"Far less fragmented" is anything but a good description of what I remember. There was a time when you had reasonable chances of running applications which used five different toolkits -- Qt, GTK, WxWindows, Motif, Xaw (, FOX, the various "incarnations" of Xaw and so on, like Xaw3D and the one that gave it the NeXT look, I don't remember its name). You could drag and drop within a Xaw application, but not from a Xaw application to a Qt application. A lot of distros routinely shipped with like 5 windows managers, besides KDE and Gnome.
There was a very brief period, between cca. 2005 and 2011 or so, when the fragmentation was less obvious because there was some degree of integration and cohesion (we had things like QtCurve, for instance). Then everyone started having delusions of grandeur again, in an early-90s-Unix manner, and things have been pretty much degrading ever since.
Today's "Linux desktop ecosystem" (God I hate this particular sequence of words) seems less fragmented largely because application development outside the major desktop environments has been largely abandoned, except for very relevant niches (photo/video editing, web browsers). A development that's largely unsurprising between KDE's architecture astronautics and Gnome's see no feedback, speak no feedback, hear no feedback attitude (that GTK, sadly, adopted for a pretty long time).
The unpleasant consequence is, of course, the "app problem" you speak of. A long time ago, the default KDE installation in Slackware 10 (which shipped with KDE 3.5, I think) shipped with a huge suite of applications, including graphical diff tools, VCS frontends, several multimedia players and so on. Two major rewrites later, they haven't re-accumulated this wealth of applications (and some of the ones that they do ship or advertise today are practically abandoned or remnants from the KDE 3 days). Some of the developments have been outright catastrophic, like KMail, which was turned from a very useful mail client to something that borks in a gazillion unpredictable ways as soon as you try to configure more than one account.
Similar things are happening in Gnome land, where they've chased feature-parity with Gnome 2 for years as they've been scrambling to fix everything that wasn't wrong with it and the horde of bugs that ensued from these fixes. Which, in fact, is why they have three forks in the first place. There was a lot of negativity about the Gnome 1 -> Gnome 2 transition, too, but that never resulted in forking. Nowadays we have people trying to keep a KDE branch that hasn't been developed in almost ten years alive, and actively using it (TDE). That's because, for all its flashiness, Apple and design fetishism, the super-disruptive community of desktop developers has failed to develop anything that's convincingly better than what was available ten years ago. The discussion, for some reason, is centered on whether the UI metaphors are adequate, ignoring users' feedback that focuses on far more obvious things, like "this thing KEEPS FUCKING CRASHING", "I just did apt-get update && apt-get upgrade and now all my applications look weird" and "everything is huge on my screen and this would look great on a tablet but this is not a tablet".
The Linux desktop today is far less fragmented, but that's because a) most of the people who could fragment it by developing fragmenting applications have long given up and use Macs and b) a lot of the traditional functions of a computer's desktop and applications have been eaten up by the web. There's little fragmentation to have when virtually all you use now is a web browser, the terminal and maybe a mail client.
Sadly, it never gathered much attention outside the former NeXTStep users crowd, save for a small resurgence of interest back OS X adoption soared and a lot of people began to actually like Cocoa.
Which is quite a shame. Mail.app, for instance, was really good.
Seems Darling is still being developed as we "speak". The Github repo is showing commits as recently as 2 days ago. And it builds on Gnustep in an attempt at supporting OSX software running on top of Linux from what i can tell.
GNUStep is still being developed, much to my amazement, but somewhat slowly. Back in 2002-2003, it was actually in use for some commercial development, too, so it's not like it's a long-abandoned turd. But the development rate is fairly slow and many applications that use it haven't been updated in a while. I expect it will be all but dead in the 3-5 years it will take the lovely members of the Linux Foundation to push Wayland everywhere.
The biggest hurdles to GNUstep's adoption were a) its lack of documentation (you were generally expected to use the NeXTStep documentation, although some functions were not implemented at all and others were slightly buggy) and b) the fact that Gorm and ProjectManager were really buggy for a really long time. This made GNUStep development not significantly more pleasant than GTK or Qt development -- not to mention difficult to get into for people who had never written NeXT or OpenStep software before.
There was a very brief period, between cca. 2005 and 2011 or so, when the fragmentation was less obvious because there was some degree of integration and cohesion (we had things like QtCurve, for instance). Then everyone started having delusions of grandeur again, in an early-90s-Unix manner, and things have been pretty much degrading ever since.
Today's "Linux desktop ecosystem" (God I hate this particular sequence of words) seems less fragmented largely because application development outside the major desktop environments has been largely abandoned, except for very relevant niches (photo/video editing, web browsers). A development that's largely unsurprising between KDE's architecture astronautics and Gnome's see no feedback, speak no feedback, hear no feedback attitude (that GTK, sadly, adopted for a pretty long time).
The unpleasant consequence is, of course, the "app problem" you speak of. A long time ago, the default KDE installation in Slackware 10 (which shipped with KDE 3.5, I think) shipped with a huge suite of applications, including graphical diff tools, VCS frontends, several multimedia players and so on. Two major rewrites later, they haven't re-accumulated this wealth of applications (and some of the ones that they do ship or advertise today are practically abandoned or remnants from the KDE 3 days). Some of the developments have been outright catastrophic, like KMail, which was turned from a very useful mail client to something that borks in a gazillion unpredictable ways as soon as you try to configure more than one account.
Similar things are happening in Gnome land, where they've chased feature-parity with Gnome 2 for years as they've been scrambling to fix everything that wasn't wrong with it and the horde of bugs that ensued from these fixes. Which, in fact, is why they have three forks in the first place. There was a lot of negativity about the Gnome 1 -> Gnome 2 transition, too, but that never resulted in forking. Nowadays we have people trying to keep a KDE branch that hasn't been developed in almost ten years alive, and actively using it (TDE). That's because, for all its flashiness, Apple and design fetishism, the super-disruptive community of desktop developers has failed to develop anything that's convincingly better than what was available ten years ago. The discussion, for some reason, is centered on whether the UI metaphors are adequate, ignoring users' feedback that focuses on far more obvious things, like "this thing KEEPS FUCKING CRASHING", "I just did apt-get update && apt-get upgrade and now all my applications look weird" and "everything is huge on my screen and this would look great on a tablet but this is not a tablet".
The Linux desktop today is far less fragmented, but that's because a) most of the people who could fragment it by developing fragmenting applications have long given up and use Macs and b) a lot of the traditional functions of a computer's desktop and applications have been eaten up by the web. There's little fragmentation to have when virtually all you use now is a web browser, the terminal and maybe a mail client.