Client-server Computing Part III (September 1996)

PDM concludes the three part series introducing client-server computing. This month, the downside of C/S.

What's bad about C/S computing?

Now for the down side. Client-server has a number of negative things going for it too. It's hard to put them in order, but certainly near the top of any list, should be the amount of misleading information surrounding the whole idea. Client-server has been grossly oversold. And since it lacks a single, clear, definition, the door is wide-open to general misunderstanding. It's widely believed that client-server is a modern-day replacement for hierarchical computing solutions, for example. But as we discussed earlier, that's usually not true (nothing processes heaps of information better, and keeps data safer, than the good old mainframe).

misconception

Another popular misconception is that client-server is inexpensive (at least compared to the style of system development and implementation we've known in the past). Again, not true. Client-server may use inexpensive PCs as clients, but when everything else is factored in: education, more demanding application development, security, systems management and maintenance, they often turn out to be more expensive than a comparably-sized mainframe effort. But more than just weightier implementation costs, many of these same issues can threaten the very success of a client-server project. Heterogeneity also weighs heavily against client-server success. There's no guarantee that a database engine from vendor A will work with a client application built with a tool from vendor B if client and server are connected by a network supplied by vendor C -- if it's Tuesday, and the moon is full. Seriously though, when you stop and take everything into account -- operating systems, networking protocols, CPU designs -- it's amazing that it works at all. This is the price we pay for living in an non-open systems world.

not new

Client-server is still relatively new. Not only is there a lack widespread experience, but there's an acute shortage of comprehensive tools. This is perhaps, most evident in the areas of systems management and security. To date, there are few products able to simultaneously monitor and control both mainframe and LAN-based systems: bad news for client-server which often lies somewhere between the two. The state of client-server security is even less cheery. Apart from an almost complete absence of tools, client-server's "information for the people" philosophy flies directly in the face of historical security practices. This is certain to become the topic of much debate over the next few years.