I was pointing out that writing good APIs and tools for developers is effectively part of building the OS, and there is much to be done in this regard. You are complaining that Microsoft won't build versions of its tools for Linux or that they won't port Linux tools over the bare bone NT kernel. One is orthogonal to the other. But in fact, I do agree with you that it would have been great to have all those tools ported both ways, except I don't have much faith that this will ever happen.

So, would you rather prefer to have 1000 homebrew versions of the .NET framework and DirectX, for that matter? Boy, that would be one big mess and hell of a bloat. Or better even, would you rather not have any of that, and instead let each developer reinvent the respective functionality in every application they write? What would possibly be the point of that? Actually, fewer versions of each and every library means that the best developers can spend more of a focused quality time finding bugs in it and improving it rather than spread themselves thin between the 1000 different alternatives..

OS is not for the users, it is for the developers. When was the last time your mom was pinning memory, loading a texture, or creating a security token? Applications are for the users. The job of the OS is to lure the application developers with lots of great APIs and cool new features to play with. Once the developers are there, the users will follow.

Actually, there's plenty to improve in a typical OS: making the OS more componentized, programmable, adding new layers of APIs for different functional domains, and otherwise supporting the developers that write code for that OS, so that they can be more productive and write more functional code in a fraction of time. For example, things like COM, WMI, DirectX, .NET, or the new WDF toolkit for driver development in Windows Vista. I don't see how you can separate any of this from the rest of the OS. The job of the OS is to bridge the gap between the developer and the hardware, and this is all part of it. And all these things have continued to evolve and will probably keep evolving for a very, very long time.

Mono had a lot of time to catch up, but as much as I respect the effort of the Mono team, the progress has been painfully slow. It took several years to even get the implementation of generics debugged. The last time I checked, implementation of web services (olive) was mostly one guy's job, and not even in the core system. Meanwhile, the gap is widening, and Microsoft is moving fast. At this point, nobody in the sane mind except the hardcore free software fanatics will ever choose Mono/Linux over .NET/Windows because it just makes no sense: Mono is epochs behind .NET in terms of functionality, and has negligible user base. How exactly is Mono going to catch up? Turn on a hidden antimatter quantum warp drive, or some other secred weapon from Sci-Fi novels? The Linux community has not embraced this technology, even though it was clearly a marvel of engineering and popped out just in the right moment. Indeed, some folks out there are still not sure that it "really flies". Nevermind that .NET is spreading to one platform (PCs, PDAs, gaming consoles, browsers, OS scripting, headless servers), programming languages and paradigm (dymamic languages, functional languages, you name it) after another. With this sort of attitude, surely it will never fly on Linux. How unfortunate. The Linux community had a huge head start with Java, but is blowing it big time, too. By the time this community realizes that managed languages were the key technology to focus on, .NET will be light years ahead. Who cares about superior package management if the development tools lag behind because the community has not actively promoted and developed and a single consistent modern development platform?

You seem to be forgetting that a software ecosystem does not consist only of the OS vendor, it also includes millions of developers who create software for that OS, either as contractors, or as independent contributions to the community. If those developers want to migrate, as they certainly seem to, then it is better for everyone, including the army, to do the same. Besides, good software is not written by monkeys in the basement of a factory building, but rather by talented people with enthusiasm and vision. Why would any talented person with enthusiasm and vision want to keep maintaining Windows XP for the army? It is better for the army to migrate, so that they can get a quality product built by quality software engineers. They certainly have the money to pay for that.

Windows XP is no longer being actively developed, and Microsoft cuts support for it in a few years. Upgrading to a supported and actively developed version of the OS seems like a good idea to me.

Actually, it is more like a newer version of Windows Vista/2008, with two years worth of of bug fixes and optimizations to the core system. The interface does seem to be noticeably more responsive, perhaps even more so than Windows 2008 used as a workstation, which has already been a significant improvement over Vista. The RC version released a month ago has been very stable, I dare say more so than the RC versions of Windows 2008. From my perspective, your rant seems totally disconnected from reality.

