Cloud computing
Not really a Java subject but it attract people. We tried to define the cloud computing and we fixed that there is 3 offers :

• IAAS (Infrastructure as a Service): this is typical Amazon services products, S3 for storage and EC2 for virtualized servers. Amazon offer very basic service now with a very powerful management tool in Eclipse plug-in (see the demo). There is also Elastic Grid proposing to develop and deploy easily on the Amazon infrastructure, GoGrid an Amazon concurrent. I believe that the recent IBM / Sun merge will create new offerings.
• SAAS (Software as a Service): we find a lot of solutions (often based on the previous offer), for example Amazon SimpleDB, Amazon SQS, Google Apps, Microsoft Azure Services CloudMQ, ZumoDrive … and I could continue like a long time…
• PAAS (Platform as a Service): is hosting the application on a common and scalable platform, it is typically Google AppEngine, it is possible to deploy all yours web application if you know Python. Microsoft probably has a deal in Azure (I should have a look) and Sun has just launched Zembly.

A lot of discussion on what about offline, security, and where is java in the cloud. For me offline mode is really important in a world of increasingly nomad people. Cloud is primarily storage space allowing me to share my data between my devices, then an area of services, and finally deployment platform of my apps.
Finished managing a backup that is never done, losed time finding  a way to share data and finally used USB key, now my data are in the cloud and synchronized on all my devices. I have set up Zumodrive in my company and it’s very cool, the documents are shared even outside the company and I don’t care about backup.
Security is the most bigger difficulty for acceptance in the company, I hear the same remarks when talking about the payment on Internet ten years ago. All these services are secure and there is no zero risk.
Java have his place in the cloud on client and server side. The multi platform aspect facilitates the developments on the client (eg: ZumoDrive client is in Java), I want to see more and more Java APIs « cloud-ready » facilitating the integration of service in code. Similarly on the server side I look forward to Google AppEngine in Java.
Finally the advantage of cloud computing is primarily economic, smalls company are the first customers and have found lower cost and flexible capabilities.

DDD (Domain Driven Design)
I had little success at the last barcamp with this subject, this time it was proposed by others much more stronger than me and have made relevant arguments on the benefits of the concept. One of the important point raised we used talking too much technical and framework implementation than focus more on the reality, something that we tend to forget wanting to put our new framework in our code. I talked about Qi4j, that is not pure DDD implementation, but is for me the best way to modelling reality. I want make a demo of my medical record implemented with Qi4j to really prove that this approach is relevant.
Of course I’m convinced that a DDD refactoring of an existing code is difficult. DDD is a best practice and a new way of development vision. Have to follow for sure …

Thank you again to the organizers. It is always a good opportunity to exchange. And i hope Google will open their doors as often as possible;)

1. Dataset that contain only one set of data (characteristics of a customer for example)
2. Dataset multi-lines (customers list)

I used Birt Designer to configure the various elements.

1. Dataset with one record

You have to add to the dataset a column which will represents the visible status of the bound component. Edit the dataset properties and create a computed columns by clicking on new

Then you have to set:

• The name of the column (isVisible for example)
• Its type: boolean

Then you have to build the expression allowing to test the other columns:

• Click on fx
• Build the expression by testing the wished columns

2. Dataset with multi-lines

First you have to create a computed column with the following attributes:

• name : length
• Data type : integer
• Aggregation : COUNT

We create then an « isVisible » column which test the value of the column lenght with the following expression :

row ["length"]!=0

3. Visible status of the component

In the tab properties and in the visibility column

• Select the « Hide element  » checkbox
• Set the expression field with the value: ! Row [" isVisible "]

If the return value is true the component is not display in the report.

1. synchronization (between clients),
2. response time,
3. offline,
4. reliable data.

The MOM is used mainly for the first two aspects and least for the next two. Most of time it is a dedicated server where messages are transmitted. If the server go down or the network go down the application show transfer error. This is typically the case in Life Cycle Data Service: the messaging mode lets you synchronize and improve response time, but if the network go down, messages sent by the client are lost. In addition, the application I currently developing, uses Eclipse RCP and embed a Web application that communicates with messaging but with a local queue, make sure of message delivering. The solution is based on IBM MQSeries and MQ Everyplace and implementation is complex but it responds to the 4 aspects. 

Is this the end of HTTP ? This is the disadvantage of asynchronous architecture as the value of HTTP is to pass throw firewalls, using another protocol limit visibility of a web application. To resolve this problem MOM have to use HTTP and this is already done.

Don’t forget that this model implies that user is not dependent of response, or most of request are in this case. Hence the need for a local database allowing offline mode. The exchange client / server is then reduced to synchronize database to database. Is this becomes too heavy in terms of deployment and management ? not really because databases are already on board (Java DB via Java Plugin, SQLite via Google Gears). Yes, but what about the database loading ? Indeed it’s necessary to load the database once and it could be long but it depends of the expected need.

This architecture represented in the diagram cons is surely ideal for certain types of Web applications that require autonomy. It’s still difficult to implement because it requires a complicated setup and tools are not intuitive. Eventually this should become completely masked to developers,  using methods send / receive without worrying about the rest. Messaging is a proven technology we just need tool or framework to realize such architecture in a simple way.

