Interesting project. Unfortunately the code is kind of messy: no strict and no warnings, globals all over the place and no comments. It's quite an achievement to get the program running without using strict/warnings - must have been a nightmare to debug? It would most likely be better to, as you suggest, start the whole project from scratch.

Also there is an IMDB::Film module at CPAN which uses HTML parsing instead of Regular Expressions. Think that would make the data extracting more robust.
Anyway, is there anywhere we can see this DB in action (since I don't have access to apache at the moment)? And thanks++ for sharing.