DBMS Server Comparison Supplement, November 1996
If the turbulent application development tools market resembles a vibrant and densely populated city, then the relational DBMS market is more like a sleepy small town dominated by a few big players. RDBMS upgrade cycles are longer, and even major revisio ns usually do not alter the fundamental assumptions underlying the server's architecture and functions. Over the next year or two, however, leading vendors will shake up the RDBMS market by rolling out support for complex (non-tabular) data managed by in creasingly object-oriented techniques built into the server.
A modern RDBMS server is already a complex piece of software. Intelligently supporting complex data types (other than dumb BLOBs) and adding support for class definitions, inheritance, methods, polymorphism, and other object-oriented features will only m ake tomorrow's Universal Servers far more complex.
Yet it is crucial to realize that support for complex data types will supplement -not supplant - support for good old-fashioned tabular data. Neither transactions nor tables are going away. A Universal Server will also manage documents, video, time-serie s data, and many other classes of data, but these servers must also be backward-compatible with today's systems. If not, they will be relegated to niche player status just like object DBMSs.
"Not one of the Big Six [RDBMS servers] supports domains. It is crucial that these products address this fundamental concept of the relational data model," Martin Rennhackkamp states in his feature introducing this supplement. If RDBMS vendors skipped th is beat, what kind of foundation do we have for supporting more complex data? If you agree with Chris Date that relational domains are analogous to object-oriented classes, this point becomes rather distressing.
The term "Universal Server" may harbor a bit of irony. Today's RDBMS servers all support at least the basics of SQL-89, and if you avoid fancy vendor-specific enhancements you stand a fighting chance of porting data and SQL-based applications from one RD BMS product to another. Vendors may seize this opportunity to differentiate their servers with unique (read: non-portable) features. Most large companies use several different DBMS servers, so ask your vendor hard questions about the future of interopera bility in a heterogeneous world.
I am not trying to scare anyone out of town. Rather, this special DBMS Server Comparison supplement will help you take stock of where the leading RDBMS servers are now.
