Writing about the best and worst in operating system is like a crap magnet: I'm pressing the big red button. People develop a personal relationship with operating systems, whether on servers or their personal machine or phone. The love/hate relationship becomes anchored with deep emotions about the merits/detractions of the devices they use-- through the lenses of operating systems.
Great in this case means possibly really great, and alternatively, really awful. Some made us happy. Others made us sad, or worse, mad. Some we toiled happily with the winners, while others burned up precious weekend time in fits of compatibility issues, installation nightmares, and startling kernel traps and various blue/black/red screens of death. Some deserved to become dumpster fodder, and others are still humming away, quietly, and doing their job. This is about both kinds.
Windows Millennium Edition (WinME)
This was the last version of Microsoft Windows that ran on top of Microsoft DOS, and it wasn't well-designed. Microsoft released it as a stop gap version to address slightly more memory and disk before the two Windows code bases would be merged together into Windows 2000 client and server editions. Technically, it arrived late in the 1990's, but its inclusion here is to remember the pain of the name.
Initially, it was boasted that Vista would be the most costliest-ever operating system to develop, and the results were uniformly disastrous. Confusion over what hardware characteristics would be needed to run Vista, lack and dearth of appropriate hardware drivers, OEM confusion, perceived (generally untrue) software compatibility issues, and vastly mixed messages caused this operating system to be avoided by most businesses, and used mostly by hapless consumers who had no choice when they bought a system. Although an architecturally safer system then predecessor Windows XP, Windows XP SP2 heralded an similar architectural change to Vista that demoted the user from root, and stanched Windows security problems, although certainly not altogether.
The world was looking for the joiner of Novell's time-honored and rock-solid NetWare network operating system to be joined fully to Linux. Novell had just purchased SuSE Linux and it looked as though the world might have a powerhouse to rival the initial foibles of Microsoft's then-embryonic Active Directory. Would eDirectory become a rival and gain authentication market share. Today, that goal is an unfulfilled dream for most.
The 'smartphone' hardware genre changed everything. When PDAs were merged into cell/mobile phones, the list of possible winners were neck and neck with each other. Palm, RIM, Symbian, and other rivals raced forward only to be stopped in their tracks (like Asimov's 'Mule' in the Foundation Trilogy) by Apple's iPhone. Now iPhone is the one to beat, and Windows Mobile lags behind dramatically. Even my own HTC Touch Pro is a poor excuse for a phone, though I've become used to its foibles under Windows Mobile 6.1. Embarrassing at best, way behind at worst.
If you've used Linux, you've used the GNU utilities that are the bedrock of Linux distributions. the GNU rewritten Unix utilities tool set was invented through the purity in effort of Richard Stallman-- the pillar of free software. Yet the GNU Hurd kernel, the ultimate free re-write of the Unix kernel, has languished. Yes, something like it can be used inside of the Linux Debian-Hurd construct, but almost no one does this. The pure kernel, it seems, may never see the light of day. It's the ultimate in free vaporware.
Great Operating Systems
Windows Server 2008 R2
Linux adherents and Mac fanboys may cry foul, but Microsoft's latest server operating system (for all of its sole-source lock-ins) has the best in terms of enterprise federated security and authentication. For those that use its admittance controls, there are ways to audit and vet logged on mobile users in ways that take lots of components and manual construction in Linux, Solaris, and BSDville. Yes, it's damnably captive to Windows platforms and treats all others as second class citizens. This is Microsoft we're talking about here, and their business is to sell operating systems and allied components. It's not easy to nominate them here as their business practices aren't very kind, but in this case, Windows 2008 R2 is an embarrassment to enterprise Linux distro makers everywhere. Microsoft's Hyper-V was late, but it's made inroads in terms of enterprise NOC hosting configuration. Even a leopard can change its spots, sometimes as scar tissue.
It just works. Darwin BSD underneath, mostly luxury on top. The upside is beauty, quietness, control, and stress-free existences. The downside is that it isn't a business plan for computer consultants and virus removers. Onerous is the fact that the most recent release of MacOS-- Snow Leopard-- had a sufficiently large number of post release patches to make our PTSD of Microsoft Windows patching come to mind. Apple's QA now faces a bit of what Microsoft does: so many hardware platforms that QA is difficult as Apple releases new hardware platform variants. The OS isn't pricey, and this isn't about hardware captivity, this is about quality and architectural philosophy in an operating system. Yet MacOS is also the underpinning for the cell/mobile OS to beat on the iPhone. Attention to detail pays.
This is another case of excellence through simply being rock-solid. Sun (soon Oracle) has built the most reliable Unix-derivative. It's boring in some ways, until you get to components like dTrace, Sun Containers, and ZFS. The technical innovations never capitulated to Microsoft Windows, and Linux and BSD developers have an instant affinity once they explore Solaris. Now, Solaris has become open and comparatively 'free'.
GNU/Linux (especially 2.6.18+)
Never has their been such an uproar in computing as a free kernel and free utilities-- all done very well with rapid, mindful if darwinian skill. Linus Torvalds crafted Linux, and has been holding on for dear life ever since. Coupled with the GNU utilities and two main window manager branches (Gnome and KDE), Linux underpinnings now grace objects from tiny wristwatches and clever cell/mobile phones, to IBM mainframes and everything in between. The promise of Linux for civilians is slowly but surely being realized through distros like Ubuntu, Novell/SUSE, Mandriva, Knoppix, and others, but the enterprise server market belongs to Red Hat, Novell/SUSE, and communities formed around each of these. That doesn't mean that there isn't worth in the literally hundreds of distros out there.
An open OS for phones has been difficult. OpenMoko tried but got little traction. Android may challenge, along with other open platforms, established platforms. A natively 'jailbroken' open phone will test carrier promises to just deliver wireless pipe. Breaking old models invariably causes reverberations in the marketplace that will swamp some, while others surf.
And there are more of both. Discuss. Please, no swear words.