Raible Designs

AppFuse not for Rookies

Adrian Sutton hits the nail on the head:

...for seasoned veterans who understand (its) technologies well, AppFuse should be a big time saver. I can see a lot of beginners getting into very deep water with it though.

AppFuse can help you get started quickly, but it won't help you really learn how to use the different technologies. Its tutorials are designed to show you how to CRUD a simple object, but that's about it. For learning the different frameworks, that's what books are for. Technical authors put a lot of time and energy into writing. A good way to show your appreciation is to buy their books.

My Recommendations:

• Java Development with Ant - you'll be amazed at how much this book explains AppFuse's build.xml. This book inspired AppFuse (thanks Erik!).
• Pro JSP - AppFuse was the sample app for my chapters in this book. Explains many of the security features in AppFuse.
• Hibernate: A J2EE Developer's Guide or Hibernate in Action.
• Web Frameworks: Struts in Action (or update-to-date with Struts Live), Tapestry in Action, Core JSF.
• Spring Books.

So Struts is Dead, huh?

So most folks think Struts is Dead, huh? I've got news for you folks. Struts has been dead since 2002. For that matter, most Java frameworks have been dead for the last year. Most folks think Struts is dead because there won't be any new and whiz-bang features added in future releases. I'm willing to bet the other Java Web Frameworks won't add any whiz-bang features in the next few years either. Sure, WebWork might get a little XMLHttpRequest lovin' for client-side validation, and Tapestry might get pretty URLs - but both of these are features that these frameworks should've had in the first place. So what's the big deal? At least the Struts Developers have the guts to stand up and say "we're in maintenance mode". Shouldn't all framework developer's be able to say this? It's really quite an accomplishment.

IMO, the only developers that shouldn't be saying this are Tapestry and JSF folks. Both of these teams should be saying - "we're going to make our frameworks easier." If these frameworks are going to be the future of Java web development, there's some work to do. The JSF folks should be saying, "we're going to fix all the stuff that's broken". This "stuff" includes POST for everything, lack of bookmarkability, and bad validation messages (the good news is the validation messages should be fixed in JSF 1.2). I think Tapestry needs some simplification too. In my experience, most Tapestry pages (by design) require 4 files. One for the Java class, one of the HTML template, one for the page specification and one for the i18n keys. This should be easier. It'd like to see two required files (template and Java class) and the others are optional. Maybe annotations could eliminate the page specification? I think there's a lot to be learned from frameworks like Ruby on Rails: default everything, allow overriding.

I think that Shale will be good, but only if it learns from the other MVC frameworks available. If I were on the team, I'd take the good things from all the others: IoC, Interceptors, HTML templates, etc. I wouldn't stop there either - there needs to be good examples of how to integrate it with middle-tier frameworks like Spring, Hivemind and EJBs. Often developers will take examples as recipes - so the more detailed and simple, the better. The Struts developers have quite an opportunity to make something great, let's hope they don't just create another framework.

Short and sweet contracts

This has been a good month for Raible Designs. Not only did I manage to land a new gig, but I've also had a couple of 1-day contracts. The first one was an architectural review (at the beginning of the month), and yesterday I delivered a presentation on TDD and Spring for a client. These short-n-sweet gigs are a lot of fun. It's a good way to get out and meet members of the Java Community and see what tools they're developing with. The first client is using Struts, Spring and Hibernate. The second client wanted to use Struts, but after my talk, they're thinking about Spring MVC. They plan on using Spring to make EJBs easier, and they're using TopLink on the backend.

The only bad part about yesterday's experience is I developed a full-body ache as soon as I left the client's site. I managed to catch a cold from Abbie this weekend, and that turned into a cold+fever last night. I've had the fever ever since and didn't go into work today. It's really shitty timing for getting sick - we have a deadline on Thursday and we're heading to Oregon to see my parents this weekend. Last I checked, my fever was at 102°F. Hopefully, I'll wake up tomorrow and it'll be gone.

AppFuse at Dallas JUG

Apparently, the Dallas JUG will have the pleasure of listening to a presentation on AppFuse this Wednesday. I can't seem to find any information on their site, but someone forwarded me an e-mail that looks genuine. You know a project is getting popular when other people are giving presentations on it!

AppFuse Startup Video?

Dion wants to see an AppFuse Startup Video like Mike Clark's CruiseControl Action Movie. While this sounds like a good idea, I think I'd be shooting myself in the foot if I created it. Why? Because then more folks would start to use AppFuse, and hence, I'd have to answer a lot more e-mails on the mailing list. Being a top-ranked project on java.net doesn't help. You might think that there's a lot of issues with AppFuse, and that's why the mail traffic is so high, but that doesn't seem to be the case. Most questions seem to be along the lines of "Why did you do this?", "What do you think about adding X technology?" or "My Hibernate relationships don't work."

Few of the issues relate to AppFuse directly (i.e. build file and directory structure), but many of them relate to the technologies it depends on. Good HowTos should lead to a lot less Hibernate questions, and I hope to work on that before the next release. As far as the other questions, I need to add some links from the FAQ to the mail archives so I can quite repeating myself. I think a lot of the mail traffic is just an indication of a successful open-source project. In other words, when you get popular - you have little time to develop anymore. I probably spend 1-2 hours per day just answering AppFuse e-mails.

