At first glance, this looks completely reasonable especially if you consider that each corresponding environment declares a container-level variable whose value is one of the directory names (Development/Integration=dev, UAT/Test=test, etc.). Developers get to configure each environment and at deployment time the server simply knows how to behave based on the preset environment variable. Sounds OK, right?

NO! It isn’t OK. It is horrible and you should never do it. Packaging configuration in with code leaves the System Administrator powerless to administer the system. See the problem? No? Here are more reasons:

• Configuration changes require a redeployment of the application (possibly even repackaging)
• Developers may not (indeed, should not) be privy to certain config information (passwords, let’s say) for production environments
• Test environments of the same “level” can’t have differing configuration unless you jump through hoops like branching source code
• Turning a dev/test/stage box into a production one (or vice versa) with one small change is just scary
• A developer can change production and nobody will notice until the code gets to production!
• A developer can change production and nobody will notice until the code gets to production!
• A DEVELOPER CAN CHANGE PRODUCTION AND NOBODY WILL NOTICE UNTIL THE CODE GETS TO PRODUCTION!

Get it?

The way to manage configuration is to provide a nicely-commented prototype config file (that can’t/won’t be read) with your application and leave the runtime configuration in a file outside the deployed application path. There are many ways to make such files available to your program regardless of whether you’re using a container or not. You can:

• Add a conf directory to the runtime classpath
• Put the config files in a known directory such as common/classes on Tomcat
• Jar them up and name them according to environment, if you just absolutely must not let the admins touch them.

Any of these tactics work. As long as you aren’t marrying code and configuration, you can’t miss.

I think we can all agree that choice is good, but is the sheer number of non-trivial frameworks really unifying the java community, or are we breeding segments of the population who know only a subset of what’s out there?

When you’re an ASP guy, you pretty much have you work cut out for you. Same goes for the new kid on the block Ruby [on Rails]. But Java, in its maturity, has brought so many innovations to market that one could make an entire project out of evaluating frameworks for use in any given project!

The days of Struts & JDBC are pretty much over, kids. Do you use Spring at the web tier? How ’bout JSF or Tapestry or WebWork? Do you decorate with SiteMesh, or assemble with Tiles? Maybe you are into JSP includes or writing a lot of custom tags. Do you manage your middle tier with Spring, or opt for the latest appserver’s EJB3 implementation? It might seem like everyone is using Hibernate for persistence, but it aint the only kid on the block. TopLink Essentials is bundled with GlassFish and other EE5 offerings will be supporting JPA as well. This is just the tip of the iceberg.

How do you decide?

But no more. The EJB expert group, now composed of some of the very creators of modern persistence framework, has bequeathed upon us something wonderful. The lightweight POJO programming model we have come to love is apparent everywhere in the EJB 3.0 specifiction (JSR-220), which was unanimously accepted by the JCP earlier this year. EJB 3.0 is just one goodie in the larger Java EE 5 (JSR-244) bag – but this post is going to be long enough as it is.

When I returned from Las Vegas with my newfound interest in EJB, my first task at hand was to create a little test application that demonstrated the scenario that is most applicable to my work: EJB3 services remoted over RMI consumed with POJO web applications that already use Spring. This use case stems from the need that we have realized the need to segment a growing middle-tier application from the monolithic beast it is destined to become into smaller, more manageable services. The client programs (mostly web applications) already access the business interfaces using Spring.

I’m going to skip the gory details of setting up the server, creating a project and all the like. There are plenty of explanations of this elsewhere. Besides, this post is going to be long enough as it is.

First we need a business interface. How about something really complex:

[java]
public interface SimpleService {
int stringLength(String string);
}
[/java]

Nothing fancy there. Now let’s implement it with some logic:

[java]
public class SimpleServiceBean implements SimpleService {
public int stringLength(String string) {
return string == null ? 0 : string.length();
}
}
[/java]

Still totally straightforward. Now let’s turn our bean into an EJB. This is the really complicated part. Watch for the annotations:

[java]
@Stateless
@Remote(SimpleService.class)
public class SimpleServiceBean implements SimpleService {
public int stringLength(String string) {
return string == null ? 0 : string.length();
}
}
[/java]

Did you get all that? The @Stateless annotation tells us that this is a Stateless Session bean, and the @Remote annotation specifies the remote interface. Everything else is handled by the container. Although the footwork is managed by the container in a somewhat vendor-specific way, the annotations are standard Java EE, and so you’re not tied to an implementation. When you catch your breath, let’s move on.

Since Java EE 5 has embraced configuration by exception, we don’t even need any depoyment descriptors. These two files can be compiled, JAR’d up and deployed to a server of your choosing — I used JBoss 4.0.4.

In order to have a client, we need to provide, at a minimum, the remote interfaces and any other classes or exceptions it references. For testing, I just tossed the same JAR on the classpath since I know it contains everything I need. Additionally, we need the Spring JARs (this test used 2.0RC1) and the JBoss client JAR’s.

The only real programming we have to do here is to create a Spring container that wires up to the server’s JNDI directory. To do this, we’ll use Spring’s JndiObjectFactoryBean. This part actually confused me because try as I did, I could never accomplish this feat using the more aptly-named SimpleRemoteStatelessSessionProxyFactoryBean. The difference, apparently, is that since EJB3 places no specific requirements on the business, home or remote interfaces, there is less to do, and thus the client simply grabs the object and narrows it down as prescribed. Here is my Spring configuration (minus the XML header and DTD info):

[xml]

org.jnp.interfaces.NamingContextFactory org.jboss.naming:org.jnp.interfaces jnp://localhost:1099
[/xml]

The JNDI properties provided are JBoss specific, as is the provider URL. These are included with the JBoss client JAR in the form of a jndi.properties file. They are exactly the same as what would be used to construct a traditional InitialContext if this were a non-spring J2SE client.

