oday, the overwhelming trend in custom software development is toward systems that run on web browsers and web servers. But as recently as two or three years ago, companies requiring systems that could service multiple users were focused on "client/server" technology. What happened?

Just as the Internet has been the dominant buzz since the mid nineties, Client/Server was the favorite IT flavor of the eighties and early nineties. Let's take a look at what happened.

Historically, the shift away from Client/Server amounts to a 180-degree turnaround: from a trend toward distributing processing power to users' desktops to a trend toward re-consolidating that power in huge centralized data centers. Beginning with the first time a teletype was connected to a "timesharing" mainframe computer [circa 1961], multi-user software followed a simple pattern: all the processing capacity, all the memory and disk storage, resided on the mainframe; only a "dumb terminal" sat on the user's desk.

During the early 1980s, as personal computers became serious tools, people became accustomed to more responsive systems whose graphical interfaces made them easier to use. PCs rapidly lowered the raw cost of computing, and a compelling case was made for shifting software tasks away from centralized mainframes and onto personal computers distributed over a network. Users were no longer dependent on MIS for all their data needs.

Companies large and small set out to develop software systems that used Windows- or Macintosh-based "front ends" in conjunction with specialized database servers. Under this arrangement, the PCs on users' desktops that drew on the resources of the database server were called "clients," a sometimes confusing term that was perhaps a cold-war-era analogy to Soviet "client" states. Throughout the late 1980s there was fierce competition in the software industry to provide the tools that people would use to build client/server systems.

It was in this market that Oracle, Sybase, Informix, Gupta, and many other companies made -- and sometimes lost -- their fortunes. Corporate systems departments underwent turmoil as mainframe programmers were given early retirement and replaced by young PC programmers. The "hot technologies" now were not COBOL and Fortran but rather Visual Basic, Fox Pro and PowerBuilder.

Multi-million-dollar projects were undertaken to convert mainframe applications to client/server. IBM began selling its mainframes as mega-servers for corporate PC networks. Despite becoming the official religion, client/server usually didn't live up to the promise, proved difficult and expensive, and often failed. In retrospect, we may have thrown out the baby with the bath water (or the retirees with the pool water). Procedures, methodologies and tools for mainframe development, which had matured for 20 years, were hastily abandoned.

Trends and styles effect IT just like any other facet of corporate life. Lots of times, however, what's in fashion doesn't always make sense. When it came to the robustness that the "enterprise" required, PC networks couldn't support big corporate systems like mainframes could. While availability of distributed raw power and apparent ease of setup made corporate networks grow at a staggering growth, systems suffered. Though PC networks were improving, they were still plagued by security, backup, reliability, and uniformity issues. While long on promise, client/server was in trouble. The big problems of client/server came from (a) loss of central control, (b) complexity and fragility of PC operating systems, primarily Windows, and (c) lack of proven development tools and techniques.

The nineties have been a very interesting time for distributed computing. In 1990 no one could have possibily imagined how fast things would move, or how important the Internet would become. But when Tim Berners-Lee first developed HTTP and HTML (1990), he was just looking for a better way of keeping track of documents at European Laboratory for Particle Physics (CERN). Like a lot of good ideas, momentum started to build.

First there was Mosaic (1993), and later Netscape (1994) and Internet Explorer (1995) which took Berners-Lee's model and made it truly usable across a variety of platforms. The rapid emergence of the browser model of computing provided a glimpse at a simplified, cross-platform approach to computing that bypasses the complexity of PC operating systems while providing graphical point-and-click interface.

Lo and behold, there was another new paradigm, and this time a better balance seems to have been struck. Yet once again, everyone's scrambling to develop software using immature tools. Mainframe's heyday had 20 years. Client/server had about 12 years to evolve before the internet meteor drove it nearly to extinction.

At Chalem Systems Group we're very excited about browser-based development. But we think it's important to keep history in mind. It's important to realize that the web and tools for building web-enabled application are still in their infancy. Of course, if you want us to do client/server, we here at Chalem Systems Group are happy to build the best possible applications for your project. In the end, no fashion or trend drives us. We continue to hold to the notion that there are different technologies suited to different purposes. It's a matter of picking the right tool for the job.

IEEE Computer History Timeline
Understanding Distributed Computing [Garnter Group's DataPro Report]

home  |  corporate overview  |  consulting services  |  articles  |  demos  |   success stories  |  products  |  glossary  |  search  |  contact  

© 2000 Chalem Systems Group
comments: feedback (at) chalem.com
site designed by: Chalem Systems Group