Navigating the Web-development Framework jungle

The landscape of application frameworks, especially in the web tier has really matured and grown where they are many quality choices available for the developer. My team and I are in the process of starting some major development work on a bunch of new J2EE applications and I’m sticking my head up and trying to look at what’s available in terms of frameworks in the web-tier.

Most of the applications that we are going to write or re-write are going to consume existing services that encapsulate business logic, integration, persistence, etc. In my company, I’ve standardized on the use of ‘services’ and I use the word loosely, to enable loosely coupled application and reuse of business logic. Most of our services are stateless, built using EJB’s (no entity beans, CMP or otherwise) and these services isolate the consumer from all of the transactional, integration type of stuff. All of these services are exposed directly via the EJB interface. Some services are also enabled as Web Services with things like JAX-RPC, Axis and WebLogic’s EJB to Web Services Ant tasks. All of the Web Services interfaces or methods are custom coded methods that return a String of XML as it’s output. From the Java side, the object graph is serialized to XML and the client can consume the XML being returned using the Schema provided, if they choose to do so. If not, the XML can be deserialized back into Java using XStream or XMLBeans. Non-Java clients can also consume these services and parse the returned XML using their own mechanisms. These services have enabled loose coupling and clean integration by .NET clients to our Java/J2EE backend.

Back to the web tier – My first Java application at this company was written in mid 1999. This was a transaction trading system deployed under WebLogic 4.5, written as a bunch of servlets communicating to a stateful EJB in the backend. Since then, our application development has gotten a little more complicated with things like Broadvision and other personalization and portal technologies. Most of our web development today is done using Struts with a good mix of JSTL, interacting with our backend services. As I look to design and architect our next applications, I am starting to look around and wonder if Struts is still the best way to go or should I really explore some other frameworks. To answer my question, I’ve spent a little time of the last few weeks digging into my choices are here are some of my choices.

  1. Struts – This is a known entity and my team has extensive experience in building Struts applications and probably a safe choice, but Struts has its issues. (That’s another blog entry ;))
  2. Struts + Tiles – I am not a fan of Tiles and it’s probably because I haven’t given Tiles enough time but the integration with Struts is nice and the ability to have templating, multi-channels support, layout control and reuse is pretty powerful.
  3. Struts + SiteMesh – I’ve only briefly read about SiteMesh but everything I read makes it look attractive enough where I should take a critical look at it and give it a deeper look. SiteMesh is another web-page layout and decoration framework that integrates with Struts.
  4. TapestryTapestry is another framework I’ve heard great things about but haven’t given it any time. Just bought Tapestry in Action, written by Howard Lewis Ship, the creator of Tapestry and I am hoping it will help me get more information. The issue of learning curve also comes into play with anything new and I’m trying to balance the benefits the new framework provides and the learning curve required to get proficient in the framework and its subtleties.
  5. WebWork – I knew I would get yelled at if I didn’t mention WebWork in this list. Rossi, are you happy now? :-) The same issue of learning curve comes into play here.
  6. SpringSpring includes a configurable MVC web framework that’s very similar to Struts and I love the idea of lightweight containers and IoC but going with Spring, like other framework includes the standard learning curve. I’ve personally played around with Spring a lot and feel that the learning curve isn’t as steep when compared to other frameworks, but there is time required to come up to speed.
  7. JavaServer Faces – I attended 2 sessions by David Geary earlier this year at the local No-fluff-just-stuff Java symposium and I didn’t walk away with a warm-n-fuzzy feeling about JSF. It’s an option, but I think I am probably not going to explore this option.

I know I am leaving out tons of other choices like Turbine, Velocity, WebLogic Workshop and Page Flow among other frameworks.

I’ve narrowed down my choices and have gotten down to Struts/SiteMesh, Tapestry and Spring as my final choices. I am in the process of creating some proof-of-concept applications using each of the frameworks to get a better understanding of the strengths and weakness of each framework as well as create some blueprint apps for my team to use as a starting point for their development effort.

I would love to hear from anyone that has any thoughts and opinions on my choices above and any jewels that you can share before we jump in and start our development effort.