Blogging Roller

That's the topic of the talk that I will present to the Triangle Java User's Group on February 16th. More information is available on the TriJUG site.

Weblogs are making their way into every aspect of online life. Everybody from gossipy teenagers, mainstream journalists, tech CEOs, and politicians to Java developers and even James Gosling, seems to have their own weblog.

In this talk you will learn why weblogs are an important part of the Java development community, how weblogs can be used in software development projects, and you'll learn about the XDoclet-driven architecture of Roller - the open source and Java-based software that drives the JavaLobby's JRoller.com site.

Inspired by Rick Hightower's attempts to get AppFuse running with MyEclipse, I've decided to give MyEclipse another try. I already had a license from work, but I've bought another license for my home use. I also upgraded both my workbox and my homebox to use Eclipse 3.0M6. I've already bumped up against the inflexible project structure issues that I hit before, but I'm having better luck than last time around and the JSP editor seems much faster than before. I'm let you know how it goes.

I'm working on extending the Struts HTML form tags so that they are aware of the Struts Validator. I'm doing this to meet these two requirements:

• All fields that are required fields will be indicated with a red star or some other indicator.
• All fields fields that have failed Stuts Validation will show up with a red box or some otner error indicator.

I've already made some progress on this and I have a good solid proof-of-concept, but I get this nagging feeling that I am reinventing the wheel here - even though googling has turned up nothing.

Is the extended-tags approach the best way to meet my two requirements above? Has this work already been done for Struts - that is - am I reinventing the wheel?

I don't even know what that means, but somehow it makes sense to me.

Jon Udell: This weekend, I was working with the Java API to Sleepycat's Berkeley DB XML, and it felt like one of those bad dreams in which you're slogging through molasses toward an ever-receding goal. I switched to Jython and quickly got the job done. And it was the same job (indexing and searching content) using the same engine (Berkeley DB XML).

I think Jon would find Groovy just as easy to use as Jython. One important advantage of Groovy over Jython, and JRuby by the way, is syntax. Joe average Java programmer is going to be able to pick up Groovy much more quickly than Jython.

Now it's certainly possible that McClanahan et (Sun ?-) al are
secretly planning for JSF to "replace Struts," i.e. be a complete,
mo', betta' MVC Solution For All Your Web Application Needs(tm), and
this is all part of some elaborate double game. Alternatively, it
seems more reasonable that, longer-term,

* JSF specializes in view-space and supporting Model1- and RAD-ish tooling
* Struts specializes in control foo, e.g. struts-chain, workflow.
* SFIL continues to be maintained
* sane web frameworks specialize and maintain integration interfaces
* developers choose what makes their little hearts go pitter-pat

More reasonable to me, anyway--am I missing something?

Somebody get this Thomas Roche guy a weblog.

I certainly don't. Groovy is fascinating, fun to play with, and, for a Java developer, much easier to grok than Jython. I've already written a couple of simple Groovy SQL scripts and now I'm trying to wrap my mind around Groovy Markup. What a cool new toy!

Later... Here is one small nit. Groovy could use better error reporting. When an error occurs, you don't always know what line in the script caused the problem.

Read all about it. I've heard Websphere's eBay customer story, but I would love to see the new JSF features in Websphere Studio. Unfortunately, I'm snowicebound.

It is difficult to tell what's really happening here, but CRN is reporting that IBM is putting the Websphere Studio toolset under the direction of the Rational group. Just the other week IBM announced that Eclipse, which is the foundation that the Websphere Studio teams build upon, will be managed by an separate corporation with it's own leader who is independent of IBM.

Juha Komulainen I find myself writing a Lisp-interpreter every couple of years.

Matt's been busy. He just released Struts-Menu 2.0 and AppFuse 1.0, two truly useful projects for twisted Struts-monkeys like me. In case you haven't heard, Struts-Menu is a JSP tag library for creating XML-configurable web application menus with support for role-based permissions and for Velocity-based menu rendering. I'm going to be getting started with Struts-Menu myself this week, so this is perfect timing. AppFuse on the other hand, is a skeleton project to help new Struts-monkeys to get started with Struts, XDoclet, Hibernate, JUnit, Cactus, and a host of other open source Java development tools and libraries.

Andy wasn't a big fan of political correctness until he realized that it was yet another way to stir up controversy, draw attention to himself, and bash The Apache Software Foundation.

