If Microsoft has its way, the term network computer (NC) will get lost in the noise surrounding Terminal Server. Microsoft would like to co-opt the whole NC phenomenon by pitching Terminal Server as a superior server platform for NCs. In Redmond-speak, the NC is synonymous with the Windows terminal.
Of course, this latter assertion is a bald-faced lie. The NC, as core proponents Sun Microsystems, Oracle, and IBM define it, is a very different animal, one that follows a more traditional client/server computing model. To argue that Windows terminals are a kind of NC is like saying that an automobile and a commercial airliner are both forms of transportation. The statement may be generally accurate, but it tells you little about the objects or their respective applications.
The differences become obvious when you examine both Windows terminals and NCs from a network architect's point of view. Specifically, when you review each solution in terms of its system requirements and application support, you see how different they really are.
Counting the Deployment Costs
In terms of integrating the solution with an existing network infrastructure, NCs win hands down. Their download-and-execute model is likely to be far less bandwidth hungry than the continuous connection model of Terminal Server.
With a Windows terminal, every keystroke, mouse click, and screen update goes over the wire. This process ties the solution's interactive performance to the network's performance and necessitates that you maintain a constant level of available bandwidth to ensure responsiveness. In contrast, NCs need to access the server only when retrieving application code or posting updated datamuch the way networked Windows users share server-based Windows applications. Even though bandwidth requirements spike from time to time, they still follow traditional download-and-execute patterns familiar to any veteran IS planner.
Of course, with some of the computational overhead off-loaded to the client, an NC-based solution will require less server horsepower than an equivalent Terminal Server-based configuration. In fact, with the new breed of self-booting NCs from IBM and Oracle, the NC server need be nothing more than a traditional Web server. Imagine trying to justify the cost of a four-way Symmetrical Multiprocessing (SMP) gigabox to the CFO when the UNIX camp is pitching NCs tied to a Linux/Apache backend.
Reality Check: Application Support
Microsoft's strongest argument against the NC (and ostensibly, for Terminal Server) is the issue of legacy application support. A platform that can take most off-the-shelf Windows applications and deploy them in an ultra-thin fashion across a series of low-cost terminals has appeal. If Windows and the Win32 API are the corporate development targets of choice, then Terminal Server is a no-brainer.
The NC's Achilles' heel is its reliance on Java as an application interface. As Java goes, so goes the NC. This message plays much better in heterogeneous shops where cross-platform support is critical (Java's strength), and plays less well in firms where Windows and Win32 applications set the standard.
Politics Will Rule the Day
One player that has seemingly been lost in the shuffle is Citrix. The developer of Terminal Server's MultiWin core holds the trump cards that may give it considerable leverage as the Terminal Server-NC story unfolds.
First, Microsoft's contract with Citrix supposedly has a clause that prohibits Microsoft from developing Terminal Server client technology for non-Windows platforms. The worlds of UNIX, the Macintosh, and other third-party platforms (at least at the client level) belong to Citrix and its ICA protocol.
As non-Windows systems, most NCs will also fall under this mysterious clause. Because many big-name NC manufacturers are reportedly pledging Terminal Server support, Citrix may find itself in the enviable position of holding the keys to the Windows terminal kingdom for these vendors. No matter which side eventually wins the thin-client war, customers will reap the rewards of intense price and performance competition and technical innovation.