The ServerSide has some amusing arguments back and forth about the advantages of RoR vs. J2EE. In my opinion, nothing beats a J2EE stack (or, more likely just the parts that you need) for complex and robust web applications. In recent history I wrote a "Sharepoint like" web application using a partial J2EE stack that simply runs forever without any intervention except for system backups. I also helped a customer architect another partial J2EE stack large scale web application based on JMS (i.e., "duct tape") that my customer recently told me has been running for 6 months without any fuss.
That said, I believe in the right tool for the right job. The only exception to this is that as developers we can not spend all of our time learning new stuff. Last year, based on the raves on Python from several Google engineers that I spoke with, I invested time working through Mark Pilgrim's fantastic book "Dive Into Python" - that was time well spent.
This year, I am investing real time getting into Ruby (the second edition of "Programming Ruby" is fantastically good). I have implemented a few simple web applications using Ruby on Rails - I like it so far. I have just been using WEBrick so far, but I plan on starting to use Apache + FastCGI + Ruby on Rails in the future. The only thing about RoR that is still missing for me is the the personal experience of writing a Apache + FastCGI + Ruby on Rails based web application and have it run for 6 months with no intervention or hassles. My gut feeling is that this combination will approach the solidness of J2EE.
In the mean time, I use Python and Ruby where those languages make sense, and stick with J2EE where that makes sense.