Here's the full text of something by frequent contributor Anton Philidor during last week's discussion of various approaches to Lintel/Wintel TCO comparisons:
"Linux is more expensive than Windows."
By mid 2004, however, enough people had bought into Red Hat's non license licensing to make Linux arguably more expensive to license than Windows...
There was a time when Linux was considered free software instead of the commercial product of two companies.
Because one of those companies was Novell, which competes with Sun to produce the least financial gain from estimable products, Red Hat was well on its way to a monopoly.
So Microsoft decided to sell Novell's Linux, and proved to be the best sales staff the software ever had. But that has ended.
IBM helped Novell buy SuSE while carefully evading liability for open source IP violations, and then ignored the company when it was in difficulty. Perhaps because IBM treats Linux as a way to reduce costs or as part of a bait-and-switch scheme rather than software to be sold.
So now there are one-and-one-half Linux distributions (SuSE being the half and inevitably fading away) in the commercial market, the only one that produces enough money in sales to matter. The remaining Linux distributions are for enthusiasts.
It's reasonable, then, to simplify: Linux is Red Hat.
Given this reality, how is it possible for Linux ever to be substantially less expensive than Windows? (Over the typical corporate time period.)
I don't share Anton's view of reality - and in particular I hope I don't usually argue that assuming something proves its truth - but he does ask an intriguing question: "how is it possible for Linux ever to be substantially less expensive than Windows?"
As asked that's a no brainer: if you're not dumb enough to pay someone like Red Hat to impose a license on you, Linux really is free - meaning that it's always possible to get Linux for less than Windows.
To extend that simple argument over the life time of the system you need to make a lot of assumptions - but reasonable assumptions produce reasonable results: i.e. if you're fair about the assumptions, Wintel falls further and further behind Lintel as you extend the time frame simply because the cost of acquiring, installing, and supporting Linux applications is always less than that for Wintel applications.
(Note that the Wintel people frequently argue that it takes about as much time to administer a Lintel system as a Wintel one and that Lintel skills cost more -thus giving Wintel a long term cost of support advantage. In reality, however, a properly configured Lintel machine will typically require no further support until the hardware fails or something changes - and as the average MCSE ages his expectations about salary, benefits, and working conditions rise to the point that the new grads being hired to run Linux now cost most organizations considerably less than their more senior Wintel colleagues.)
It's at this point in the analysis that you hit, however, the limitations built into the question as Anton asks it. Specifically, "how is it possible for Linux ever to be substantially less expensive than Windows?" focuses on cost when really we ought to be talking about expected benefits net of costs. Try to do that across a range of Lintel/Wintel usage scenarios and you'll discover that you simply can't make a case for Microsoft on most server applications but can make one for Microsoft on most desktop applications.
Notice that I said "most server" and "most desktop" applications because these generalizations don't apply across the board. What does seem to produce a somewhat more general differentiation is whether Microsoft or open source leads with respect to the core application in a bundle.
That leadership can be expressed in two ways: through market dominance and through what the product does.
As far as I know the office suites offer the clearest example of this effect with respect to leadership achieved through market dominance. What happens there is that there's no objective difference in benefits obtainable through use of either StarOffice/OpenOffice or Microsoft Office, but there is a consistent difference in the cost of ownership because open source users are forced to incur the costs of being different - providing additional training, re-assuring users unhappy about being held out of the perceived office automation main stream, and converting third party documents to and from Microsoft's proprietory formats.
Look, however, at the product areas where Microsoft routinely struggles to catch up to the open source competitor -essentially all server applications from the OS up - and you see that technical leadership can have the same effect: forcing someone doing a long term TCO study to add the costs of following - in these cases often expressed as foregone benefits - for those not choosing the industry's leading applications.
So what's the bottom line reality here? simple: Lintel is always cheaper than Wintel to get into, but it usually only wins on long term TCO where the applications, like Apache and Firefox, lead Microsoft's and, conversely, it loses on long term TCO where the applications, like StarOffice and Gnome, follow Microsoft's lead.