This article is wrong in many areas.<p>First of all getting the NT kernel and building upon is a thing you can do without getting the source code. The driver development kit of Windows is order of magnitude better in terms of documentation and ease of use (as far as kernel development is easy) than the Linux or BSD driver kit (I've worked on all of three).<p>You don't really need the source code of the kernel when you have all the documentation you need and a powerful API.<p>If you really need the source code for what you are doing, Microsoft gives access to this source code. You get a remote access via smartcard and can browse through the whole source code (remember the 2000 source code leak).<p>The source code availability, is really not, I think, the explanation why Windows NT doesn't dominate the server world. You don't need to recompile or tinker the kernel to administrate a Windows machine, this need simply doesn't exist.<p>The reason is that it took a lot of time for Windows to adopt a server philosophy. If you've played with Windows Server 2008, you realize Microsoft is getting there.<p>One of the things I like with Windows Server 2008 is that it doesn't install the whole universe and let you precisely pick what you want to deploy on your machine.<p>The addition of a more powerful shell and a more transparent administration panel helps as well.<p>Another problem is the very rich, but complex, ACL (access control list) system of Windows NT. Although you can deploy extremely fine grained and subtle rights and authorizations with this system, in UNIX you just type "chmod 0755" and 99.9% of the cases you're fine with it.<p>Last but not least, Windows carries a bad reputation in terms of reliability and security.<p>In terms of performance or features, NT has got nothing to envy to Linux or BSD.<p>The reason to me is really a "packaging" problem, the technology is capable.