Username or EmailPassword
I completely agree with you that weirdly-capitalized-NeXTSTEP is by today's standard what should be emulated in terms of tightness, flexibility, lightweightness (is that a word in english? I'm from Quebec too like that other guy above..), FUNCTIONALITY, and solid architecture. The object model is solid enough that it's actually not dependent on a single graphic display engine, unlike GNOME/KDE/etc who are still dependant on X11.
It's true that the current needs of computer users couldn't be served that well by a 640k PC, provided that you need at least to be able to interact with users having more recent OSes, but one thing that has been lost in the hardware performance expansion is the tightness of code. Embedded people still live in the age of 640k, 512k, 32k... and they are among the smartest coders in existence for being able to squeeze every bit out of their systems. In the so-called olden days, there _was_ a macho pride in doing much with less (remember the Demo scene), but the ability to do great tight OSes/application has vanished from the mainstream.
I use Mac OS X at home and GNUSTEP on my Slackware machine at work, and I dream of having a world united in the OPENSTEP specification. The original architecture has been extended, but not patched to death and forever hacked like Windows. Which proves one thing: the OPENSTEP developer got it right, and they are still right. You can have a TCP/IP stack, a solid OO GUI, consistent UI, clear interfaces, flexible object model (see what Apple is doing with it right now!), cross-plateform applications, simple development, and easy dynamic library management, but NOT at the expense of tomorrow's computing power.
Yes, the original NeXT OS was demanding cutting-edge hardware at the time, but nowadays it does not. Windows, on the other hand, is still surfing on the upgrade wave, and the Linux GUI plateforms are making that error as well.
Honestly, if we could abandon the X11 implementation we have now once and for all, and relegate it to a compatibility layer outside of the core needs for a modern desktop OS, by following the *STEP model we could capitalize on a great architecture that is still relevant 10years later, and that WILL be relevant in 10years from now. For whatever goodness there is in the X11 protocol, there is a comparable badness in the X11 implementation. And please, coding a GUI in C is not useful anymore: it sadly leads to more bloat than tightness.
I'm saddened that even after reading stuff like the "X Window disaster" chapter in the Unix-haters handbook, which was written in the early '90s, no one has still taken care of leveraging a better display model, because that is perhaps the single most important cause of bloat and failure in the free *NIX desktops in 2005.