Andrej Koelewijn

Here’s how i create an update of a vagrant box. It’s basically like creating a new box. I start a virtual machine based on an existing vagrant box, update it and shut it down. Next create a package and add the package:

vagrant package --base <name-in-virtualbox> --output <package-filename>
vagrant box add <box-name> <package-filename>

My mule configurations files often contain proxy settings for connectors that are used to communicate to the outside world. However during development and testing I can’t use the normal proxy, since i’m mocking the outside world.

Littleproxy is a java proxy that you can use to replace the real proxy with a testing proxy. This avoids you having to install a proxy on your development machine and on the continuous integration server.

The following is part of a mule config. It defines an http connector that is configured to talk through a proxy.

<http:connector name="http.connector"/>
<http:connector name="http.connector.proxy"
                proxyHostname="localhost" proxyPort="50103"/>
<flow name="http.gateway.and.proxy">
    <http:inbound-endpoint host="localhost" port="50102"
                           path="gateway.in"
                           exchange-pattern="request-response"
                           connector-ref="http.connector">
    </http:inbound-endpoint>
    <logger message="http.gateway - #1"/>
    <http:outbound-endpoint host="localhost" port="50100"
                            path="dummy.service"
                            connector-ref="http.connector.proxy">
    </http:outbound-endpoint>
</flow>

Include little proxy in your maven pom.xml:

<dependency>
    <groupId>org.littleshoot</groupId>
    <artifactId>littleproxy</artifactId>
    <version>0.4</version>
</dependency>

Here’s the unit test. First i start the proxy (you could also use @Before or @BeforeClass), and then i run a regular mule unit test:

public void testShouldReturnHelloWorldThroughGatewayAndProxy()
    throws Exception {
    String msg = "Hi!";
    HttpProxyServer proxyServer = new DefaultHttpProxyServer(50103);
    proxyServer.start();
    /*
    * send message
    */
    MuleClient client = new MuleClient(muleContext);
    MuleMessage result = client.send(
            "http://localhost:50102/gateway.in?connector=http.connector"
            , msg, null);
    String payload = result.getPayloadAsString();
    proxyServer.stop();
    Assert.assertTrue("Did not receive HelloWorld!: " + payload, "HelloWorld!".equals(payload));
}

I’ve been using When can I use… quite a bit recently. It enables you to quickly determine if a html5, css3 or svg feature is enabled in all the different browsers. Very useful.

A project manager is responsible for the succesful execution of a project. Success is usually defined as delivering the project on time, within budget, to the predefined requirements.

A product owner is also responsible for the succesful execution of a project. He manages a team, determines the requirements, the order in which the team delivers these requirements, and he determines when and how often these requirements are released.

The difference between a product owner and a project manager are the instruments used to manage a project.

Scrum is based on the fact that projects have large number of unknown aspects. Requirements may be unknown, productivity is unknown upfront, technical challenges will be unknown. Because of all these unpredicatable aspects, it’s impossible to draw up a plan upfront which predicts time, budget and delivered features.

A product owner manages the project by focussing on the requirements. What do the customers need? He makes sure that the most important aspects as done first. If you don’t know exactly what you need to ship to your users, and you don’t know exactly how long it will take to create it, you need prioritize.

“Do we still need a project manager if we’re doing scrum?”

Yes, scrum needs a project manager. It’s the product owner.

I’m working on a small javascript library to use svg drawings for presentations. This would allow you to display a single svg drawing similar to normal presentations on slideshare or speakerdeck. You can see the current state here: SVG Presenter test.

I wanted to see if you can use webfonts in an svg drawing in Inkscape. Found out it is doable. You can download fonts using the css rule @font-face. SVG supports css @font-face.

Here’s an example with four different fonts, all webfonts downloaded from google web fonts.

Inkscape doesn’t have official support for webfonts, but you can add the required css to the xml source. Here is an example, usually the defs element already exists and you need to add the style element:

  <defs id="defs2987">
    <style
       type="text/css"
       id="style7">@font-face {
  font-family: 'Shadows Into Light';
  font-style: normal;
  font-weight: normal;
  src: local('Shadows Into Light'), local('ShadowsIntoLight'), url('http://themes.googleusercontent.com/static/fonts/shadowsintolight/v3/clhLqOv7MXn459PTh0gXYHW1xglZCgocDnD_teV2lMU.woff') format('woff');
}
</style>
  </defs>

You can import fonts that you don’ have installed locally, but Inkscape will not allow you to select the font. You can however edit the svg source code to modify the style of some text to use the webfont. In the drawing above the first three fonts i used i have installed on my laptop, the fourth isn’t on my computer. Inkscape will show an error icon, but you can ignore it.