Guice.createInjector().getInstance(aClass);

aClass is not a service but an object using injected services, then you can’t create a ViewPart including service, you must use an intermediate object : a controller.
For example you have a ViewPart (MissionViewPart), a controller (MissionViewController) and a service (MissionService). The declaration look like this :

– MissionService :

@ImplementedBy(MissionServiceImpl.class)
public interface MissionService {
...
}

and create a singleton MissionServiceImpl that implements MissionService.
– MissionServiceImpl :

@Singleton
public class MissionServiceImpl implements MissionService {
    public MissionServiceImpl() {
	...
	}
...
}

– MissionViewController (where the service is injected) :

public class MissionViewController {
    MissionService service;

    @Inject
    public MissionViewController(MissionService service) {
        this.service = service;
    }
...
}

– MissionViewPart

 public class MissionView extends ViewPart {
    public static final String ID = "ResUrgencesSMUR.missionView";
    MissionViewController controller;

    public MissionView() {
        controller = Controllers.getInstance(MissionViewController.class);
    }
...
}

I use a Controllers object because the Guice User’s Guide say :

Your code should deal directly with the Injector as little as possible. Instead,
you want to bootstrap your application by injecting one root object. The container
can further inject dependencies into the root object’s dependencies, and so on
recursively. In the end, your application should ideally have one class (if that
many) which knows about the Injector, and every other class should expect to
have dependencies injected.

– Controllers :

public class Controllers {
    static Injector injector = Guice.createInjector();
    static Map controllers = new HashMap();
    public static  T getInstance(Class type) {
        T controller = (T)controllers.get(type);
        if (controller == null) {
            controller = injector.getInstance(type);
            controllers.put(type, controller);
        }
        return controller;
    }
}

So now when the controller is called by the view, to create UI and bind it and do whatever you want, the service is automatically injected. To use differents service’s implementation you must use the Module concept to define, at the beginning, which implementation have to be uses by each service interface, the @implementedBy key is just the default implementation.
It’s a kind of injection use but if the application is complex and big, and also use a lot of services and differents technologies, i think it’s best to use enterprise SOA frameworks (Improve Foundations, Spring).

<w:composite background="#82cff1"> <a:layout x:class="gridLayout" f:numColumns="2" /> <w:label background="#82cff1" text="Raison :"> <a:layoutData x:class="gridData" f:horizontalAlignment="GridData.END" /> </w:label> <w:text w:style="SWT.BORDER" x:id="callReason" font="Arial,10,SWT.BOLD"> <a:layoutData x:class="gridData" f:widthHint="320" f:horizontalAlignment="GridData.BEGINNING" /> </w:text> <w:label background="#82cff1" text="Type :"> <a:layoutData x:class="gridData" f:horizontalAlignment="GridData.END" /> </w:label> <v:comboViewer x:id="missionType" v:style="SWT.READ_ONLY" font="Arial,10,SWT.BOLD"> <a:combo.layoutData x:class="gridData" f:widthHint="120" f:horizontalAlignment="GridData.BEGINNING" /> </v:comboViewer> </w:composite>

– TableViewer, Group, ToolBar

<w:group text="Victimes" background="#82cff1" x:id="victimGroup" enabled="false"> <a:layout x:class="gridLayout" f:numColumns="2" f:marginHeight="0" f:marginWidth="0" marginTop="5" f:verticalSpacing="4" /> <v:tableViewer x:id="victims" v:style="SWT.BORDER | SWT.H_SCROLL | SWT.V_SCROLL | SWT.MULTI | SWT.FULL_SELECTION | SWT.HIDE_SELECTION" table.linesVisible="true" table.headerVisible="true"> <a:table.layoutData x:class="gridData" f:horizontalAlignment="GridData.FILL" f:grabExcessHorizontalSpace="true" f:heightHint="95" /> <v:tableViewerColumn column.text="Nom" x:id="victim.name" /> <v:tableViewerColumn column.text="Sexe/Age" x:id="victim.sexeAge" v:column.style="SWT.CENTER" /> <v:tableViewerColumn column.text="Motif" x:id="victim.motif" v:column.style="SWT.CENTER" /> <v:tableViewerColumn column.text="Transport" x:id="victim.transport" /> </v:tableViewer> <w:toolBar background="#82cff1" w:style="SWT.FLAT | SWT.VERTICAL"> <w:toolItem toolTipText="Supprimer" x:id="victim.remove" /> <w:toolItem toolTipText="Ajouter" x:id="victim.add" /> <w:toolItem toolTipText="Exporter" x:id="victim.export" /> </w:toolBar> </w:group>

– Combo, ListViewer, CDateTime (Nebula Widget)

