Insight may run in several modes (standalone, rmi server, rmi client), so it should support being configured in several ways :

• From a java program, like a test case
• From a Cocoon configuration file
• From an external configuration file
• Not configured, using defaults

Here is the basic anatomy of a Insight configuration file :

<insight
    xmlns:insight="http://shaman.paris5.fr/insight-conf"
    rmi-server="localhost"
    run-style="standalone"
>

  <foo-service
      role="fr.paris5.shaman.insight.service.unexposed.Foo"
      class="fr.paris5.shaman.insight.logic.unexposed.FooImpl"
  />

  <hello-service
      role="fr.paris5.shaman.insight.service.exposed.HelloService"
      class="fr.paris5.shaman.insight.logic.exposed.HelloServiceImpl"
      exposed-as="Hello"
  >
    <hello-message>Hello, world</hello-message>
  </hello-service>

</insight>

The exposed-as attribute makes that only service known as fr.paris5.shaman.insight.service.exposed.HelloService will be seen from the RMI client side.

The rmi-server attribute gives the DNS name of the RMI server.

The run-style attribute can be one of following :

• standalone
• rmi-client
• rmi-server

The role and class attributes of each service do keep their meaning as defined by Avalon-Excalibur Component Manager.

Insight maybe instantiated / configured from an Excalibur-compatible file (like cocoon.xconf).

Here is a sample of an embedded Insight configuration file, with Insight running in the same VM as the Exacalibur Component Manager. Please note the role and style attributes which appeared, telling the Excalibur Component Manager how to instantiate the Insight Manager.

<my-excalibur-configuration-file>

  <!-- some declarations here... -->

  <insight
      xmlns:insight="http://shaman.paris5.fr/insight-conf"
      role="fr.paris5.shaman.insight.component.InsightManager"
      class="fr.paris5.shaman.insight.component.InsightManager"
      rmi-server="myhost"
      run-style="standalone"
  >

    <foo-service
        role="fr.paris5.shaman.insight.service.unexposed.Foo"
        class="fr.paris5.shaman.insight.logic.unexposed.FooImpl"
    />

    <hello-service
        role="fr.paris5.shaman.insight.service.exposed.HelloService"
        class="fr.paris5.shaman.insight.logic.exposed.HelloServiceImpl"
        exposed-as="Hello"
    >
      <hello-message>Hello, world</hello-message>
    </hello-service>

  </insight>

  <!-- some declarations here... -->

</my-excalibur-configuration-file>

If the configuration should describe a rmi-client behavior, it would look like :

<my-excalibur-configuration-file>

  <!-- some declarations here... -->

  <insight
      xmlns:insight="http://shaman.paris5.fr/insight-conf"
      role="fr.paris5.shaman.insight.component.InsightManager"
      class="fr.paris5.shaman.insight.component.InsightManager"
      run-style="rmi-client"
      rmi-server="myhost"
  >

    <hello-service
        role="fr.paris5.shaman.insight.service.exposed.HelloService"
        exposed-as="Hello"
    >
    </hello-service>

  </insight>

  <!-- some declarations here... -->

</my-excalibur-configuration-file>

Note that unnecessary options have been removed (they could still appear, but would be ignored) :

• foo-service is not exposed, so it won't exist in a RMI client run style
• hello-message is a parameter supposed to be used by server only

Since Cocoon configuration files aleready tend to be huge, it may be preferable to not embed configuration. For achieving this, an Insight Manager declaration in an Excalibur configuration can declare a reference to a configuration using an URL :

<my-excalibur-configuration-file>

  <!-- some declarations here... -->

  <insight
      xmlns:insight="http://shaman.paris5.fr/insight-conf"
      role="fr.paris5.shaman.insight.component.InsightManager"
      class="fr.paris5.shaman.insight.component.InsightManager"
      configuration-url="./insight.xconf"
      run-style="rmi-client"
      rmi-server="myhost"
  />

  <!-- some declarations here... -->

</my-excalibur-configuration-file>

The configuration-url defines a reference to the file to use. The run-style and rmi-server will override the values (if any) in the given ./insight.xconf.

Since defaults for services will be used most often, it's possible to let the Insight Manager use its own bundled resource for configuring itself. In this case, the declaration in the Excalibur configuration file will look as follow (rmi-server attribute may be forgotten, defaulting to localhost) :

<my-excalibur-configuration-file>

  <!-- some declarations here... -->

  <insight
      xmlns:insight="http://shaman.paris5.fr/insight-conf"
      role="fr.paris5.shaman.insight.component.InsightManager"
      class="fr.paris5.shaman.insight.component.InsightManager"
      run-style="rmi-client"
  />

  <!-- some declarations here... -->

</my-excalibur-configuration-file>

When referencing (e.g. declaring a configuration-url attribute, it's not allowed to declare nested elements.

Avalon's Contexts are supported, for allowing a Java program to define some context values without defining a Configuration object.

All Key objects are defined in fr.paris5.shaman.insight.InsightContext.

Paragraphs above pointed out there are several ways to configure the Insight Manager. Configuration items are taken in the order of preseance defined below :

1. Excalibur element
2. Context
3. The class-bundled configuration is used

1. Excalibur element
2. Configuration file
3. Context
4. "localhost" used as default

1. Excalibur element
2. Configuration file
3. Context
4. "standalone" used as default