What was it?OK, maybe the best way to start this would be a little background information. Basically I was relocating my primary network operations to a different area of the house, but due to physical restraints and wiring issues, I had to leave the server where it was. This was a problem, because at the time my primary interface with the server was an RS-232 connected VT420 serial terminal. So my solution was to build a serial terminal directly into one of the walls in the new area of the house. I would run the RS-232 signals down the unused pairs in the Cat 5 via some breakout modules I was working on, so both the serial connection and primary data link would be run through the same line, preventing me from having to run a second cable down the run.
Now...it would probably be wishful thinking of me to believe you have read this far and didn't already ask yourself "Why not just use SSH..."; of course, any normal person would. I had naturally considered that at the time, but was prevented from going this route initially because of two primary concerns.
First, I had no constant-on alternative to the serial terminal. Or in other words, if there was some issue that needed immediate attention, I would have to startup one of my laptops, connect to the network, authenticate with the server, and by that time the crisis (I never got far enough to postulate what would necessitate such urgency) could have been too far out of control. This was in stark contrast to the serial terminal, which would greet me with a flashing prompt no more than 10 seconds after I flicked the switch. One of the little joys of serial networking that is rarely experienced today is the fact that even if you turn the terminal off, your session never stops on the server. So you could not power the terminal on for a week, but it would still be sitting exactly where you left off when it came back online.
The second issue, as I recall, was that I had an inexplicable fear that either the network itself would go down, or any remote access daemons I was running on the server would fail. As I saw it, the serial terminal was the most reliable line of contact with the server. Even if somehow the server lost all TCP/IP connectivity, I would still have a direct connection from which I could diagnose and fix the issue. To be fair, this was my first Unix server after coming from a lifetime of Windows administration, so perhaps I simply expected there to be multiple unexplained failures as I had come accustomed to under Microsoft's offering.
How did it work?The idea was really very simple. Basically, I would take an old 386 laptop, put the screen on top of the motherboard using stand offs, and put the whole thing into a metal panel that would have been flush with the wall. The real hook though was the keyboard, I had devised a way for the keyboard to fold up to cover the screen when the terminal was not in use. The easiest way to describe it would be to think of the trays on the back of airliner seats. When you wanted to use the terminal, you pulled the keyboard down towards you, and turned on the main power.
The laptop's HDD was replaced with a CF-IDE adapter and 32 MB CF card. This contained an MS-DOS 6.22 installation and the appropriate software and autoexec.bat file to launch the terminal emulator immediately. I wanted to use Linux on the board, but frankly there was no way to get the kernel and initrd to boot as quickly as DOS; Linux took over a minute to boot, while DOS was up in roughly 8 seconds. RAM was also a concern, as the machine only had 8 MB in the first place, and even a frugal kernel and run-time environment would be a tight fit.
Aesthetically, I was going for a Cold War Era "You can't fight in the War Room" sort of look. I had envisioned this as looking totally at home in a missile silo, so to that end I used a red lighted safety switch that you might find on the dashboard of a B-52 (or alternately, a '91 Civic). The front panel was to be cut from a single sheet of stainless steel, as was the fold-down keyboard tray.
So what happened?Thankfully, I never got as far as actually blowing the hole in the wall. Sometimes I look at the spot the terminal was going to go into and just breathe deeply with an inner peace. Not that I wouldn't have liked the look of it exactly, but as time went on and I got more accustomed to how Linux would work for me on the server side, I realized there was absolutely no reason to be able to have constant access to the server. When I got to the point that I trusted SSH enough that I decided I could rely on it in a (still unquantified) emergency, my skills in scripting and running the server had gotten good enough that the machine basically ran itself.
I recently ran into the box that held the half-completed terminal and felt a slight pang of regret as I pushed it back into the cavernous interior of what I consider my technology graveyard. While I still think the look of a built-in terminal would be very cool for a server area, the idea of it being a serial terminal is perhaps not as appealing anymore.
On the other hand...it did give me an idea. I wonder if I couldn't resurrect the old VT420 and put it to some useful work with more modern technology...