OPINION
Getting the Facts About Windows and Linux

I found a whole bunch of examples just like this the other day when I followed up on a Microsoft (Nasdaq: MSFT) ad urging me to "get the facts" on the relative cost performance of Linux versus Windows. Not all of those reports, available on Microsoft's Get the Facts site, are based on surveys, but they all illustrate the magical power of money and emotional commitment to prove all kinds of otherwise counter-intuitive things.

What's interesting about these reports isn't that they're intended to deceive, but that the deception is carried out by carefully selecting, slanting and wordsmithing bits of truth to the point that you can rant about omissions and unfairness but not demonstrate a pattern of deliberate lies.

For example, there's one report by IDC that almost perfectly illustrates the survey recipe for proving the Earth flat. This one, showing that Microsoft's Windows 2000 has a lower five-year total cost of ownership (TCO) than Linux , starts -- as usual -- with the assumption that TCO is just a matter of opinion. If you can establish that point by assumption, then proving the improbable is just a matter of interviewing the right people and doing a little hand wave over any conflicts that come up.

Three-Step Technique

Now you might think getting the right people for this would be hard, but it isn't. Just start with a list of people who have publicly declared deep personal commitments to Microsoft a few years ago, and you'll find that those who are still in the same jobs tend to have just the opinions you want to hear.

There are some minor issues in applying the survey methodology to this particular case. It's tricky, for example, to get reliable five-year cost data for Windows 2000 if you're doing the survey early enough in 2002 that none of the respondents can have much more than 18 months of cost experience with it. That's nothing, however, compared to the difficulty of getting accurate Linux cost data from them when only 40 percent of the companies interviewed admit to having Linux somewhere.

On the other hand, this is the kind of nitpicky thing most people don't notice, so just keep muttering the liar's mantra -- "don't ask, don't tell" -- and you'll soon find that proving Linux more expensive than Windows isn't any harder than proving that the world is flat.

The nicest thing about this basic three-step technique -- first, assume quoted opinions define facts, then pick the right people to quote, and lastly wave your hands to make any uncomfortable realities disappear -- is that it applies to everything from alien abduction to Windows security because there's never any shortage of people eager to strengthen their own faith by voicing it in public. But don't be misled into thinking that convenience makes this the only way, or even the best way, of honestly delivering dishonest results in service to a client.

There are at least two equally good or better ways of achieving the same results: omission or denigrative diminution, and sleight of word. Just remember that lying truthfully is like any other magic trick: You just have to believe with the audience that if they don't see you doing it, then it isn't happening.

Denigrative Diminution

In denigrative diminution, you pass off a true but negative statement about the stuff you want to hide as a discussion of it. The trick here is to create just enough of a negative perception to prevent your audience from wanting to explore an obviously unrewarding topic and then redirect attention elsewhere by quickly moving on to something positive. That way, the emotional response will block the intellectual response, and few people will notice that you skipped any discussion of the facts.

For example, here's how the widely used Lauden and Lauden Management Information Systems textbook describes Unix: "an interactive, multiuser, multitasking operating system developed by Bell Laboratories in 1969 to help scientific researchers share data." That's actually wrong, but the magic trick isn't in the errors -- which are probably unintentional. It's in silently dropping the next 34 years of continuous technical progress to enforce the impression of Unix as old, dated and geeky.

None of that's true for Microsoft, of course, whose insanely great Windows 2000 is used, according to the same textbook, as an "operating system for high-performance desktop and laptop computers and for network servers." Denigrative diminution can be very effective, but it's easy to go too far. In this case, for example, the authors go a bit over the top by ignoring security in their multipage paean to Windows but adding that "Unix also poses some security problems, because multiple jobs and users can access the same file simultaneously" to their second and final paragraph on Unix.

It takes real chutzpah to do this, relying on the dog that doesn't bark to denigrate Unix while hyping Windows XP in a college textbook. But sleight of word offers the ultimate professional challenge in the genre. What makes it so dangerous and exciting is that the artist pulls the audience into the act, showing them exactly what he's doing, while he's doing it, but in such a way that most of them will see something entirely different.

Questionable Numbers

For example, in another of those flat-Earth reports available from Microsoft's Get the Facts site, Meta Group introduces its proof that Windows Server is cheaper than Linux at the database level with a couple of sentences that are absolute marvels of sleight of word construction:

"Windows or Unix costs a few hundred dollars more per server than a Linux distribution (Red Hat, SuSE). For example, Red Hat's Advanced Server edition costs about $800-$2,500 per year (annual subscription license) depending on support service levels. Microsoft's Windows 2000 Advanced Server edition costs about $4,000 upfront."

As examples of the art, this is awesome. That linkage of Windows and Unix as comparables against Linux establishes authorial fairness, while the single word tacked on at the end, "upfront," makes the absurd comparison between support fees and license fees completely defensible. That's great stuff by itself, but the real genius lies in that phrase "a few hundred dollars." After all, what's a few hundred dollars in a systems decision? And what kind of curmudgeon would notice that "a few" ranges from 15 to 32 here?

The numbers are questionable too, of course, so next week I'm going to look at what Linux actually costs. But meanwhile I'd urge you to look at some of Microsoft's flat-Earth findings. There are valuable lessons there, on quoting fools, assuming results, omitting data, or fine-tuning emotive content to mislead that you simply won't find nicely collected in one place anywhere else.