Another unfortunate side-effect of this is that there seems to be a lot of newbies. When AppFuse was first released in April 2003, it seemed that only experienced, smart developers used it. Maybe this was because there wasn't any documentation (besides Pro JSP and Java Development with Ant, which explains the entire build.xml file), so folks had to really understand the dependent technologies to use AppFuse. Now there's questions about the basics of different frameworks. In most cases, I'd like to respond to a link to the framework's documentation - but sometimes the documentation just isn't there. I guess that's why frameworks like Ruby on Rails succeed - all the dependencies are part of the framework. If I tried to do that in the Java Community, it'd be project suicide. I'd spend all day answering questions like, "Why aren't you using Hibernate?", "Why Not Spring/JSF/Struts, etc." Furthermore, I'm not as smart as the framework developers, so it'd simply never happen.

But I digress. What's in it for me if I create an AppFuse Startup Video? I can see what's in it for Mike - his video is about a project he doesn't support (AFAIK) and the video should lead to more book sales. I suppose I could try and hook users that AppFuse is explained in Spring Live, but that's not really the case. Maybe I should just do an Equinox Startup Video.

Comments are flakey?

I've gotten several e-mails from readers that they haven't been able to leave comments on this blog. I suspect it has something to do with twisty comments, but I just tried them in Firefox 1.0 on Windows XP and they seem to be working fine. In order to help me track down this problem, try posting a comment to this entry and see if it works. Make sure and copy it before you post it, just in case it doesn't work. Clicking on the "Permalink" or title for this post will take you to a backup comment form that seems to work regardless. Thanks for the help!

Update: From the various comments, there doesn't seem to be a common thread. It seems that half the folks that used Firefox + XP it worked for and the other half it didn't. I doubt it's a caching issue, but I have been seeing a lot of the following in my logs:

ERROR 2005-01-22 19:51:01,944 | RequestProcessor:processMapping | Invalid path /comment was requested

Roller recently changed from using /comment to /comments - so I don't know if it's spammers hitting the old URL or what. My guess is yes since it seems to happening in blocks. I'll try to add a preview button tonight so users will get taken to the permalink form for previewing. Either that or I could just change to always using the permalink form. It's up to you guys - let me know what you prefer.

RE: Hype: Ruby on Rails

Patrick thinks that Ruby on Rails is all hype.

Now maybe I'm just a bit biased since my framework isn't getting all the slashdotters oohing and awwing over it, but I think Ruby on Rails is way over hyped. The tutorial here is great and gave me a very good overview of what it does. At the end of the day, RoR is simply a RESTful CRUD framework.

I'd like to agree with Patrick, because that is my natural tendency when I see a project that everyone praises. But I know better. I think it's better not to speculate on the productivity or usefulness of a framework until you've used it to develop an app.

That's what I did with Spring, WebWork, Tapestry and JSF last year. Now I feel like I know "the truth" and whether one framework is better than the other. The truth is they all have strengths and they all have weaknesses. While one might work well for one project, it might not for the next. I think the best thing is that you don't setup yourself for framework lock-in. If you only know one web framework for Java, you should probably pick up a book and develop an app with another framework - just to see how things are done differently. Now that I've used all of the Big 5 in Java, I don't think it would be that hard to migrate an app from one framework to next.

So what am I trying to say? Don't bash on a framework until you've tried it. And I don't mean toying around with it on a Tuesday night, I mean using it for a real-world project. I'll probably diving in and doing a little Rails development later this year. Why? So I can see if all the hype is accurate.

Maven Console and setting properties

I've been using Maven at my new gig and the Maven Console in order to avoid its painfully slow startup times. Using a slow-ass PowerBook in conjunction with Maven makes my face turn read and my ears smoke sometimes, but I'm getting used to it, much to my dismay. While the console has made things tolerable, Maven itself keeps getting in the way. I hate how it *requires* me to run my tests everytime I build or deploy. So I've turned that off by creating a build.properties file with "maven.test.skip=true". The problem with the Maven Console is it doesn't let me turn tests back on, so I'm stuck with running "maven test -Dmaven.test.skip=false" when I want to run my tests.

There's two ways I can think of to solve this problem:

• If "maven idea:multiproject" allows me to setup my project so that Tomcat/Resin/whatever can point to my source directory and I don't have to deploy. I'm a web developer, and I typically have to run "maven deploy" to test simple UI changes. That's why I turn the tests off - because I want a 1-2 second turnaround to see my changes. BTW, it's too bad there's no "eclipse:multiproject" goal.
• Enhance the console so it's possible to set properties. For example, typing "-Dmaven.test.skip=false" would set the property so the next time I run "maven war", my tests would be run. That, or allow me to run "maven war -Dmaven.test.skip=false". Allowing this would also make it possible to run a single test from the command line, instead of all (the only current option).

JCIFS and jWebUnit

On my current project, we're using JCIFS to integrate our application authentication process with NT Domain logins. While I found it quite easy to integrate, the one issue I found is I couldn't replicate the login process in a jWebUnit test. I tried setting the WWW-Authentication header to NTLM, but couldn't get it to work. The solution I ended up using is to subclass the NtlmHttpFilter and disable authentication when the User-Agent is "httpunit".

Hopefully this is useful for others. If you've managed to get regular jWebUnit authentication working with NTLM, I'm all ears.

Using JasperReports with AppFuse and Spring

In Spring 1.1.3, support was added for using JasperReports with Spring MVC. Today, Gregory Beumer posted a nice overview of JasperReports. This inspired me to dig up Gilberto's post on How to integrate JasperReports with AppFuse. If you're looking for a reporting solution in your AppFuse-based application, and you're using Spring MVC ... enjoy! I plan on adding this to the wiki in the future, along with howtos for integrating JasperReports with Struts, WebWork, JSF and Tapestry. If you happen to know of tutorials for integrating JasperReports with these other frameworks, please let me know.