An "extensible DBMS" is a data manager where the "data model" (say,
relational or SQL data model) is treated more as a framework than as a
finished product. For example, the relational data model has logical
notions of domains, relations, and relational operators. In an
extensible DBMS users can 'extend' the physical implementations of each
of these. For instance, they can add to the set of domains--adding a
new user-defined data type and operators--relations--if you don't like
page oriented disk access but still want to be able to write queries
you can implement your own storage layer--and even relational
operators--want one of Date's fancy temporal operators? or got a new
join algorithm you want to add?. Note that this doesn't mean that an
engineer working for the vendor can add these things, but rather that
an "after market" in extensions can be created which "plug into" the
DBMS and extend its functionality.
The marketing term most frequently associated with this idea was
"Object-Relational". Early work on building this class of system was
undertaken in the late 1980s. Notable extensible DBMSs today would
include PostgresSQL. Most commercial SQL DBMS products today support
such extensions to a greater or lesser extent.
Received on Wed Dec 08 2004 - 17:28:42 CST
