The promise of mashups is to leverage visual programming and the power of integration of different data sources and existing GUI-components. Gregor Hohpe from Google (http://www.eaipatterns.com/ramblings.html) did the presentation on mashups, and using an RSS-feed from his conference Google Calendar , he managed to create a web application where all his conferences were mapped out on Google Maps. He used Yahoo Pipes to transform the RSS feed adding lattitude and longitude for the conferences, and feeding the transformed data into Google Mashup Editor to present these in a slick user interface, including Google Maps. All this in less than an our, not writing more than 50 lines of code and HTML.

Grails is a Groovy version of Ruby on Rails, combining the power of Spring, Hibernate and Groovy. Grails will generate all the scaffolding needed to run a web-application, and almost no code is necessary, except the writing of POJOs. Because of that, creating an application from scratch is fast as lightning, and the generated code is straightforward and easy to change. When a POJO has been created, Grails will both create the GSPs (Groovy Server Pages) and the database. The POJOs are mapped to the database tables using GORM (Groovy Object Relational Mapping) which is nothing more than a wrapper on top of Hibernate. Of course there’s an extensive tag library for use in the GSPs, which does all the dynamic stuff you need, ajax included. Needless to say, the most popular ajax-frameworks are also easily available from within Grails.

And with that, I’ll go and pack my suitcase, as my journey home starts tomorrow at 0515.

It was an interesting day, where I managed to learn what REST is. the right approach to webservices and how it is possible to structure code by using a mix of POJOs, dependency injection and AOP.

Although the session on POJOs did not present any eye openers, it was nice to be reassured that the path we’re currently taking in the project I’m working on, is the right path. DRY SOCs where the keywords, which in plain english is Don’t Repeat Yourself and Separation of Concerns.How? The speaker used Spring to tie together the parts of his system, although Guice could have been another approach. We don’t use either, but we do design our system using interfaces and constructor injection so as to avoid tight coupling. We can easily test the parts mocking the dependencies using JMock, which the speaker also demonstrated.

He went on to show how boilerplate code on his service layer including security, transaction management, logging and auditing, is added using abstract oriented programming, which is also a part of the Spring framework. With AOP handling these concerns, the mock object test cases also becomes a lot simpler! He shows an example where AOP is used to add certain behavior to classes with certain annotations. He does this by making AOP make the annotated classes implement an interface, and adding the appropriate methods defined by this interface. Great stuff!

AOP seems to have matured into something ready for production code.

I went to two different sessions on REST. The first one took a very academic approach to the subject matter, and I didn’t leave that session with a very clear idea about the concept. Luckily, I went to another session on the same subject, and I was reminded that speakers are more important than the subject matter. The latter guy managed to explain the concept in clear understandable examples, and in just about 5 minutes I had understood more about the concept that the 1½ hours with the first guy. The very short version is hard to give, but REST is the architectural style in which the web has been built, and more specifically, the design principle behind HTTP 1.1. Unfortunately, the existing browsers are not using the protocol as it were supposed to be used. All URIs are supposed to represent a ressource, and the HTTP commands GET, POST, PUT and DELETE are the CRUD operations available upon them.

I enjoyed being confirmed in my belief in continuous integration as a basis for the production cycle. Sadly, we do not do a full build and deploy cycle on the project I’m currently working on, but that will soon come to change. In my never ending quest to find a better tool than CC, the speaker introduced Continuum, apart from being a part of the Maven2 project, it will run ant tasks as well. I’m currently trying to get it to run our build scripts.

next up a session on the more advanced aspects of the Groovy scripting language. Here, the notion of Builders were introduces. An extremely efficient, powerful and easy way to build tree-like structures, such as ant scripts, HTML, XML and finally Swing user interfaces, as they are essentially trees consisting of nested frames, panels and widgets.

To give you a taste of it, here’s an example taken from developerworks,

import groovy.xml.*
import java.io.*

class HTMLBuilderExample{
static void main(args) {
writer = new StringWriter()
builder = new MarkupBuilder(writer)

builder.html(){
head(){
title("Groov'n with Builders"){}
}
body(){
p("Welcome to Builders 101. As you can see " +
"this Groovlet is fairly simple.")
}
}
println writer.toString()
}
}

The resulting HTML will look like this:

<html>
<head>
<title>Groov’n with Builders</title>
</head>
<body>
<p>Welcome to Builders 101. As you can see this Groovlet is fairly simple.</p>
</body>
</html>
</pre>

OK, there’s tons of interesting stuff in Groovy. Groovy can be imbedded in other java code, either as compiled bytecode, ie classes, or as scripting. it even has nice meta functions to intercept method calls, much like AOP, but integrated in Groovy.
Another very nice feature, is the SOAP server module, which enables you to create a running soap-server in these few lines of code:

def client = new SOAPServer("path-to-wsdl-file")
def result = client.method1

My visit to the security 101 session turned out to be a waste of time, as I really didn’t to learn about SSL and basic cryptography all over again. I spent the session experimenting with Continuum and Axis2.
Next up were a session on SOA and SaaS with the charismatic David Moskovitch. David has thing with chocolates, and always brings a handful of goodies from the nearby Colorado Chocolate Factory, prizes for attentive audience. As it turned out, I won a price for honesty, as I admitted to my ignorance as to what Time Sharing is. Apart from the prize, it was an interesting session. I won’t get into the details, as they were plentyful, but the very short version is: SOA describes business processes, SaaS is revenue centered, no ownership, no risk. David stressed this repeatedly: The Business process is the major driver, not the technology (BPOA). And finally SOA Governance: who, what, when, where why and how, which in my opinion is a bit abstract. Perhaps this article could be helpful: http://www-306.ibm.com/software/solutions/soa/gov/lifecycle/

My final session before the Q&A was Axis2. A good presentation, but unfortunately, the speaker was at the same time non-native english and extremely fast speaking, so I missed almost all his points, apart from what he had on his slides. Ah well.

