An often repeated axiom among the Internet cogniscenti is ``the network is the computer''. Personal computers have proved to be very useful tools to manipulate information, but the real power of information comes only when information is conveyed from one party to another. It is this realization that has created the intense interest in the Internet.
When the Internet was in it's infancy, standards were developed to allow computers to talk to each other in a common language, independent of their operating system, architecture, or vendor. More than anything else, it is this capability that has resulted in the tremendous impact of the Internet. This will be particularly true for medical information systems, since information will need to follow us around as we go from one place to another. Providing security for a static, physically isolated network is hard enough. Extending this to include roaming remote computers greatly complicates the situation, but must be accepted as a requirement for future systems.
The common language of network protocols that has been used to develop the Internet is commonly known as TCP/IP  (but should also include UDP). More than anything else, TCP/IP has demonstrated its scalability, adaptability, and suitability to a wide variety of applications. Riding on top of it is a set of protocols to support worldwide machine naming conventions (DNS), worldwide electronic mail transmission (SMTP), network management protocols (SNMP), file sharing (NFS), and an almost every imaginable service beyond these basic ones. TCP/IP has been identified most closel with UNIX machines, but there are now numerous implementations for Microsoft Windows, Apple Macintosh, and mainframes. It is notable that TCP/IP will be incorporated into Windows 95.
The type of system that I believe will ultimately be the dominant architecture for clinical information systems will consist of a distributed relational database system, spread around multiple institutions and/or locations. This database system will be accessible in a client/server fashion, using a variety of client workstations that are tailored to the particular user and/or facility. The client workstations may be fully capable desktop workstations, laptop computers, personal data assistants, or dumb terminals. The networking used to accomplish the client/server communication might be a combination of existing wiring or new wireless networks. We should expect to be able to buy server and client hardware and software from a variety of vendors, all of which can be expected to interoperate.