Research in the field of relational databases has been ongoing for over 30 years.  Researchers have developed remarkable techniques for optimizing the storage and retrieval of highly-structured data.

One large limitation of relational systems is that they are inherantly two-dimensional; they have no notion of hierarchy, and thus, are not well-suited to storing semi-structured information, where hierarchy is often a key component.

To overcome this problem, new database technologies and query languages have been developed specifically for the support of semi-structured data.  In these systems, each record has a set of properties that can be specified hierarchically.  Moving back to the games example, one might specify the "American" name for Football as:

game.name.american = "Football"

 There are also corresponding special query languages that have been developed for these databases, which give the user access to these hierarchical properties, thereby allowing for easy storage and querying of semi-structured data.  Unfortunately these query languages were not standard XML query languages -- they were database specific.

The Major Drawback of semi-structured databases is that they are SLOW.  None of the prominently developed databases have proven to be scalable beyond a collection size on the order of 10MB, which is clearly unsuitable for large-scale information processing.

Find out how we solved this problem in final part of the lesson here.

Last updated: May 01, 2002.