MS finally offers vision of network computing

MS finally offers vision of network computing

The rise of application service provision will facilitate the rapid adoption of network computingImay be jumping the gun once again, but I suspect network computing will get a big boost this year from an unexpected source - an increase in Microsoft Windows-related security problems and viruses.

People aren't going to abandon Microsoft software anytime soon, no matter how badly it's designed. (You know, this whole phenomenon reminds me of the old Buick slogan - `Wouldn't you really rather drive a Buick?' I always imagined that the slogan could easily be preceded by something like `Sure, you can have a well-designed automobile that is safe and efficient, but . . .)?

One solution is to reduce the problems invited by Microsoft software by consolidating your processing power and putting it at the server. This doesn't eliminate the problems. You still have to run antivirus software, for example. But imagine how much easier it would be to react to a virus if it is restricted to a ser-ver farm instead of spanning hundreds or thousands of client computers in addition to the server farm?

Even Microsoft can see the writing on the wall. That's why the company has promised to deliver its own vision of network computing. Microsoft's technology is still less sophisticated than what you can do with X11, so I expect its approach will have to mature. But you don't have to wait for Microsoft to catch up, as you'll see in a moment.

Let me throw in one more determining factor that should contribute to network computing's success: the rise of the application service provider (ASP). I promoted the ASP concept as far back as February 1998 (although I suggested the needlessly complex acronym ISISP, which stands for Information Systems Internet Service Provider). It may take another year for ASPs to really get started; but as ASP momentum continues to build, the idea of putting most of your applications on a server farm will seem more and more like a natural evolution path for IT computing.

Now let me throw in one more interesting variable: The Santa Cruz Operation's (SCO's) Tarantella. I predicted big things for Tarantella more than two years ago when I thought Java-based network computing would explode. I was obviously mistaken about Java-based computing (although it is only mostly dead - it has been feeling better lately).

But I still think Tarantella could be a pivotal technology for the growth of network-based computing.

Tarantella is middleware that pretends to be a client to a heterogeneous mix of applications platforms. It homogenises those client platforms by delivering the presentation of all the applications to a Java-based client.

OK, let's do that again in English by looking at a practical example. You want to get your users off Windows boxes and provide them instead with thin-client access to an office suite running on Windows NT Terminal Server Edition. You also want your users to be able to run a CAD package which runs best on a Sun Solaris server.

You'll run into several problems if you try to do this by providing your users with Microsoft's thin-client protocol RDP and/or the standard Unix protocol, X11. Your users will likely have to authenticate with multiple systems, and they will be stuck with the finicky performance characteristics of both RDP and X11.

You can reduce your problems by choosing only one protocol, but an ASP probably won't have that luxury. So put a Tarantella server between the servers and the clients. The upcoming version of Tarantella will understand both RDP and X11 and will present both types of applications to any Java-enabled client. And the Tarantella protocol automatically adjusts to the available bandwidth.

Look at it this way. You can log in once from your Java-enabled browser at work. Tarantella will detect and use the available bandwidth as you run your office suite and CAD package. You log out without closing your applications, drive home, and log in again from your home PC connected to the Internet via Digital Subscriber Line or cable modem. Then you pick up where you left off in both the applications you were running at work.

Anyone who has tried to use network computing should be able to see immediately how much nicer Tarantella can make the network computing experience. So if SCO prices Tarantella right, it could be just the thing that tips the scales in favour of the future success of ASPs. I'll be watching.

Nicholas Petreley is editorial director of LinuxWorld ( Reach him at, and visit his forum at

Follow Us

Join the newsletter!

Error: Please check your email address.
Show Comments