I particularly liked this part:

Windows 2000 Holds Networking World Speed Record As part of the bi-annual Internet2 Land Speed Contest between corporations and research institutions, senior networking engineers from Microsoft put Windows’ performance to the test. With two off-the-shelf PCs running Windows 2000 Professional, the PCs sent a stream of random text coast-to-coast at 750 megabytes per second - faster than ever before on a land-based network. That's fast enough to carry all the music on a CD across the U.S. in six seconds, approximately 15,000 times the speed of a standard 56 kilobytes per second modem.

I want to know where they live that the standard modem transfers 56 kilobytes per second. I'd also like to know how they consider a jump in network speed record to be solely attributable to the computer at the end of the conversation. Also, although admittedly I didn't read all of the documents, none of the benchmarks I see run on the same make and model of hardware even though there are Unix and Unix-based OSes which do run on the same hardware as Windows. They make a big deal about Solaris and AIX on proprietary hardware being slower per cost and calling it 'analogous hardware'.

Also, they call it a Windows vs. Unix TCO comparison, but...

• They often put in a section header saying something like "Better Reliability". They then mention that they reengineered NT 4 to come up with 200o or 2000 to XP, or 2000 to 2003 in order to make it more reliable and thast that worked. Well, hooray for them, but that's misleading and has no place in a Windows vs. Unix comparison, as it's a comparison of two versions of Windows. No improvement stats ar egiven for any Unix implementation.
• They mention uptime guarantees given by Windows system integrators and omit any info about similar guarantees in the Unix market, which obviously exist. They are trying to intimate that they are the only OS vendor whose product is covered by integrator's guarantees.
• They make severl claims about Windows devoid of any comparisons to Unix of any flavor, and sometimes without any quantitative analysis.
• There seems to be this big attack of some form of Linux by the "impartial" IDC, commissioned directly by MS. I see no reference to a particular product, distribution, configuration, etc. These were Windows shops that had recently adopted Linux, too, no doubt.
• A five-year TCO study on a 3-year old OS? Please! Not to mention that you lose support form MS after the product has been on the shelf for 5 years and must pay your license fees all over again. That leaves two years for Windows 2000.

All in all, these papers are good reading if you're practicing developing your own FUD campaign.

Christopher E. Stith