Expert Core: Semantics

An enterprise repository stores complex objects engaged in complex relationships. Unlike most databases, which are designed for simple, highly structured records engaged in fairly simple relationships with other records, repositories must be able to support:

• Complex objects
• Bi-directional, many-to-many relationships
• Deep inheritance
• Recursion
• Rich versioning
• Extensible metamodels

Interestingly, many so-called repositories are implemented with conventional database technology, which typically does not support all of these requirements. When vendors use conventional database technology for applications requiring these advanced features, they’re pushing the envelope.  This is something to be aware of when you're evaluating repository solutions.

By conventional database technologies, I primarily mean relational and object-oriented. Also included, however, are less common (these days) database technologies, such as hierarchical, network and multi-value.  Relational databases, for example, while very flexible don’t scale well when too many runtime joins are required to assemble complex objects. Object oriented databases overcome some but not all of the limitations of relational databases. The relative capabilities and limitations of both types of databases are summarized in the following table:

Configuration Items (CI) refers to the contents of a Configuration Management DataBase (CMDB), which provides a convenient frame of reference for defining repository requirements since CIs can be complex and engage in complex relationships.  We can see from this table that, even with object oriented databases, there are limitations.

There is actually a relatively unknown class of database technology addresses the limitations noted above. That class is a topic for a future posting.