So far, the use of Linux on the IA-1 has been formulated in the following way:
* Add USB Ethernet device
* Add Bootable CF media
* Run X and Netscape to access internet
There have been tweaks and adjustments here and there for allowing some limited stuff, but so far, there hasn't been a purpose defined or a set of variables that users (application engineers!) can change to suit their needs more effectively. I think doing so would lead to a project series that would maintain focus and allow a larger number of users to participate.
By creating an "IA-1 Cookbook" for example, we can gen together recipes for a variety of circumstances. The most popular seems to be "The X windowing machine with a web browser connecting via ethernet." However, I can see other recipes worth formulating as well. Among these might be:
1. A NAT firewall with text based menus
a. Configured for Dialup/SLIP/PPP etc sharing to ethernet
b. Configured for broadband internet sharing to ethernet
c. Configured as "b" with added dial-in server with PPTP for your home LAN.
2. A FAX server! Send/Receive faxes to be printed on a SMB or other network printer.
3. A stand alone and dedicated X terminal and nothing but. (Easy on, Fool-proof usage)
4. An MP3 player device
a. For the home
b. For the car!
I am sure there are other possibilities that haven't popped into my head yet, but the fact is, the machine is built with very limiting hardware and all projects so far seem to be in directions that push the machine too close or even over the limited capabilities of the machine resulting in crashes and instability.
I love what's going on! Rasmus is still my God. I just think there is room for some organizing and improved focus for what we can do with this device and appropriate uses for it.
I can see a lot of potential for the internet router/NAT firewall project. The expenses to the implementor might be a maximum of $250 if you include the cost of hardware that might not be present on your PC for CF module work. That's not too shabby if you think about it. And the specs of the device itself fits nicely for such a purpose since anything graphical immediately uses a lot of memory and saps performance from the machine.
At any rate, by focusing on specific purposes rather than generic ones, the various resources the device contains can be efficiently dedicated to it with fewer concerns about having the already cramped internal 16MB CF filled with generic tools for user manipulation which will probably not be used once its use as an appliance has been established.
Here's another project that I think has merit, especially when it comes to loading up new images and such. Let's create a small image for use in the CF slot with the power to update the internal 16MB CF. Instead of using DOS tools and such, shouldn't it be possible to create a bootable Linux that will allow use to dump image files from a network or USB data source onto the internal 16MB CF memory? One such stable Linux load formulated to that end would almost completely ween the implementor from using anything DOS at all. I'm sure that would be very appealing for the more hard-core Linux users and I believe it would serve to better streamline the actual development process as well. Why mess with LARGE CF images when a mere (inexpensive!) 8MB module would do in order to facilitate the update to the internal 16MB loading. Imagine the simplicity at that point? Boot up on the "load module", select/configure the source (SMB, NFS, FTP, HTTP, USB drive, whatever), then tell it to install. Done! One image never to be used for anything else and at a cost to the user of what? (What's an 8MB module going for now? Pretty cheap I'll bet!)
The added efficiency, increased use of "Linux-only" methods and ease of use could do amazing things to the sale of IA-1s and make Tiger Direct very rich. ;) It could also serve as a VERY inexpensive introduction to Linux for many people out there. This situation holds a lot of good potential so long as focus is maintained and clear-cut goals are defined. Rooting ourselves in the terrific work already done by Rasmus and all is a great start at a uniform set of building blocks that could comprise any number of IA-1 projects.