Objectrad Project: Java Metadata Server

Objectrad recently completed development of a Java Metadata Server for a global IT company. We expect them to introduce it to the market soon. This project illustrates the technologies we employed, and shows a little about our architectural philosophy.

The application is a metadata server, serving metadata to client applications and/or web servers. Users can define new types of metadata. The server will generate the needed code, making it capable of serving the metadata of the user-defined type.

Although focused on metadata, the application is a kind of model-based application development environment. Users define classes, associations, attributes, operations, constraints, etc., using UML-like elements. The application uses the model to generate and compile the Java source code. The result is added to a running server. Users supply needed method implementations.

Imports metadata from XML streams into the server, and export metadata, in the format required by the particular type of metadata. Each metadata type has its own DTD. Provides a completely generalized capability, to handle any form of XML, with no code generation.

By abstracting out the metadata storage and retrieval requirements into a generic persistence layer, persistence-independence is achieved. New implementations of persistence can be added by implementing interfaces defined in this layer.

Various kinds of metadata is representable, including SQL Schemas, UML models and Java source code. New kinds of metadata can be added, by defining the metadata type with a metamodel. Developers can extend existing metamodels to add new capabilities.

When a new type of metadata is needed, a metamodel (metadata about metadata, in a UML-like form) is used to define it. From the metamodel, the Java code required to support the new metadata is generated and compiled.

Operations are defined in a metamodel, to give behavior to the metadata objects. Implementations of the operations are plugged-in by the developer. They can be added dynamically, while the server is running.

Constraints can be defined on metadata types, using the Object Constraint Language (OCL). These constraints are translated into Java code, as part of code generation. Developers can also plug-in their own constraint implementations, dynamically, while the server is running.

Client applications can use generic interfaces, which support all types of metadata, or interfaces specific to a metadata type, which are generated from a metamodel. The client application does not change when the form of middleware changes. So whether the client is using CORBA or is locally in the VM with the server, the client application doesn't change.

Our architecture emphasized:

	separation of domains, with careful management of dependencies - to provide the greatest flexibility and extensibility; the result was software that could be configured into dramatically different applications

	pluggable services and features - we define interfaces for plugging in components, and use this approach for many of the application features; others can replace the features or add new features

In the diagram below, move the mouse over the architecture components to see the description of each part.

This metadata server is extremely tailorable. One configuration is for a single-user desktop application, with the metadata loaded from and saved to XML files. But in other configurations, the application supports the definition and management of metadata distributed across the enterprise. Metadata reference remote metadata; metadata can be used to define relationships among other collections of metadata. From a Java client, through JNDI, the enterprise's metadata is just held within a directory structure, with no concern over the physical location of the elements.

The configuration below is n-tier. Metadata Servers can share the same Database, or each use separate databases. The Metadata Servers act as clients in managing the links between metadata elements on separate servers. The Metadata Server supports Java clients directly, through JNDI; using a Web Server, browser clients can access the metadata.

Technologies Used
The application includes these technologies:

	J2EE - JNDI, JDBC, CORBA, CosNaming
	XML
	UML metamodel
	Java - reflection, JFC, sockets, security
	JavaCC parser compiler