It is something I have never truly understood but the basic fact is that things will tend, in the course of time, to go tits-up: Call is the second law of thermodynamics, call it bit-rot, blame it on Bill Gates; whatever - Systems that once worked well can for no readily apparent reason stop working.

Of course, there is always a reason: It just isn't immediately obvious. The problem is that when you start working on sufficiently large and complex systems - such as something like a bunch of CGI scripts running off a database, so not terribly large - silly little things in the code can cause what was once a nice system to stop being so happy.

Example: If your code is run as a cron-job, for example, and does something like file-archiving or logging and it hits a snag you might not notice until, three weeks and several hundred/thousand/hundred-thousand log entries later when you really need something urgently. Then you notice that those last three weeks your script has been appending just so much garbage to your nicely comma-delimited file.

While I don't really mean what I have written in the title (except for publicly available CGI, where paranoia isn't the problem, it is a lack thereof) for business applications you have to write and maintain in a defensive manner.

Elgon