The Lab Guys need your help determining the best thin-client solution

I have a favor to ask: I need your help settling an argument that's raging inside my head. I usually don't open my head up for inspection because its contents are not a pretty sight. But this argument centers on thin-client technology, a subject that might be bouncing around inside your head too, so I thought I'd get your opinion.

Two types of thin-client terminals can connect to a Windows-based Terminal Server or WinFrame/MetaFrame system: instant-on terminals and downloadable devices. Instant-on terminals load their operating system (OS) from ROM and work as soon as you turn them on. The Wyse Winterm product line provides the best examples of instant-on terminals. Downloadable devices are terminals that retrieve their OS and parameters from a server; network computers (NCs) fit into this category.

Some of the voices inside my head argue that downloadable devices are a bad idea because they require the support of a specific infrastructure. You must set up a Trivial File Transfer Protocol (TFTP) or NFS server to handle download requests. If you're a conscientious systems administrator, you'll also want to install a backup server to ensure that the devices can function if the primary server goes down. To run the backup server, you'll want to implement file replication and IP address failover products. Suddenly, your thin clients require fat-server solutions.

Personally, I'd rather not have to work so hard to set up a few terminals. Remember, I haven't even described the steps for implementing Terminal Server or WinFrame/MetaFrame here. I just described the process of setting up an environment that lets you initialize your terminals so that they can potentially access applications residing on Terminal Server or WinFrame/MetaFrame systems.

The supporters of downloadable terminals (the other voices inside my head) argue that an infrastructure lets you have centralized control of terminal configuration and the software your terminals run. These voices declare, "There is no better way to roll out a new version of terminal software or introduce a terminal configuration change than to download the necessary information from a centralized server." In their opinion, downloadable terminals' central administration is worth allocating a couple servers to handle the download requests. The voices ask, "Don't centralized management and download fit the model of thin-client computing? If you're going to centralize your applications, the next logical step is to centralize your terminals' configuration and management."

When I mention the simplicity of instant-on technology, the voices of the downloadable-device supporters rise to full volume and point out the price of instant-on terminals: After you install one of these devices, you can't update the terminal's software. For a while, these voices swayed me because the inability to update terminal software is a major drawback. After all, software changes are fast and furious in the NT industry. Then I learned that most instant-on terminal vendors now support flash downloads from a central server so that you can update software and configuration information, and my doubts returned.

The voices have reached a deadlock (although they haven't gotten any quieter), so I need your help. I hope your passionate and articulate opinions about terminal technologies will sway me one way or the other. Do you have a preference? Do you think implementing a download infrastructure is worth the price of admission to thin-client computing, or do you prefer the instant gratification that instant-on terminals offer? Write me at john@winntmag.com. Maybe the problem really just boils down to the fact that I have a thick head when it comes to downloadable thin-client terminals.