Java does nothing to solve problems that occur with programs written in Cobol, so what does that prove? If the old architecture is flawed, then best thing you can possibly expect is for them to build a completely new architecture, with support for all the APIs that the developers used in C++, so that the need to use the obsolete architecture is eliminated. And this is what you're getting. I can't think of a single major API that doesn't have a managed replacement, either in place or rapidly emerging. Now, as much as the 15-year old COM architecture was obsolete, the Linux community never even tried to developed a proper object-oriented set of system interfaces. Why don't you propose something superior? From what I see, the Linux community has nothing to offer. The only thing comparable to .NET is Java. Remind me what part of the Linux kernel or core Linux ecosystem is written in Java...

Instead of asserting the alleged inconsistencies in my response, why don't you state what you are trying to prove because I somehow fail to see any coherent message. 1. That Windows lacks mechanisms that prevent DLL hell and should be blamed for that? I pointed to the existence of .NET as well as to the fact that Microsoft is pushing it wherever it can, on all existing and emerging platforms, and even in the driver framework. Can you respond to this? 2. That Microsoft should be blamed for the fact that many developers are still coding in C++? I've pointed to the fact that they have virtually no excuse for doing so. Microsoft is not a church that controls the minds of its followers. I'd like to see how Linux manages to maintain discipline in its developers community if it grows to the same market share. I see discipline in the developers community lacking even today. 3. That there are some problem with .NET because Microsoft is internally using C++? What does that prove? Certainly this doesn't prove in any way that developers should be doing the same. Does it prove that .NET is not usable? I fail to see that logical connection. As you yourself have pointed out, perhaps Microsoft doesn't have any problem with things like DLL hell, and since the codebase is controlled by one organization, perhaps it has no compelling reason to switch to coding everything .NET within just a few years from the introduction of this platform. Besides, what is Linux kernel internally using? Last time I checked, it was written mostly in C, a retarded programming language that lacks any modern features. Microsoft introduced things like COM some 15 years ago, and even if it's C++, its internal structure is much more object-oriented that Linux ever was with its archaic filesystem interface. Following your line of reasoning, one would argue that since Linux is not internally coded in Java, therefore Linux lacks any sort of high-level programming features and is therefore useless as a platform for developers.

The majority of people can't count to 10 and believe the universe started 6000 years ago, but that's not the problem with modern science, that's just their inertia. Windows has been aggressively moving into .NET on all fronts, and is introducing .NET all over the place: as a way to interface the OS (does Linux come with a consistent set of interfaces like this for all parts of the platform, or is it still all based on the archaic filesystem interface), as a way to interface all their applications, in Windows Mobile, in XBOX, in Zune, in Silverlight, and you're also recently getting driver kits that let you write managed code in user-mode. Sure, much of Windows itself is still written mostly in C++, but this process can't happen instantly with a large software organization such as this, eat your own dog food is not always easy. External developers don't have much excuse for not moving into .NET, other than their reluctance to learn new technologies.

You must be mentally still in the 1980s because I haven't seen a single instance of DLL hell for the last 15 years. In case you missed that, there has been this thing called ".NET" evolving for, like, the past 7 years, which had essentially solved the problem once and for good. Are you still running Windows 1.0, on a steam-powered, 3-bit CPU, or are you just watching a lot of porn these days. If the latter is the case, then you might consider avoid that "premium" content that requires additional "codecs".

I think the fact that there are people out there who aspire to do mundane jobs is a blessing, we should thank them for being there and not insult them for no reason. From my perspective, I see around myself all kinds of primadonnas that don't know anything and won't get their hands dirty because of their fucking princessdom. Just about every M.Eng. student at my school (no offense to M.Eng. students) believes it is below his dignity to do anything less than reinventing polymorphic tensor models for quantum strangelet algebras over functional operating system kernel manifolds in Banach spaces, but they can't write a 50-line program or burn a fucking CD. People whom I could ask to install stuff, run those damned experiment for me, fix my crappy website, or grpah my data for me without showing attitude, and be happy that they could be useful, I praise such people and I wish there could be more of those. Instead, all I get is guys who live in some kind of virtual reality and just can't honestly get in touch with their skill set.

Wrong, MS matches 1:1 donations of its employees. What do Google or Google's founders do, beyond being in vogue?