New developments like XML, OLAP, and object-oriented technologies continue to change the ways we are collecting, storing, and consuming information; the very nature of the information keeps changing and often involves new media and new formats.
XML emerged as the de-facto information exchange standard and not surprisingly, relational databases responded by incorporating XML into their cores. The approaches taken by each of the RDBMS vendors might be different (XML documents might be mapped and parsed in familiar text-based records, or be stored as complete documents), but the details of these implementations have become increasingly irrelevant to the vast majority of developers and users.
OLAP became a standard for BI — business intelligence. With the enormous amount of data — structured or otherwise — accumulated since the dawn of civilization, it was only a matter of time before someone would take data comprehension to the next level, which is to discover statistical trends. While not part of the RDBMS technology, BI does not make much sense without some kind of a database — relational, in our case. The main processing unit of this information is a multidimensional CUBE, which can be manipulated using either some general-purpose language (like Java) or some proprietary language (like Microsoft MDX). Some vendors bundle business intelligence tools with their RDBMS, some BI tools are stand-alone tools built by third-party companies.
The object-oriented approach became the de-facto application programming standard, and as such made a compelling case for object-oriented databases. As we model the surrounding world in terms of objects, we need a place to store these objects. An RDBMS maps the objects to words; an OODBMS will accept them as they are. You may compare it to a book, where images are created by your brain from mere words; the movie stores and communicates visual objects directly to your senses, bypassing the verbalization step.
OODBMS may well be the wave of the future, which is notoriously unpredictable. As of today, many companies have implemented object-oriented databases, designed to store and retrieve objects created within some particular language (Java, C++, Smalltalk). Eventually, new standards will emerge and performance gaps — if any — will be eliminated, making RDBMS outdated. For now, RDBMS remain the pillars of the business community, though they do pay lip service to the objects, incorporating them as data types but warning against the inefficiency of using them.