Update. Integrate Aptana Jaxer into Default Apache Instance

I finally found the solution (and, more importantly, an understanding of the issue) I was looking for to fully integration the default Apache instance on my development machine with Railo and Tomcat. I've previously written about setting up Railo and Tomcat on Aptana Cloud, however, for my local machine, I setup Railo with Resin to connect to Apache.

After giving it some thought, I decided that I needed to bite the bullet and mirror a production setup for my development environment (my production servers are using Tomcat, not Resin). Getting Tomcat, Railo and Apache to play together really wasn't that hard. There are tons of good blog posts on it (links below) from Sean Corfield, Matt Woodware and others. All very helpful.

A significant part of the integration process was to use AJP (Apache JServ Protocol) to proxy HTTP requests from Apache to Tomcat. The standard Apache/Tomcat integration results in Apache, essentially, turning over all HTTP requests to Tomcat.

The problem, for me, was that I wasn't getting how Apache could proxy requests to Tomcat. A combination of posts from Sean Corfield's and Matt Woodward's blogs had the explanations and code sample I needed.

I won't restate what's already been well-written but I encourage anyone looking to get Apache, Tomcat and Railo going on their dev (or production) boxes to check out those posts. An important bit to note is that AJP is part of (or installed with) Tomcat by default. It's freaking been there the whole time :)!

Following is a sample change on Matt's blog, I made the following change to my VirtualHost(s) so that I proxied my requests to Tomcat rather than letting Apache turn over the entire HTTP request.

Note: The SyntaxHighlighter utility alters my Apache code and some tag attributes will be incorrect (e.g., the opening VirtualHost element, which should read <VirtualHost *:80> will be altered)

<VirtualHost *:80>
    ServerName dev.mysite.local
    ServerAdmin somepin@gmail.com
    DirectoryIndex index.cfm index.html index.htm
    DocumentRoot /Users/myusername/Sites/mysite
    # here comes the proxy details!!!
    ProxyRequests Off
    <Proxy *>
        Order deny,allow
        Allow from all
    </Proxy>
    ProxyPass / ajp://localhost:8009/
    ProxyPassReverse / ajp://localhost:8009/
</VirtualHost>

In addition to the Proxy instructions, I removed the line of code that mounted mod_jk to my virtual host (JkMount /*.cfm coldfusion). The result of this has been that all three servers play nicely (Apache, Tomcat, and Railo) AND all my SES URLs work perfectly without UrlRewriteFilter in place. By proxying the requests, I think I can even use Apache Rewrite Rules in my VirtualHost element. I'll try that out this weekend. Yeah! I love it when servers play nicely together.

It's moments like these that I realize why my wife and friends make fun me all the time. I'm such a dork...but I love it :)!

This post isn't meant to be a rant against .NET; rather, it's an expression of general frustration with the .NET Web Services I have been forced (project requirements) to work with the past two months.

Update
I received some good comments (below) that pointed the issues I've dealt with are more design than .NET. That's actually great to hear because it's been incredibly frustrating. As a CFML and PHP developer (learning Ruby and Java in greater depth this year), I would look at these services I've been dealing with and think "this could be so much simpler...and easier to work with". Thanks for the feedback!

The Background
I'm working on a Flex application to be used in a medical research project. The back end technology driving the application are .NET Web Services. Until last Friday, I was working with two Web Services: one for authorization and another for dealing with the main user type (a researcher).

The authentication service is pretty generic. Collect and user name and password and pass to the service, which returns a boolean indicating whether or not the authentication was successful or not.

The researcher service is a bit more complex. It enables the application to get and set a wide range of data related to the researcher's study. It's been a bit of a pain for two reasons: first, many of the methods are, in my opinion, quite convoluted. One way I think they're convoluted is in the returned data type. For many of the methods, what comes back is dependent upon what you request. Second, dealing with conditional data types being returned (or expected in a call) from .NET Web Services can, depending on your setup, result in encoding and decoding errors. These are known bugs in Flex 3.x (it's supposedly fixed in 4). Still a pain, even if it is Adobe's 'fault', because I can't help but think that the Web Service methods could be less complex, yet equally robust.

Going Over The Edge
The third service, a messaging service, was deployed last Friday to our test environment. This is the Web Service that just killed me. The entire purpose of this Web Service is to take a string and send this string, via email, to a pre-determined recipient (it's not even a dynamically generated email address to which we send the note).

Following are the methods in this service:
GetEmailEnumeration
GetTemplateParametersByTemplateId
Send

Seriously?! .NET needs three methods to send an email? Seems like overkill to me.

Here's the process I have been instructed to take:
First, call GetEmailEnumeration() to get the EmailId class, which contains all the Email templates. This class returns a few names of templates. Fine. But if I have to call this method and still set which one I want to use, why can't I just pass a string parameter to the send method to get the freaking template?

Second, call GetTemplateParametersByTemplateId() to get an array of Parameter objects which are the email template parameters for sending out the email. The resulting array of Parameter objects just contains name/value pairs I need. Again, why not just let me set these as parameters of the send method without making a call to the server to get, essentially, static data?

Third, call the send method, passing in the email id and the parameters object (array of parameters). The user content is set as a pre-defined parameter named 'Verbatim'.

In summation, I need to make two server calls to get information that I know anyway (we know what parameters we are going to receive and then regurgetate the name/value pairs we receive when calling the send method. This just seems ridiculous to me, not to mention inefficient. Why force the server to handle two extra calls when, for the life of me, I cannot see why we need them.

Less Code, Greater Efficiency: The ColdFusion Way
Here's a basic CFML version of a similar Web Service:

  
    <cfargument name="params" type="struct" required-"true" />
    
      
      
        
           
        
      
      #Trim(params.message)#
    
  

Of course, I have only stubbed this out with the most basic of functionality but, really, how much functionality is needed to send a "Contact Us" email? That's all we're doing on the client side. Sending a freaking generic contact us note to a dedicated email address.

I just don't see the benefits, on any level, to the .NET Web Service's approach. Why make something that is so bloody common and easy to do, so difficult and cumbersome.

While it's frustrating to have to deal with .NET Web Services on any level, I must say that this experience (and others like it) deepen my appreciation for ColdFusion, as both a language and server-side platform.