understanding application scoped components in a cluster

[codelion]

Making sure I get this right:

An application scoped component is good e.g. to keep settings. It has the same contents on all machines in a cluster. Nice magic working for us inside JBoss.

Now if I change (i.e. code changes) the value of a member field it will be available on all machines in the cluster, right?

Now if that application scoped component e.g. contains a map (e.g. LinkedHashMap), and the map contains settings (entity beans in memory), which contain a name and a value (logically pairs), and if I change (i.e. code changes) the value in a setting object, will that be available on all machines in the cluster?

The question is, how far into that tree from the component to its map to the settings does that replication work? Anywhere in the tree? (Assuming replication is the right word, is it?)

[norman.richards]

I don't think application scope components are clustered.

[codelion]

If they aren't clustered, then how does this work in Seam reference section 3.1.7. Application context:

Application context is mainly useful for holding static information such as configuration data, reference data or metamodels.

How then to update configuration data?