde time-estimate bar during a file copy is just painful).
If you want a distro which doesnt take 1G of ram to run a filemanager, the latest one you can try is Ubuntu 7.04 - runs much faster on my 6 year old laptop (with 200MB ram) than my modern desktop (with 1G ram).
I'd be interested to know why you consider upstart "seriously broken". I've just used it in anger for the first time, and it seems like a big improvement on the old, arcane system of alphabetically named init scripts.
The technology is great! I love the launchd/upstart way of doing things. What is completely unacceptable for a core piece of system software is the pathetic level of documentation. I should NOT have to source dive to write an init script! Whoever let upstart out of the door with the piss poor documentation should be lashed repeatedly with a wet noodle.
I migrated from Kubuntu to Xubuntu when Lucid came out, it runs greatly on older notebooks for that matter, even Xubuntu Natty today.
While the issues mentioned in the article are valid, it's still so much easier for me to maintain all family (and several friends) pc/laptops around, after I migrated them from Windows to Kubuntu/Xubuntu. The desktops just work and don't break, which is not something you can say about the Windows tamagotchi thing - you need to care about this one, or keep it restricted.
GNOME and KDE are monolithic and encourage everything under their umbrellas to depend on everything else. The Linux desktop has fallen into the same pitfall that so many enterprises do: reaction against integration issues in a chaotic environment has led to monolithic approaches that provide uniformity and integration but prevent the mix-and-match flexibility that allows individual projects to fail and succeed on their merits. The monolithic approach, the architecture astronaut approach, means high commitment to each project and leads to projects like PulseAudio becoming "too big to fail" even if they fail at their technical objectives. The next step in evolution is supposed to be a loosely coupled services architecture that provides the best of those two worlds. I won't venture to guess where it goes after that.
however, I've noticed that striving for loosely coupled architecture has often fostered its own brand of, if you like, architectural astronomy, as the architects try to design the ultimate meta-framework that will make it easy to integrate disparate components without the components knowing anything about how they can be used.
1. Glorious design
2. Shitty first pass, released anyhow and made a standard
3. Laboriously fixing it over the period of years until it converges on "Slightly-but-only-slightly less sucky than what it replaced"
4. Goto 1
has become the standard operating procedure of the Linux desktop world.
I am not convinced that the Gnome and KDE projects are good organizational strategies. Replacing a lumbering behemoth accountable to its many customers with lumbering behemoths accountable to essentially nobody is not an improvement, no matter how "open" the latter may be.
(Oh, and I was happy only because I'm a professional programmer and I still prefer the command line over any "semantic desktop". I freely acknowledge this is not the common case.)
[1]: http://www.joelonsoftware.com/articles/fog0000000018.html