A few years ago, thin clients were all the rage. Leading the charge was Sun Microsystems, driven perhaps by a disdain for Microsoft, but many others were producing a variety of thin-client products on both the client and server sides.

While stationary thin clients never really caught on, I became - and remain - a proponent of wireless thin clients.

Now, I'll grant that anyone who's ever had a dropped call or experienced the high degree of variability in data throughput that's endemic to most wide-area (cellular) wireless networks probably thinks that moving to the thin-client model for mobile computing is just plain nuts. But let me explain.

If you've not experienced thin-client computing, it's very simple. It's just like using any other computer but with the processing and data on the server side of the network, and the clients looking and acting very similarly to the familiar Windows-based interface that we've all come to adopt.

The idea is that one shouldn't care where one's data or processor is; all we need to know is that we'll have access wherever and whenever we need it, with the same degree of utility and responsiveness that we have in a more classic client/server implementation.

And, indeed, on a hard-wired LAN connection, the thin approach should work just fine. There are many examples of successful thin implementations, and companies like Citrix, Hewlett-Packard and Wyse have been doing just fine in the thin-client business for some time. Microsoft has also been offering Windows Terminal Server.

Microsoft, and most IT managers, it would seem, prefer traditional (robust) PC architectures. You might also want to check out Real VNC, which essentially allows any kind of client to talk to any kind of server. This is very cool (and, in one version, free) technology. It's also a great way to reuse older devices that otherwise wouldn't be useful as a computing platform today, or to use limited-function mobile devices as thin clients.

But "PC" and "mobility" are usually in fundamental conflict. Even tiny notebooks are usually too much to carry everywhere. (For a really tiny notebook that's an exception, check out the OQO 01+ - which we review here) So for most, smart phones are likely the answer.

This combination of phone and PDA makes a dandy phone, but it's not a computer. And that's OK, because it will play nicely into the idea of a thin client. There's just enough horsepower on most smart phones to run a browser, and that's the key to a mobile thin-client implementation.

I really like the idea of Web services, as it combines the thin-client concept with the write-once/run-anywhere ethic that I think will ultimately be the savior of IT budgets everywhere. But notice how the pieces come together - browser-based thin clients that run potentially on any mobile device, accessing the same applications that are available on the office desktop. What could be better?

Throughput is the problemBut there's a fly in the ointment here. In the office, we have continuous, essentially predictable throughput. On wireless, we don't. What happens when someone needs connectivity to run a vital application but can't get it? This is the counterpoint to the vision I outlined above, and it's a more than valid point.

Thin-client mobile computing can't work until there's a critical mass of wireless connectivity, at least where a given worker needs to be most of the time.

As I discussed in my column on convergence, though, two networks really are better than one. I think the combination of cellular-based, wide-area wireless data and metro-scale WiFi deployments will become the preferred approach to providing an assurance of connectivity.

But we are clearly not there yet, and it will be at least a couple more years before we begin to seriously think about wireless thin clients as the default.

In the interim, I am experimenting with thin-client e-mail using the beta version of Yahoo Mobile e-mail on my Treo 650.

So far, so good. I don't worry about the storage limitations (on either the server or client side) imposed by services that push e-mail to a mobile device, and I operate on the same set of data, whether at my desk or on the road.

I'll have more on this approach in a future column, and more on microbrowsers in general, but for now, I have no plans to try any other mobile e-mail strategy, at least in a production environment.

And there is one other benefit of a mobile thin-client strategy - security. I'm sure you've heard about the bozos who stored critical, live and unsecured data on their notebooks, only to have these computers stolen and the data potentially compromised.

In a thin-client world, information is fundamentally secure, and that's another major incentive for adopting this strategy as wireless coverage becomes, if not ubiquitous, at least reliable.

Craig J. Mathias is a principal at Farpoint Group, an advisory firm specialising in wireless networking and mobile computing. He can be reached at [email protected] This article first appeared in Computerworld.