Ovidiu Predescu has posted an interesting presentation on Cocoon's continuation-based control flow or "FlowScript". Coocon is a web application framework that is build around an XML pipelining technology. Since Java does not support continuations, FlowScript uses a version of the Rhino JavaScript engine that has been hacked to support continuations. Very interesting stuff.

via Christopher Kemper: part one of an excellent article by Ed Burnette about building applications on the Eclipse Rich Client Platform.

I think this is a very interesting topic and I've been exploring it further. Here is my current reading list.

Papers and talks on continuations and web programming:

• Modeling Web Interactions (Graunke et al., Apr 2003)
• Lisp in Web-based Applications (Graham, Apr 2001)
• Web Programming with Continuations (PDF) (Byrd, Nov 2002)

A couple of weblog posting on the topic:

• Ovidiu Predescu explains Cocoon's continuation control flow layer
• Blog discussion between <a href= "http://www.cincomsmalltalk.com/userblogs/avi/blogView?showComments=true&entry=3243274443"> Avi Bryant, Daniel von Fange and others on web continuations.

Java web application frameworks that support continuations:

• RIFE: a web application framework with continuation support
• Cocoon's Flowscript supports continuations by using a specially modified version of Rhino

Often in web applications, a user will be filling in an HTML form on one page and will realize the need to go to a second or even a third page to enter information needed by the first page.

For example, you are writing a weblog entry by typing text into an HTML form on the Edit Weblog Entry page. You type in about half of your entry and then you realize that you need to create a new category for your post. You click the Add Weblog Category button, arrive at the Add Weblog Category page, create your new Category, and click OK. At this point, you are returned to the Edit Weblog Entry page to find that your half-edited weblog entry is still there. You are ready to continue right where you left off.

How do you achieve this type of behavior in your web application? How do you allow users to continue working where they left off? Do you treat each case as a special case or do you have a general solution like a breadcrumb stack or a web continuation to this problem? It can get pretty messy. Do you cache half-completed forms in your HTTP session or in your database? Or, do you avoid the whole issue by carefully designing around it.

I didn't get a good response on the Weblogic forums for this question, so now, dear reader, I bothering you.

If you deploy a simple Servlet/JSP application with an error page on Weblogic 8.1 and your application throws an error, the Servlet Container will save and then restore the contents of your ServletContext. So, if you have any non-serializable ServletContext attributes, you will lose them. (Strange thing is, and this part is definitely a bug, if you deploy the very same application, but without an error page specified in web.xml, then the save and restore will not occur and all of your ServletContext attributes will be preserved.)

This doesn't seem quite right. I realize that the best practice is to only store serializable attributes in your context and in your session, but it is not a requirement of the Servlet specification. If it was a requirement, then setAttribute would take a Serializable rather than just a plain old Object.

Seems to me, non-serializable attributes should only break your application if you try to run in distributed mode. Also, the behavior of ServletContext attributes should be the same regardless of the presence or absence of an error page.

Is this save-and-restore-on-error behavior a bug in Weblogic?

Must context attributes be serializable?

Is there a way to turn off the save-and-restore-on-error behavior on Weblogic?

Via the Struts-User mailing list: FWA Software announced Visual Tags for Struts, a plugin for Dreamweaver MX that provides full support for the Struts custom JSP tags in Dreamweaver's code view, design view, and live data view.

How did I miss the OSCache 2.0 release?

Via Kevin Dangoor: the updated Eclipse 3.0 plan shows that we'll have to wait until June 2004 for a final Eclipse 3.0.

In my work, the most important new features are in the extended Java theme: "Improve support for Java-like source files" like JSP and "Support Java references outside Java code". Unfortunately, there is no indication that JSP support will extend beyond the basic code-completion, refactoring support, and syntax coloring for Java code embedded in JSP files. I'm still holding out some foolish hope that maybe, just maybe, IBM will move Websphere Studio's WYSIWYG HTML/JSP editor and Struts support into Eclipse while reserving the upcoming JavaServer Faces support for paying customers. Please? After all, by June 2004, Struts support will be the old hat to JSF's brand new bag.

I followed up my "not impressed with MyEclipse" post by posting some of the problems I encountered to the MyEclipse "private messages" support forum. That was yesterday. I didn't get a response, so today I posted each of my issues in the Bug Reports and Fixes forum (see below). The MyEclipse site is very slow and I haven't been able to muster the patience to figure out if there is a separate forum for paying customers (like me).

• JSP editor is very slow to start
• MyEclipse's inflexible project directory structure
• Can't remove MyEclipse from project
• JavaScript breaks the JSP syntax coloring
• Can't handle TLDs referenced by URI