The Q&A session were many facetted, but the most interesting piece of information was from a geologist, who pointed out that the amount of energy released to the atmosphere by humans were only a fraction of the energy released by a single volcanic eruption, of which we have about ten every year. This said in the context of the fact that Colorado Software Summit has abandoned bottled water this year, for the reasons stated here.

As for java, I’ve been to four sessions today, neither of which were essentially boring or uninteresting.

GROOVY

We’ve been using a tiny amount of Groovy scripting, so nothing surprising came up in the Groovy Session, but I were reminded, that the syntactic sugar offered by Groovy, which is a superset of the java language, running on any proper JVM, is pretty powerful.

JBOSS

The COSMIC electronic patient record we sell and enhance at the WM-data Healthcare division is deployed on a JBoss-platform. For that reason alone it was somewhat an eye opener attending the session on “Management, Monitoring, and Deployment of Tomcat and JBoss”. As it turned out, JBoss is distributed with a couple of very interesting surveillance applications, of which none is available in the distrubution we are running. The most interesting one being the web-console. Memory snapshots can be created and monitored from the jboss web-console. Snapshots and monitoring of changes in jmx values can be created across all the mbeans of the jmx-console. Alerts can be created, mails sent, etc. Pretty powerful!

Also, JConsole could be interesting in a JBoss context. JConsole is a part of the JDK, and the run.bat in the JBoss bin directory must be edited to include -Dcom.sun.management.jmxremote. After JBoss has been started, the $JAVA_HOME/bin/jconsole.exe can be executed. See: http://wiki.jboss.org/wiki/Wiki.jsp?page=UseJDK5JConsole

A couple of other very interesting points: OutOfMemory can occur because of too many statics overflowing the permanent generation memory, not the heap! We’ve seen this problem, where the amount of free memory looks perfectly fine, and all of a sudden an OutOfMemoryError is thrown. More at: http://java.sun.com/docs/hotspot/gc1.4.2/faq.html. Further: Actively invoking the GC from the mbeans will free *all memory*, including the hotspot-compiled classes. This is rarely desirable!

Google Gears is in a very few words a javascript library enabling ajax applications to go offline. As of now, Google reader is a prominent example. An essential property of Google Gears is the embedded SQL-database, which allows the client side javascript to do CRUD-operations on a client side data source.

Spring vs. Guice

Finally, I attended a session on the difference between Spring and Guice, two prominent exponents of dependency injection. The question really boils down to this: do you prefer the strong typing of the java language in Guice, or the more flexible, compile free approach of XML-hell in Spring. I’ve been really interested in trying out Guice for a while, but I still can’t quite see the point: why not just inject the dependencies via different constructors, and have the test case inject mock versions? After today’s session, I’m still looking for that rock solid argument…

Anyway, the topic of healthcare was especially interesting from my perspective, as I’m an architect with the danish and Århus-based Healthcare division, and, according to Soyring, a part of the solution to the approaching healthcare problems.

In an american view, the “baby boomers”, a nick name given to the large birthrates of the post-WII years, are now looking for retirement, and this is popularly refered to as “the silver tsunami”, as this large body of people are now leaving the workforce, and will require large amounts of help from the healthcare services, doctors, hospitals etc. We’re facing similar problems in denmark, though we have a less glamoruous name for it, as we call the problem “the elder burden”. He actually mentioned the danish healthcare sector as a prime example of how to do healthcare. In denmark, we may have a bit more nuanced opinion on this. He also mentioned personal electronic medical records, a field in which we are working in at the moment.
IBM has put out a report on this: http://www-03.ibm.com/industries/healthcare/doc/content/landing/2955767105.html

As a solution to these problems, he somehow managed to link them to massively parallel processing environments, being either based on server stacks or community computing, two fields in which IBM plays a major role. I found the arguments a bit stretched, even though he’s most certainly right about the forth coming change in the architectures we build systems from. Here, he also tied a knot to the emergence of SOA and SOA governance, which he found particularly important, as it helps companies predict the value generated by IT.

One of the more interesting side stories, was that of out sourcing to India and China. Apparently, Indian companies are now actively outsourcing their outsourcing businesses, as they are in lack of qualified personnel(!) Also, the labor rates in these parts of the world are rapidly increasing, which in the end will make it less profitable for western companies to outsource their production.

As for skiing, my picture of the coffee mug revealed that we do have snow here in keystone, and plenty enough to actually go skiing, but alas, the season does not open before november 6.

Now, off to the keynote at 0830 with John Soyring from IBM.

When I last attended the Colorado Software Summit back in 1999, I was without computer and internet access for a week. Back then, the slides for the 50 different presentations were handed out in a very heavy ring binder, and sore shoulders were garanteed when carrying this load of papers around for the entire week.

Now, the presentations are available on a CD, in editable PDF, and the bag which is handed out this first day of the conference is designed to hold a laptop. High speed WIFI abound, both at the conference center and in the hotel.

Some things do remain the same. The americans are still polite, and the coffee mug, which is a part of the merchandise, is as distinctive as ever!

I shall be blogging during the week’s presentations, and hopefully I can present you with a glimpse of the many coloured splendour which is the Colorado Software Summit, held in the beautiful, but oxygen deprived Rocky Mountains.