<a:layout x:class="fillLayout" /> <w:composite background="#82cff1"> <a:layout x:class="gridLayout" f:marginHeight="0" f:marginWidth="0" /> <w:combo x:id="filter" w:style="SWT.READ_ONLY"> <a:layoutData x:class="gridData" f:horizontalAlignment="GridData.CENTER" f:grabExcessHorizontalSpace="true" f:widthHint="140" /> </w:combo> <v:listViewer v:style="SWT.MULTI | SWT.BORDER | SWT.V_SCROLL" x:id="resources"> <a:list.layoutData x:class="gridData" f:horizontalAlignment="GridData.CENTER" f:verticalAlignment="GridData.FILL" f:grabExcessVerticalSpace="true" f:widthHint="140" /> </v:listViewer> </w:composite> <w:cDateTime background="#82cff1" x:id="takeAt" w:style="CDT.TIME_SHORT | CDT.BORDER | CDT.SIMPLE | CDT.CLOCK_12_HOUR"> </w:cDateTime>

Now i make the dream that VE could be able to generate EXSWT in WYSIWYG.
Tom also implements JFace Data Binding in EXSWT (XBL) and it’s easy to bind your model directly in your XML.

static Browser browser;
public static void main(String[] args) {
    Display display = new Display();
    Shell shell = new Shell(display);
    shell.setLayout(new GridLayout(2, true));
    shell.setText("Use Mozilla's Design Mode");
    try {
        browser = new Browser(shell, SWT.MOZILLA);
    } catch (SWTError e) {
        System.out.println("Could not instantiate Browser: " + e.getMessage());
        return;
    }
    browser.setLayoutData(new GridData(GridData.FILL, GridData.FILL, true, true, 2, 1));
    Button searchButton = new Button(shell, SWT.PUSH);
    searchButton.setText("Search");
    searchButton.addListener(SWT.Selection, new Listener() {
        public void handleEvent(Event event) {
            search();
        }
    });
    browser.setUrl("http://www.google.com");
    shell.setSize(400, 400);
    shell.open();
    while (!shell.isDisposed()) {
        if (!display.readAndDispatch())
             display.sleep();
    }
    display.dispose();
}

public static boolean search() {
    nsIWebBrowser webBrowser = (nsIWebBrowser) browser.getWebBrowser();
    if (webBrowser == null) {
        System.out.println("Could not get the nsIWebBrowser from the Browser widget");
        return false;
    }
    nsIDOMWindow window = webBrowser.getContentDOMWindow();
    nsIDOMDocument document = window.getDocument();
    nsIDOMXPathEvaluator xpath = (nsIDOMXPathEvaluator) document.queryInterface(nsIDOMXPathEvaluator.NS_IDOMXPATHEVALUATOR_IID);
    nsIDOMXPathNSResolver res = xpath.createNSResolver(document);
    nsISupports obj = xpath.evaluate("//input[@name='q']/@title",
                                                 document,
                                                 res,
                                                 nsIDOMXPathResult.STRING_TYPE,
                                                 null);
    nsIDOMXPathResult result = (nsIDOMXPathResult) obj.queryInterface(nsIDOMXPathResult.NS_IDOMXPATHRESULT_IID);
    System.out.println(result.getStringValue());
    return true;
}

4. via a socket : by executing an XMLHTTPRequest with execute() method which call a socket and get the content.

private String getDataFromBrowser() {
	String javascript = "try {";
	javascript += "xhr_object = new ActiveXObject(\"Microsoft.XMLHTTP\"); ";
	javascript += "xhr_object.open(\"POST\", \"http://localhost:9091\", false);";
	javascript += "xhr_object.setRequestHeader(\"Content-type\", \"application/x-www-form-urlencoded\");";
	javascript += "xhr_param = generateXML();"; // appel d'une fonction retournant un contenu
	javascript += "xhr_object.send(xhr_param);";
	javascript += "} catch (ex) { alert(\"Failed to save document\"); }";

	String xml = null;
	XMLPicker e = new XMLPicker();
	Thread t = new Thread(e);
	t.start();

	// wait threa run
	try {
		Thread.sleep(10);
	} catch (InterruptedException eee) {
	}
	// ask browser
	boolean ok = browser.execute(javascript);

	int wait = 0;
	while (xml == null && wait < 10)
		try {
			Thread.sleep(10);
			wait++;
			xml = e.getXML();
		} catch (InterruptedException eeee) {
		}
	System.out.println(xml);
	return xml;
}

private class XMLPicker implements Runnable {
	private String xml;
	public String getXML() {
		return xml;
	}

	public void run() {
		ServerSocket server = null;
		try {
			server = new ServerSocket(9091);
			server.setSoTimeout(5000);
			System.out.println("Waiting for response");
			Socket socket = server.accept();
			// stream treatment, i can give the HTTPServer class if you want
			xml = HttpServer.getXMLString(socket, "UTF-8"); 
			socket.close();
		} catch (IOException ee) {
			System.err.println(ee);
		} finally {
			if (server != null) {
				try {
					server.close();
				} catch (Exception e) {
				// Ignore.
				}
			}
		}
	}
}

5. running an HTTP server inside the Eclipse JVM : with Tomcat embedded for example or with WebObjects (Application.primeApplication()). The objects in the same JVM can be shared but be carreful of classloaders.
6. via terracota : a JVM cluster offer an easy way to synchronize objects between JVM.