Bad Idea #1

Networking two machines that cannot be connected with ethernet for some reason (either missing hardware or security concerns) via a shared scsi bus.

I actually got this to work! I currently transfer files between my RS/6000 and my VAXstation via a shared scsi bus. Since I'm really not qualified to go mucking about with low-level scsi drivers etc., I put a scratch disk on the shared bus, and the two systems transfer files by accessing it in tar format. This is obviously just batch-mode file transfers at the moment, but it's sure to get worse.

Bad Idea #2

This one is much harder to realize, but more amusing if it could actually be done. The basic idea is to use a T1 CSU to fake voice channel packets, and use a T1 demux & an array of telephones as a 24 channel synthesizer. I have a CSU, and I had a stripped demux, but I got the impression that repopulating all the missing boards in the demux would be beyond my budget, so I gave it to a friend.

Bad Idea #3

The three-digit LED display on the front panel of my RS/6000 is under software control. It provides status & diagnostic information during system startup, but after that gets shut off and is available to user-space programs. I wrote two programs that make use of this display. The first is silly, the second is just excessive.

The first program displays morse code by flashing "000" for 300 usec for a dot and 900 usec for a dash. (And a handy shell script has the machine "reading" randomly selected poems "out loud" round the clock...)

The second displays various pieces of system status data, like CPU usage, number of logins, and time of day. Each digit can display multiple-digit values, of varying length. Each "click", which defaults to two seconds, but can be specified on the command line, the next digit of each value is displayed, or that LED goes blank to indicate the end of the number. The three digits can be of different lengths, and can be out of sync. It looks like complete gibberish if you don't know what it's doing!

Bad Idea #4

A few years ago I had no access to the net, only mail via UUCP from home. There was a site that was offering an html-pages-via-email service though, so I could email a list of URLs to a special address, and get the web pages back by email.

So... I modified lynx to generate email requests when following links, and had my mail system intercept the returning pages, strip the headers, and put the html in the correctly named file in a special directory. Latency was a bit rough (hours) but it was interactive!

(I told the guy who ran the email server, and he was greatly amused.)

Bad Idea #5

Doing non-linear transformations in postscript, which doesn't allow that. Think fresnel lens effect using one pixel wide clipping regions.

Stuuuuuupid. But kinda neat.

Bad Idea #6

I have a little interface box for real world IO (digital in for checking switches and buttons, digital out for controlling relays, analog in and out, etc) that speaks simple ASCII commands over a serial line to the host. So... useless for "realtime" stuff, because that serial line is a bit pokey, but much much easier to write code for that your typical bashing-io-registers-on-the-bus scenario.

I originally bought it for a project for a friend's art installation in the 2002 Biennial at the Whitney Museum in New York. After that was over and I got it back, I decided to use it to replace various desktop widgets with "front panel" style hardware.

The first thing I did was wire up some indicator lights. One shows new mail (only mail from sources on my "white list" though) and the other, big red (genuine US Navy "CAUTION: CLASSIFIED") light indicated whether or not the word "nuclear" appears on the world news page of cnn.com. (I originally had it showing the current terrorist threat level, but that got boring.)