You may want to take a look at Catalyst, CGI::Application, or Bricolage (listed in order of my familiarity and personal recommendation). If you're willing to look at other languages, Rails (for Ruby) and Seaside (for Smalltalk) are both also very good solutions to the problem of generating Yet Another Web Platform.

As for driving your webapp using configuration files (which is, essentially, what you're talking about), you might want to look at the MVC and REST architectures (not mutually exclusive). There are a number of good articles on Perlmonks, but wikipedia, perl.com, onlamp.com, and dozens of other sites are great resources, too.

Please remember that you're working in a problemspace that has been very well explored for over a decade by some of the brightest minds of our time. This doesn't mean you cannot find a new solution. It just means that it's very unlikely that you will do so. Because of the low probability, you have a higher burden of proof. That includes having learned the current methods so that you know what they are first. If you don't know the current state, how can you improve on it?

As for it working - I don't doubt that. But, please bear in mind my signature. And, the second criterion isn't just for the code - it's also for the people using the application. How easy would it be for them to make changes assuming they don't want to know XML?