In-Depth

If at first you don't succeed...

Other than Microsoft, the long-running software story of the last quarter century has been the database business. For much of the period, the relational database has been dominant. Through the years, experts have touted object-oriented databases and hybrid object-relational systems as potential replacements, but to date none have succeeded.

The top relational database players have quietly started re-emphasizing object-relational technologies with the emergence of corporate Internet computing, but all are treading lightly due to past missteps. The term object relational has suffered a series of severe blows during the past few years for several reasons, such as Oracle failing to ship a promised system and Informix's doomed strategy to transform its full focus to object relational.

"The chalice has been poisoned, so to speak," said David Wells, an analyst at Ovum, a London-based consulting firm. "Oracle 8i has some interesting pieces of object technology, Microsoft with OLE DB has a whole range of object features and Sybase is adding object technology. But they are choosing not to talk about it because of the perception of failure."

The emergence of the Internet, the popularity of Java and some upcoming new database versions from Oracle Corp., Redwood Shores, Calif., IBM and Informix Inc., Menlo Park, Calif., could allow these newer hybrid systems to succeed where others have failed, say observers. Oracle plans to ship what it calls an Internet database, the object-relational Oracle 8i, later this spring. The new Oracle database incorporates the interMedia technology previously sold as an option for storing multimedia data types, and adds Java technology to the kernel. The new strategy is seen as the top DBMS vendor's strong embrace of object technology.

Informix and IBM also expect the spread of Internet-based applications to widen demand for their respective universal database technologies. Observers expect Microsoft Corp. to stake a claim in the object-relational territory, perhaps integrating the OLE DB and Microsoft Data Access technologies into the SQL Server DBMS. The universal database strategy of Sybase Inc., Emeryville, Calif., is unclear.

"We still see growth," said Jeffrey Jones, program manager for IBM data management marketing. "The Universal database capabilities are not the only reason to buy a database, the [object] extenders [to DB2] are a bonus, a checklist item. Use of the extenders is growing nicely, slowly but steadily."

Meanwhile, three-year-old Cloudscape Inc., Oakland, Calif., has built an object-relational database from scratch to take advantage of the emerging Java revolution.

Frank Greco, founder and chief executive of Crossroads Technologies, a New York City-based systems integrator, said the Cloudscape system is ideal for financial services firms. Java has spread quickly through the business while any database has to support SQL, which is a standard at most financial services firms.

Informix has signed an agreement with Cloudscape to utilize its Java technology in its new Centaur database, which joins the Universal and Dynamic Server products. David Applebaum, Informix vice president of product marketing and management, said the new system, set to ship by the end of 1999, will also incorporate Microsoft's ActiveX and COM technology.

While vendors are now convinced that Internet computing will boost the necessity of universal systems, Wells of Ovum said users must still be convinced of the need. "This isn't a great leap. Object relational does just a bit more than relational. There is a need for the [object-relational] technology, and it will increase as more Web-based applications are built," he said.

"There is a somewhat increased demand for bject-relational systems," added Carl Olafson, an analyst at International Data Corp. (IDC), a Framingham, Mass.-based consulting firm, "particularly in some specialized vertical niches. Any mainstream acceptance will be driven by widespread adoption of Web technology."

According to Olafson, the timetable for widespread acceptance of what IDC calls "Cybersmart Computing" is still two to three years away. In the meantime, relational databases will remain the standard, he said. To date, most observers expect today's top makers of relational systems, primarily Oracle and IBM, to hold a clear advantage in providing products for the new paradigm.

Meanwhile, top database suppliers have an opportunity to generate sales in niches once confined to OODBMS systems, said Arthur Berrill, director of the SpatialWare product line and chief developer for spatial technology at MapInfo Corp., a Troy, N.Y.-based maker of business mapping software. "Years ago, the spatial niche could be handled by pure object technology," he said. About six years ago, MapInfo built object extensions to Oracle.

A little more than 18 months ago, MapInfo decided that Oracle, IBM and Informix could provide the necessary object extensions and began bundling those products with its MapInfo software. "Today, we expect the database vendor to provide the object-relational extensions," Berrill said.

MapInfo also plans to evaluate Microsoft's SQL Server 7.0 database, which continues to utilize add-on products like OLE DB and Microsoft Data Access to add Universal server capabilities. "Our multimedia support is still outside the database," said Jose Blakely, senior architect on the SQL Server development team. "More and more people want the ability to have the logic run inside the server, and we see that as a good thing."

Efforts to supplant relational systems about a decade ago by a group of object-oriented DBMS start-ups (most with the word "object" in their title) proved fruitless. Oracle, IBM and Informix responded to the OODBMS challenge by adding object capabilities to relational databases, thus claiming to eliminate the need for the new systems. While that effort did not crush the OODBMS revolution, it pretty much eliminated any chance of those systems reaching the mainstream. "The object database market is small but steady, it's a niche market," said Mary Hope, a senior analyst at Ovum. "We see nothing to suggest the technology will become mainstream."

Object database technology does have its champions, including software giant Computer Associates International Inc., Islandia, N.Y., which called off an effort to add object extensions to its Ingres database in favor of developing Jasmine, a pure OODBMS. "We had to make compromises on both the object and relational side," said Carl Hartman, CA's vice president of information management marketing, of the effort to build a hybrid system.

Rousseau Aurelien, chief executive at WebEnsemble, a New York City maker of digital asset and knowledge management tools, says "the hype around object relational is one of my pet peeves." Aurelien's product line incorporates CA's Jasmine. "The existence of object relational shows the need for pure object-oriented databases. Object relational still requires a lot more programming than object databases."

Adds Wells of Ovum, "The object database vendors have a number of important opportunities -- the Web is an opportunity for them. But they do face significant competition from relational vendors. Unless they can do something big, then Oracle will be perceived as good enough."

Despite the early hype from Oracle, Informix and IBM, IT organizations have so far failed to buy into the need for the universal database technology in any significant numbers. These early efforts to add object capabilities to relational systems failed for a variety of reasons. Ovum's Hope maintains that object-relational technology has always been valid for IT organizations, but that sales "problems started with the way vendors positioned it. Instead of stressing business benefits, they confused the market by talking about the wonderfulness of the technology."

IDC's Olafson added that "universal systems require a fair amount of knowledge of complex data types and specialized programming skills. A lot of organizations just didn't feel they had a need for this."

The failure of Informix's highly publicized effort to shift its focus from relational to object relational about three years ago has been well documented. Not only did the new technology fail to catch on, sales of the Informix relational technology flagged due to inattention. The calamity led to significant losses, a drop in revenue, management departures and shareholder lawsuits. "Informix did an awesome job of building object extensions," said Berrill of MapInfo. "Informix forged the direction, but suffered an enormous amount."

Under new leadership, Informix has radically shifted its focus, twice repositioning its product lines, and has returned to revenue growth and profitability.

Oracle's Larry Ellison proclaimed his firm's plan for object-relational databases shortly after the birth of the OODBMS market, but the plans faded quickly -- especially after the Informix debacle. "We've seen a lot of changes over the last 18 months to two years," said Dom Linders, director of industry marketing at Oracle.

Oracle 8i has caught the attention of some users. "Oracle 8i is fundamental to our future direction," said Berrill of MapInfo. "It's clearly a step in the right direction to put the Java technology into the kernel."

However, a few more steps need to be taken before the technology can enter the mainstream.