I have known a lot of network administrators/IT support people over the years. Heck, I once served in that role at a small board-level hardware shop. Not that it was in my job description. Management gave the job to the busiest guy, hoping he would do it. In all my conversations with administrators over the years (including myself), the common thread is the pain of supporting desktop computers. Whereas servers require routine maintenance and patches on a weekly basis, desktops require almost constant care and feeding.

Of course, today's desktops suffer from bloatware--the subject of my last column--which places undo stress on the administrator in the form of licensing, upgrading, and instability. They are also susceptible to all kinds of back-door security problems like viruses and spyware. On the server, you can install one patch that protects 1,000 desktops. On the desktop, you have to install 1,000 patches. And it gets worse as desktops become laptops that travel and connect to the network remotely. Once a laptop leaves the building, the administrator has no knowledge of where it goes on the Internet, and the games it plays when it gets there. The admin just has to deal with the aftermath of these activities.

For a decade, administrators have proposed a simple solution these problems: Put the desktop on a diet. Give users access only to the software and networks they need in order to do their jobs. As I said in my last column, this is easier said than done. To do it right, you need a lot of custom programming, and what small business wants to spend more money on programmers when it can just hire more administrators? So over the years small and medium businesses (SMBs) just hired more support folks to handle the increased overhead of administering mostly Microsoft products.

Microsoft has always claimed that the investment in extra administrators paid off in increased office productivity. According to Redmond, all these extra features in Windows and Office help workers get much more work done per hour. For every extra administrator you hire, you can fire three office assistants. This was true in the '90s, when Windows and Office improvements could be measured in productivity gains. But as the products have matured, extra features actually decreased productivity by increasing downtime. Because unnecessary features make PCs less secure and less stable, users have to sit idle waiting to have their machines cleaned and fixed. As SMBs look to cut every cost they can, they look to a new approach.

As early as 1996, so-called thin-client computing was hyped as the next big thing in IT. According to the hype, we could give users network appliances that were little more than dumb terminals with access to applications that ran on the network. Users would have access to all and only the applications and data they need. The applications could be upgraded and maintained once and for all on the server. No more bloatware, no more patch as patch can, no more desktop pain.

The problem with thin-client computing was it didn't work. Networks were just too slow to run whole applications over them. Productivity took a huge dive as workers waited for the network to refresh their screens. And users needed access to local resources when the server was down. A desktop goes down and it idles one worker. A server goes down and it can idle a thousand.

Users hated thin clients. Computers are an extension of their offices. Give them a thin client and it feels like being moved from a window office with a door to an interior cube by the elevator. Administrators expect users to whine, but the whining is manageable when it comes from one or two individuals at a time. When everyone in the company sends you an e-mail at the same time, you just want to crawl under the raised floor of the data center and never come out. Or you can give them what they want: the fattest desktops with all the applications your boss will allow them to have.

In May of this year, just when we thought thin-client computing was dead, IBM introduced Workplace, a thin productivity suite based on Lotus Notes, Domino, Workplace, and a couple of WebSphere products. Customers can rent as much of the suite as they want at a cost of around $2.00 per month per user. The product is automatically updated and maintained. And it can run either on fat clients or from a server to a variety of client types, including PDAs.

Analysts gave the new suite mixed reviews. Some called it a Microsoft killer because it will inevitably eat away at Microsoft's Office cash cow. Others said it was just the same old thin client bull they debunked eight years ago. I tend to think it will be somewhere in the middle, not because of my well-known opinions of Microsoft or the little-known fact that I currently work as a contractor for IBM. I think it will eventually cut into the Microsoft Office hegemony, but it will take time.

It won't be a Microsoft killer just yet because of general Office and Windows momentum. Sure, companies are fed up with Microsoft's hold on Office. It's overpriced; it's updated too frequently; it's bloated; it's too closely tied to other products. Office and Windows represent almost half the cost of a new machine, and a large chunk of ongoing maintenance and virus protection, not to mention licensing headaches. So there is a real need for this, and a growing number of companies will dump Office for the first better option. But users will not like it. And initial overhead will make a lot of companies reluctant to change until they have to. So adoption will be slow at first and it may not seriously cut into Microsoft's share for a few years.

The main problem with thin-client computing eight years ago was timing. Network technology was just too immature. A lot has changed since the mid-'90s, however. For one thing, networks are a lot faster. While desktops have continued to validate Moore's Law, networks have kept pace and then some. Ethernet has gone from 10-BaseT to gigabit. Storage-area networks are now affordable. Even fast Internet is becoming pervasive. Also, common free server operating systems, such as Linux, have matured to scale to 16 processors that support logical partitioning, multithreading, grid computing, and other sophisticated performance management features. And a new generation of computer science graduates have learned Linux as their primary operating system. In short, you can now cheaply run applications across a network without much of a performance hit, provided that the apps are not bloated.

The other key ingredient in the new thin-client model is client flexibility. In the old model, it was thin clients or nothing. This model is client-neutral. You can run it on everything from a PDA to a workstation. Lots of people need fat clients for graphics and other development work. Others need powerful laptops for sales-force automation and other data-intensive tasks on the road. Why deprive them of their machines just because of a rigid compliance to some artificial client standard? But, while this model does not require thin clients, it does reduce the burden of office applications on the client, which reduces the need for PC upgrades for a large class of users. And it allows those with minimal client needs to work on the road. In short, it should help companies save costs in the long run while increasing productivity in the short run.

And it should keep administrators from needing to hide under the floor of their data centers.