The role of IT

In reviewing some of the comments on my Sun Ray chapter headings series I became even more convinced than before that the single biggest differentiator between the data processing view of computing and the Unix view derives from the difference on role conceptualization for IT.

As a Unix guy my view of IT's role is simple: I think it forms, or should form, a kind of communal memory and instantaneous communications network for the organization it serves - and that's a role conceputalization in which IT is a revenue generator, not a cost center.

As a result the most difficult balancing act for IT leadership involves ensuring that users get to do whatever users want to do provided that what they're doing doesn't contravene law or explicit corporate policy. In other words, the IT answer is always "Yes," unless we know that doing whatever it is contravenes legitimately established and communicated corporate policy or applicable external law. It's like being in the military: an absolute duty to obey "those appointed over you," unless doing so knowingly contravenes applicable national law or the code.

The data processing view is much simpler: process data at the lowest possible cost to produce needed reports within the time limits set in the service level agreement -a simplicity of focus dating back to the 1920s on which today's required ritual bowing in the direction of business support is just so much patina.

It's this role conceptualization that really determines how the smart display architecture differs from a Microsoft client-server deployment - and whether you use Wintel PCs, thin clients, or Linux PCs as your desktop clients affects cost, but not structure and focus.

In fact one of my favorite analogies: the probably false assertion that if you place a frog in hot water it will leap out, but if you put that same frog in cool water that's being heated slowly the frog will let itself be boiled, applies nicely.

What happened to large client-server systems during the late ninties was that such deployments got to be so complicated, so unreliable, and so manpower intensive, that the total cost of deployment started to zoom out of control and was only reined in by the evolution of a series of best practices - all of them focused on control centralization. By 2001, therefore, Julie Smith David, David Schuff, and Robert St. Louis were able to show rather conclusively that the total cost of a Microsoft client-server deployment decreases as these practices are applied; i.e. with increases in desktop control and processing centralization. (see: Managing Your IT Total Cost of Ownership, Communications of the ACM, January 2002, pp101ff)

Basically, what they found and what has now become the accepted basis for PC best practices in most large organizations, is simply that the cheapest way to run a few thousand desktop PCs is to turn them into expensive terminals - and that's exactly what most major users do when they implement a completely locked down desktop, outsource the helpdesk to isolate users from IT, and centralize processing in data center rackmounts.

The result of course, works exactly like a thin client deployment - except you continue paying for PCs and users continue enjoying the heat, noise, and downtime associated with having a powerful, but cheaply built, computer with a sixty million line OS on their desktops. Bottom line: think of the user time, frustration, capital cost, software licensing, and support cost disadvantages of the PC relative to the thin client as the cost of heating the water gradually, and you've got it.

So why is this tolerated? I think the answer is that the frogs involved have so many people telling them how cool and soothing the water is, that they don't internalize the reality that they've actually been boiled; i.e. that the need to manage the complexities of Wintel deployment have made it possible to turn the typical corporate PC desktop into an expensive and generally disfunctional modern version of a 327X terminal.

Luckily it hasn't happened everywhere. Where data processing is the empire, Linux has been the rebellion - with support from the entire Unix community including MacOS X and Solaris.

All of which brings us to an interesting group of tea leaves floated out by Sun CEO Jonathan Swartz in his Sun blog:

We see no global slowdown in IT. Despite what one competitor said. Our key customers (those that view information technology as a competitive advantage, not a cost center) are continuing to invest. They're investing to drive on-line relationships, fuel competitive advantage, and drive efficiencies - but mostly they're investing because they see a return.

It's too early to tell, at least without access to their numbers and market feedback, but it's possible to read this as a sign of hope: of the idea that more companies are investing in IT on the revenue side, not to save money, but to make it - and therefore as a sign that data processing's view of IT's role as cost center may yet be challenged by Unix.


Paul Murphy wrote and published The Unix Guide to Defenestration. Murphy is a 25-year veteran of the I.T. consulting industry, specializing in Unix and Unix-related management issues.