@crumb:
Of course if your time has no value you can spend your life customizing and recompiling linux to make it run more or less decently.
Sadly linux apps and the OS itself sucks and openoffice crawls...
Yes OpenOffice 1.0.x crawls, but OOo 1.1rc5 is the fastest I've seen yet. 1.0.x took *years* to save a big document, 1.1 drastically improves start time (but not enough) and doc load/save time (also not enough, but at least now it's quite usable).
OpenOffice uses bugger-all of the Linux API infrastructure that is there. It's a bit like Mozilla. It uses its own entirely self-contained architecutre for everything from the GUI down to the String handling.
It is its own operating system.
It is not a good example of a lean, mean Linux app. that uses all the nifty shared libraries that I was talking about. Abiword and GNUmeric might be the next best thing.
As for the Linux kernel sucking, it's about using the right tool for the job. Embedded applications? Definately. Free, open source, rock solid, flexible, small footprint. Server/storage farms? Routers? File server? Web server? Application Server? Yes. The Linux kernel scales up fairly well from what I've read. Engineering workstation? Probably, if you have a dedicated admin or a support contract. Business desktop? There's nothing intrinsically limiting about the Linux kernel here, except that it is a Unix clone so you need in-house or contracted Linux expertise.
As for the GNU fluff around the kernel, it's great for non-desktop things. Engineering workstations? Yes, Linux is solid, has good SMP and support for lots of memory and huge files, but it might not be the best for the user (I think it is though, over the Sun CDE anyway, I've only ever used an IBM AIX app over SSH/X so can't really comment there, haven't used anything IRIX).
Business desktop? Probably no, unless you are a _very_ large corporation who can afford good responsive knowledgable admins prepared to do probably more testing/tweaking work than what would be necessary with a Sun setup.
Then we get to home users. Home users generally don't want to be Unix sysadmins. They want to play games and watch movies.
On the kernel side, that means prompt, direct, serious investment from hardware vendors into writing decent drivers, quickly, that properly utilise the hardware. Just like the Windows efforts. This isn't happening now, it isn't going to happen soon, if ever.
@alx:
Finally, i hate the way that the heavyweight desktop environments (Gnome and KDE) claim to be really innovative - but all they do is copy windows
On the GNU fluff side (Gnome and friends), the GUI design is acceptable. When the Gnome people say they are being innovative, they don't necessarily mean the GUI design. They are also talking about the technical side. The API, the internel design. For me, if there is any innovaiton in the GNOME project, that is where it is happening. It frustrates me that people write off something because they can't see beyond the surface. If they don't want to see beyond the surface, they aren't really qualified to write it off, now are they?
And finally, on the home user's desktop experience side, there will ALWAYS be problems. GNU/Linux distros are a clone of the Unix environment. This environment assumes a certain amount of technical knowledge. Assumptions like basic understanding of permissions, groups, file system layout, networking protocols, file system types, mount points, scripting, user account administration, package managment, how the kernel works (loading drivers), bootloaders (setting up a new kernel with LILO).
You could say MacOS X has solved these things. But the money in Linux is not on the home users's desktop, so these issues aren't going to be addressed for quite some time, if ever. The best that I am hoping for is Linux being adopted on the corporate desktop, and POS.