JDM Strategy - Java Data Mining: Strategy, Standard, and Practice - page 103

Strategic Objective 3: Have broad acceptance among
vendors and consumers
All too often, a standards effort can become dominated by tech-
nologists enamored with the technical details of a problem and
vendors limited to focusing on the specific features they support,
rather than on ease of use or what the consumer needs. Involving both
consumers and vendors in the standards process helps to provide bal-
ance to the resulting work. The JCP encourages participation from
multiple vendors and consumers to ensure that Java language exten-
sions meet a wide range of needs. This and other aspects of the JCP, as
discussed in Section 17.2, give potential users greater confidence in the
resulting standard.
Tactically, we chose to involve multiple vendors and potential
consumers to ensure the standard reflects the capabilities and prac-
tices common in data mining tools, as well as the needs of industry.
For example, the JSR-73 expert group consisted of leading corporate
and individual members from both data mining vendors and
technology consumers, with Oracle as specification lead [JSR-73
2006]. 2

Strategic Objective 4: Be extensible
As in any healthy field, the number of data mining techniques and
algorithms continues to expand. Variations on traditional algorithms
are introduced to improve performance, quality, or ease of use. As
such, no standard is ever complete. New requirements or capabilities
are constantly evolving. Any of the proceedings from SIGKDD
[SIGKDD 2005] give testament to the research supporting advance-
ments in data mining.
To remain relevant, a standard must provide sufficient common
functionality, yet enable vendors to enhance and grow that function-
ality in advance of the standards process itself. Having a framework
through which new mining capabilities can be readily added makes
it practical for vendors to implement the standard, yet provide their
customers the same level of service and capabilities those customers
are accustomed to.
Tactically, we designed JDM to be an easily extensible framework
consisting of a small core set of interfaces from which JDM standard
and vendor-proprietary interfaces can inherit to provide new capabil-
ities (e.g., functions and algorithms) while keeping the same overall

2 See [JSR-73 2006] for a complete list of expert group members.