JBoss mounts Remote interfaces (by default) using /remote, and Local interfaces using /local. As you can see here, we’ve chosen the remote. One could just as easily expose SimpleServiceBean as a Local EJB by using the @Local annotation and providing the same business interface as an arguement. The Spring container changes only in URL and jndiName values.

Once the bean factory is configured, you can load it and request the remote SimpleServiceBean proxy by name (simpleService), casting it to SimpleService as you would normally do with a Spring-managed object. It works flawlessly.

Gavin & crew demonstrated quickly but effectively the lovely synergy that has been brought to EJB3 and JSF by their newest creation, Seam. Seam can hardly be called a web framework — you wouldn’t recognize it as such what with its lack of, well, just about everything that you’ve come to expect from a web framework.

Seam throws away boilerplate code a-la Rails, but does so in a way that leverages a lot of the untapped power of EJB3 and JSF. Seam is all about Don’t Repeat Yourself (DRY). As such, we throw away form beans (a-la Spring MVC) and a whole lot of XML (in favor of annotations).

The most surprising theme that emerges from all of this is the rampant use of Stateful Session beans. You read that right – I said Stateful. Seam begs the question, “why on earth would I represent state in a database (slow is an understatement) or in HTTPSession (with no dirty checking) when I can use this lovely gift called Stateful beans?” He makes a damn good point. By representing web conversations as SFSB’s (which are POJO’s in EJB3, in case you’re wondering) you eliminate many of the headaches associated with web apps. Multiple tabs open in your browser? Piece of cake – each browser tab/session/window has a unique conversation (they’re stateful, remember?). Double submit? Ha! Don’t even try! Want to have a clean client redirect after performing some action, but still pass messages across? No problemo – you’ve got a lightweight stateful conversation behind you.

By logically grouping your flows into SFSB’s, you can leverage more sophisticated failover because the container dirty-checks each instance and replicates automatically. Try that with HTTPSession.

So, I will be doing a sandbox app soon with Seam, and I’ll post more as I know it. Till then, read all you can.

JBoss is mutating their MicroKernel (heavily JMX based) into what they call the MicroContainer.  The MicroContainer architecture could be described by reading the outline of Spring’s core container and subtituting “Spring” for “JBoss”.  Indeed, they are not only building a lightweight DI framework; they are building their entire app server on it.  Cool.

JBoss has always been good about letting the developer/integrater customize the footprint of the server through configuration.  It looks like this is going to be even cleaner and better in v5, since the same DI engine which builds the server innerworkings can also assemble application components — all the way up through the presentation tier using Seam/JSF.

• If you use “==” to compare two objects such as Strings, your code will not compile.
• The difference between a checked exception and and unchecked on is that you must handle a checked exception and you cannot handle an unchecked one.
• Omitting the access control modifier from a method reverts to the default access control which is the same as protected.
• To pass information from a Struts Action to a JSP, you set it in the response.
• In an MVC architecture, such as Struts, the JSP’s represent the model.

Guys, if you are going to interview for a programming job, brush up on the fundamentals! May I suggest a certification or two; perhaps a black belt?

Updated 5/32/2006 -CB

In their 2.0 release the Spring team has given the community its wish, and provided a JSP form tag library. It’s been added to the manual, but here’s a key term that might pique your interest: convention over configuration.

The Spring crew has taken lessons learned from the wildfire of support for Ruby on Rails and employed them in their J2EE kit. Bravo. Time (and other ‘me-too’ projects) will tell if this is really what J2EE needs to stay on return to the cutting edge.

Good

• Designed from the ground up using good OO principles – Rod & crew really have this part mastered
• Amazingly flexible – many concrete classes are provided for each bit of functionality so that you never get more than you need
• Transparently woven-in with Spring – less glue code

Bad

• Lack of rich HTML component library – just doesn’t cut it when you’re used to having forms and select boxes built for you. Apparently this is addressed in 2.0M3 which makes a new Form tag library available, but I didn’t see any reference to this in the docs.

More to come as I work with it.

I’ve found that the easiest way to transmit text from a foreign partner using Microsoft Word. Word stores its text as Unicode by default and is widely available and understood. It also displays the foreign text accurately so it can be used to visually compare web page results to original documentation. A plus if you don’t read Korean!

Since Java will not read files stored natively in Unicode, we need to get it into encoded ASCII format. This is done by using the native2ascii tool provided with the JVM. So how does this all fit into the software lifecycle?

Here’s the plan:

1. Obtain native text using Microsoft Word, or any text editor that can store a recognizeable native format. Make sure you know the encoding of the original text; native2ascii will need this. Since we carry the data in Word, the native encoding is UTF8. Other encodings are listed here.

2. Store a native version of the .properties file using a different extension. I use .utf8 for UTF-8, since that seems to keep things clear. In this copy of the file, native text can be pasted directly into the file. Using the Eclipse editor, this maintains the original visual glyphs exactly like Word.

3. Setup an Ant task that calls native2ascii on any applicable resource directories, using UTF8 as the source encoding. The same task can write the output files with the .properties extension. Each type of native source file you maintain will need its own command, which can be called from a single precompile task, like such:

4. Deploy your application!

The key here is to treat the encoded ASCII resources as build artifacts, and not version them. So for an application like ours, I store only resources_ko.utf8, not the resulting resources_ko.properties. This allows us to maintain only the native copy, and check the properties against what comes out on the screen.