I create most of my presentations using Inkscape (here’s an example: Presentation: Introduction to Scrum). This is quite a bit of work, as i need to export all the separate images, and import them into LibreOffice Impress.

I’ve created a small javascript script that can be used to directly use the svg image as a presentation. It allows you to navigate through all the layers of a drawing. Here’s a demo: Inkscape-presenter demo. Use the arrow keys, your mouse, or a logitech remote presenter to move forward and backward (tested with firefox and chromium).

The code is on github: inkscape-presenter / slide-animation.js.

When I tried to use the following example by David Dossot mule-agent-based-sync-http-request-handling my jms endpoint times out. In the log i can see that the response message is created, but the outbound endpoint never receives it.

After some googling i found the following blog post by Claude Mamo: ReplyTo in a Mule flow. Adding the org.mule.routing.requestreply.ReplyToPropertyRequestReplyReplier processor does indeed solve the problem.

The weird thing is, i can’t find much documentation on the need for this. I searched the Mule User Guide, but it isn’t mentioned there. Also googling it, just returns a lot of links to source code.

Is this the correct way to implement a request response jms endpoint, or is this a workaround for a bug?

Here is a quick example how you can access you can access properties loaded using spring using the Mule expression language.

    <!-- configure property placeholder -->
    <context:property-placeholder location="classpath:application.properties"
                                  system-properties-mode="OVERRIDE"
                                  ignore-resource-not-found="false"
                                  ignore-unresolvable="false"/>
    <!-- configure bean to read the same properties -->
    <spring:beans>
        <spring:bean id="appProps"
                     class="org.springframework.beans.factory.config.PropertiesFactoryBean">
            <spring:property name="singleton" value="true"/>
            <spring:property name="location" value="classpath:application.properties"/>
            <spring:property name="properties">
                <spring:props>
                    <spring:prop key="key4">value4</spring:prop>
                </spring:props>
            </spring:property>
        </spring:bean>
    </spring:beans>

    <flow name="spring-property">
        <vm:inbound-endpoint exchange-pattern="request-response" path="springproperty.in">
            <logger message="Message received #1" level="INFO"/>
            <!-- set a session scoped property on the message -->
            <message-properties-transformer scope="session">
                <add-message-property key="dynamickey" value="key4"/>
            </message-properties-transformer>
        </vm:inbound-endpoint>
        <!-- use the property placeholder, set once, not per message -->
        <logger message="key1 = ${key1}" level="INFO"/>
        <!-- use the appProps bean -->
        <logger message="key2 = #[groovy:appProps.key2]" level="INFO"/>
        <logger message="key3 = #[groovy:appProps['key3']]" level="INFO"/>
        <!-- access property value based on message property -->
        <logger message="key4 = #[groovy:appProps[message.getSessionProperty('dynamickey')]]" level="INFO"/>
        <append-string-transformer message="abc"/>
    </flow>

Yet another big IT project in dutch government failed. A 25 million euro project was cancelled. Lots of press and discussions:

• Ict-project waterschappen loopt uit op debacle
• Tweede Kamer gaat onderzoek doen naar ict-problemen
• Parlementair onderzoek naar ict-problemen bij overheid

Why did they spend 25 million before declaring the project failed? Shouldn’t they be able to detect earlier that what is being build isn’t going to help the users? Or that it’s going to cost a lot more than estimated?

Most government projects are using Prince 2, a project management method created by a British government institute. It looks like this isn’t working, for whatever reason. Maybe they’re not using it right, maybe it isn’t the right tool to create succesful products.

Mary Beijleveld recently pointed me towards another method created by a British governement institute: System Error. This approach advises to use Agile methods for project management.

Looks like a good time to read it.

As far as i’m concerned the biggest problem is not understanding the difference between product development and product manufacturing, and the different types of management you need for these two.

My LAC-2011 presentation Scrum under Architecture is now available on speakerdeck.

Since it contains mainly images, some notes might be useful:

• Scrum isn’t something by software developers, for software developers
• Scrum is a tool enabling management of product development by product managers
• There are two types of projects: those implementing known solutions, and those looking for a solution for a problem
• You can manage the first type of projects in the traditional way: requirements, time, resources are known and minimal change should be managed
• The second type cannot be managed this way: you don’t know what the best solution will look like
• Agile project management is about getting the best solution within available time and resources
• Many companies struggle to scale agile: how to you go beyond one team?
• Architecture enables you to scale agile
• Agile Architecture is not just about doing short iterative architecture
• Agile Architecture should enable you to work with small independent customers focussed teams
• This needs to be enabled by technical architecture but also by organisational architecture