I'm at TS-7393 along with a pretty packed room of, I estimate, 700+ attendees. All the sessions that I've gone to have been well attended with a serious but enthusiastic-looking group of folks it appears who ask smart questions in the Q & A's. The session is titled "Shale: The Next Struts?" with Sabreware's David Geary and Sun's Craig McClanahan. Craig is Strut's inventor and David, Struts template tag library inventor, which was the inspiration for creating Tiles, one of Struts most popular features. Shale is the new kid on the block in the web application frameworks space, with David and Craig being two of the three leading original contributors on Shale. A lot of people want to know: What's the deal with Struts and JavaServer Faces (JSF), and Struts and Shale? A lot of uncertainty surrounds the topic. It is impossible to predict where it is all headed

McClanahan presents the rationale for Shale. Shale is unique among web application frameworks in that it starts by assuming that JSF exists. Recognizing that with a little tweaking you can turn JSF into a completely functional web app framework, he decided to see if something as popular as Struts could be created. So he proposed Shale as the architecture for Struts 2.0 to Struts developers, which did not entirely sit well, given the focus on backwards compatibility and support for existing Struts users. They did accept Shale as a Struts sub-project, so it is officially part of the Struts environment. The idea is to grow a community around Shale that is already used to the idea of using web app frameworks in developing. He is quick to point out that Shale has zero direct relationship with the Struts 1.0 code. Shale is all new code based on the assumption that JSF exists; it skips the step of re- implementing the features that are already there, like validation and request processing.

Features in Shale include web flow, so developers can organize a dialogue that is longer than a request but shorter than a session. There is client-side and server-side validation but with JSF components. Shale is integrated with Spring to enable transparent use of Spring's bean factory to create managed beans. David Geary has been extracting Tiles out of the Struts core, so it is independent of the underlying framework and can be used by any framework more easily. Shale will take advantage of that. Primary subtrees is another way to perform reuse that will function through a specialized library called "Play". McClanahan pointed out that a lot of people dislike Struts because it likes JavaServer Pages (JSP). JSF likes JSP as well, but JSF has components as a Java API, and developers can plug in a different view handler to use something different. With Shale, they've created an environment that looks like Tapestry with a standard HTML file for the markup presentation. It has a test framework to make it easy to test the view controller and other implementations.

He is quick to point out that, although they could use Shale to create another JSF component library, that is not their goal. The components they are creating are only ones that interact directly with the framework's facilities. Anyone's JSF components will work in Shale - why reinvent the wheel? The five tags within Shale are: token tags to catch duplicate submits; a subview for extra value-add services for a dynamically included page as well as the main page; a commonsValidator tag used to integrate the commonsValidator rules; a validator script so developers can point to the Java script so they don't have to load it on every page. and a clay tag used for parameterized subtrees.

JSF as an architecture has extension points that Shale takes advantage of. The Spring integration is a custom variable resolver and looks at the first word in an expression like "customeraddress.city" -- if a customer does not exist will create the bean for you. By default, JSF will look in the managed beans definitions, but the extension checks for a bean factory as well. Similar facilities enable access to the Java Naming and Directory Interface (JNDI) context. The navigation management is encapsulated in a specialized dialog navigation handler. A specialized view handler enables developers to treat a Tile name as a view ID to reduce work

So where is all this headed? McClanahan presents three scenarios. First, it is possible that Struts developers will see that Shale is becoming popular and choose it as Struts 2.0. Second, it could become a separate Apache project - synergy exists with the Apache MyFaces community that is building both an implementation of JSF and some interesting components to go with it. All of the JSF-related things could be encapsulated into a single project. A third possible future is aware of how long it takes to build a Java standard in the Java Community Process (JCP) - almost two years to build JSF 1.0. Shale can be seen as an experiment to see if its facilities make sense in a future JSF release, perhaps in JSF 2.0 time frame. The future is unclear and ultimately in the hands of developers.

An Interview with Craig McClanahan

David Geary's Blog

Craig McClanahan's Blog