Agile Web Development and the Lighter Languages

I have spent the past few months working with developing and experimenting with Java (JSP, Struts, Spring, Hibernate), Perl, PHP, ASP, Visual Basic and Ruby. Well, I will have to say there is something to be said about agile processes and open source dynamic languages. Recently I have had the opportunity to work with Java/JSP, Perl, and PHP. I am porting a PHP web application to JSP and intentionally sticking with a JSP Model 1 implementation. This is a bit of a digression in development techniques or a few steps backwards but I am testing out a theory. I can always refactor to an MVC design or other if necessary.

The theory that I am testing is: "It is possible to ignore elegance temporarily for attaining speed and practicality of getting a solution working using a 'quick and clean' process. (agility)" The old cliche in programming of doing something 'quick and dirty' to just get the job done is a concept I never really liked but have had to do occassionally in my career. The 'quick and dirty' approach almost always got me in trouble. I try to avoid doing this at all costs if possible. Based on my experiences with agile 'quick and clean' development, I feel that it is possible to attain software development speed practically in a sustained manner without sacrificing too much elegance. In other words, getting the cleanest job done in the best possible manner in the shortest period of time.

What I have found interesting in the days I've spent porting PHP to JSP is that it is relativley straight forward if you ignore your inner programming voice that says you should refactor the PHP to a more elegant Java/JSP design. In just a few days work, I have functioning webapps in JSP that rival the development speed of the looser PHP. I have to note that the PHP webapp is not using any PHP framework, design patterns (Five Common PHP Design Patterns), or other MVC pattern frameworks. This would of course complicate the design and make porting PHP to Java/JSP slower in my opinion but would require an experiment to prove it.

One other thing I must note is that having a really good IDE is critical to making this work. I am currently using Eclipse 3.1, Netbeans 5, and Zend Studio 5 in these 'quick and clean' experimental processes and this is definitely a must. I recently added Netbeans 5 to my pallete of java programming tools and it has yielded quite a productivity improvement for JSP and servlet development for me.

My next set of experiments is to get into Ruby on Rails (RoR) development which is a 'quick and clean' architecture. I have been talking about this all year however, have not made the time to start an experiment with RoR.