Information Technology Reference
In-Depth Information
dependencies does not mean discarding them. Indeed, they are still potential,
and may eventually become visible again with the progressive arbitration of the
other dependencies. Still, we observed that this filtering helps keeping the focus
of the end-users during this elicitation process.
5 Evaluation
To experiment and evaluate the RAINBOW approach, a validation protocol
was defined based on the
Participant-Observer
principles to monitor the use
of the RAINBOW approach, and the
Brainstorming/Focus group
principles to
analyse the resulting conceptual schemas, as defined in [20]. This protocol was
used for a first series of experiments where pairs of end-users and analysts were
asked to jointly define the conceptual schema of a future information system,
including constraints and dependencies, using the RAINBOW approach and its
tool support.
For each project, the first task consisted in preparing the experimentation
by defining the subject based on real-life concerns of the end-users, then train-
ing the participants to understand the method and to use the tools. Secondly,
the end-users and analysts were asked to apply the approach on their project
and focus on the five first phases, while observers took notes. In particular, for
the
Nurture
phase which dealt with the elicitation of constraints and depen-
dencies, the participants were asked to progressively provide data samples and
constraints, while arbitrating the candidate constraints suggestions. Thirdly, the
observations on the eciency of the approach were analysed, and finally, the
quality of the produced schemas was debated, taking in account schemas that
were designed by the analysts independently of the approach.
The analysis of these experiments notably highlighted that the RAINBOW
approach and tool support did help end-users and analysts to communicate
static data requirements to each other, inclusive of constraints and dependen-
cies. Though all the requirements could not be expressed through the toolkit, the
latter did serve as a basis for discussion and modifications. Since the validation
aspect of the proposed approach cannot be addressed more extensively in this
paper, the interested reader may refer to [21] for further details on the validation
process and methodology.
6Conluon
In this paper, we extend the user-oriented RAINBOW approach presented in
[2], and describe how it can be used to interactively elicit database constraints
and dependencies, and more specifically domains of values, optionality, exis-
tence constraints, identifiers and functional dependencies. The process, inspired
by the principles of Armstrong relations, uses form-based interfaces that were
previously drawn by the end-users themselves in order to let them provide a
limited set of data samples that will restrict the potential implicit constraints of
the underlying conceptual schema. Conversely, end-users are invited to arbitrate
