|
|
OO Process and Metrics for
Effort Estimation
|
convention centre
room 7 |
This workshop is a continuation of similar, very
successful workshops held at OOPSLA `95, OOPSLA `96 and OOPSLA `97 with the same theme.
Software development remains difficult to plan.
Object-orientation has thus far not changed the picture. However due to the greater
continuity in OO across the life-cycle - objects everywhere - there is good hope that OO
may improve the predictability of OO projects. At least two ingredients are required for
planning SW projects:
- More detailed processes than what is
currently available (waterfall, spiral, fountain, clean-room, etc.). This would help to
identify more intermediate milestones.
- Effort tracking and estimation metrics. This
would help to measure progress and to recognize earlier that unexpected roadblocks have
been encountered.
Problem Areas:
- What are good micro processes for OO
analysis, design, implementation, testing, maintenance?
- How do we preserve the developers' creativity
when development processes are defined with a finer granularity than is currently
customary?
- How do we ensure compatibility of micro
processes and macro processes like waterfall, spiral, fountain, clean-room, etc.?
- What are project characteristics for choosing
a macro process?
- How does reuse fits in the development
process?
- At what level should process instrumentation
occur? At the individual level (e.g. Humphreys PSP)? Or can higher levels of granularity
be just as good? What is the relationship between measurements taken at different levels?
- Can the Albrecht function point technique be
used (as is? modified?) for an initial effort estimate?
- Are task points a good alternative?
- How do we estimate remaining effort during
development?
- How frequently should measurements be taken
during a project and how should this fit into a given project life-cycle?
- How do we correct effort estimation for
reuse?
- Which product metrics, if any, can be used
for effort estimation?
- Effort estimation is rooted in reasoning by
analogy. Can we have a generic estimation technique or will they be domain/ company/
product line/ (?) specific?
- Effort estimation requires systematic
collection of data. At what granularity? Per subsystem? Per class? Per attribute? Per
transition network? Per member function?
- What role can/ should OO CASE tools play?
Organizers:
Granville Miller, Make Systems
Email: gmiller@makesys.com
Chris Ball, American Management Systems
Dennis de Champeaux, OntoOO, Inc.
Philip Haynes, Object Oriented Pty
Ltd.
Brian Henderson-Sellers,
Swinburne University of Technology,
Australia