Ever since the "netbook craze" (beginning with the Eee last year), affordable ultraportable laptops have descended to a price level much lower, in fact, than were the Pocket PCs I was obsessed with (the tombstone of my Windows CE obsession reads: "2000-2005. What was I thinking?").
I've written technology blog entries before, discussing the fallacy of the "do it all" gadget, but as I continue to look for a possible Christmas technology purchase, I reflect again on how technology could make certain tasks easier, more enjoyable, and more efficient. In a more recent post I had my sights on the Gigabyte m912. However, early reviews have dissuaded me: its 1GB of RAM isn't enough to drive Vista comfortably; the version of Vista it comes with lacks tablet functionality; and Intel has just shipped a dual-core version of the same processor it carries, which itself is pretty weak.
When Gigabyte or another company releases a similar machine with the dual-core Atom processor, 2GB of RAM, Vista Home Premium, and perhaps a bigger battery, I'll take it.
But let's get a little speculative now.
My philosophy of computing is that where human activities have different priorities, they need to be embodied in different devices. Thus, for example, it is incorrect to try and integrate a cell phone with a full computer. Mobile phones need to be light, convenient, simple, inexpensive, and unobtrusive. Computers need to be powerful, capacious, and have excellent human interfaces. A no-compromise solution is a multiple-device solution, and there is no way around that.
However, a multi-device solution does not necessarily mean having multiple centers of data storage. No matter how good of software you use, it will be a struggle to keep all of the software synchronized on all of them, all of the time.
People already get around this by having large USB keys that they carry around with them. The more responsible among those regularly copy the contents of said keys onto the computers.
My thought is that two bodies of technology should be separated: on the one hand, the inputs and outputs of a computer, and on the other, the data storage, processor, GPU, etc. that make up the computer itself (including the batter for portable devices). The former can be any size or shape; the latter should be as portable as possible.
The UMPC device category is ostensibly an attempt at this sort of arrangement. However, it has a terrible flaw: such devices are encumbered by UI hardware that is woefully inadequate. By trying to turn card-deck sized computers into self-sufficient PCs, companies have been forced to compromise on absolutely essential features. I am sorry: serious work cannot be accomplished on an OQO. The combination of a touchscreen, battery, thumb-keyboard, speaker, and every IO port possible ramps the price up on such machines so as to make a cost-benefit analysis very dreary.
In my opinion, a portable computer should be a combination of CPU, GPU, storage, wireless technology, and other integrated devices, but have only a single, high-bandwith port. It would have no display, no UI, and no battery. A brick, in other words. That brick would also hold the OS of the user's choice, loaded with plug-and-play drivers for various inputs and displaytechnologies. Such a device could be terrifically portable and inexpensive, relative to today's UMPCs and ultraportable laptops. This would also improve options for more capable processors and discrete graphics.
A battery could be integrated, perhaps, so that the brick could maintain periodic 3G connections to a online backup server. But otherwise, the brick would be a card-deck size machine with a single port.
Then you could have every kind of computer under the sun, each an empty shell with a single port. Desktops, laptops, ultraportables, and tablet PCs. Note that I do not include "UMPC"--I believe that this category of computer needs to die. The smallest computer usable for real mobile productivity is "netbook" sized, with at least a 9" screen.
Not every brick would be powerful enough to drive every kind of display satisfactorily; but every "shell" workstation would accept every brick. You could buy a "gaming PC" brick or an affordable "workstation" brick. Eiter brick would work in a desktop or a laptop "shell", but the "gaming PC" brick might kill the batteries in the laptop shell very quickly relative to the cheap brick.
I know this isn't an original idea.