Conclusion - User-Level Workflow Design: A Bioinformatics Perspective - page 189

domain information or by recording and later reproducing the decisions taken
by the synthesis algorithm during the assembly of the solutions. Ideal and
more substantial, however, would be to enable the framework to directly per-
form instance-based workflow synthesis and thus avoiding ambiguities from
the beginning.
Usability
In order to improve plugin usability, especially more sophisticated solution
choosing mechanisms could be applied. A very effective feature would be the
prioritization of solutions, for instance along the lines of [289], in order to
achieve a more meaningful sorting. Furthermore, it is envisaged to replace
the current functionality for choosing a solution from a list of textual rep-
resentations by a graphical, interactive solution chooser that visualizes the
entire set of solutions in a more comprehensible automaton representation,
and supports different mechanisms for choosing a particular solution. An addi-
tional helpful feature in this regard would be kind of “automatic explanation”
mechanism of the synthesis procedure itself. This would make the individual
steps that are taken by the algorithm comprehensible through visualization
of the current state of the search, possibly also allowing the user to influence
the search procedure manually.
Minor usability improvements concern, for example, the further simplifi-
cation of domain modeling, for instance by providing specific wizards for
service descriptions and thus freeing the user completely from dealing with
configuration files.
Explicit Variability Management
Model-based workflow development with the jABC framework and related
technologies has always focused on supporting the agile handling of variant-
rich software applications in different ways [208, 210]. Already the basic jABC
framework provides a workflow definition layer at which variants of workflows
can easily be built, which has been demonstrated for bioinformatics workflows
in [172]. The abstract workflow descriptions that are applied by PROPHETS
are furthermore declarative (but nevertheless intuitive) workflow specifica-
tions that represent all possible variations.
Unless explicitly configured by the user in a different way, the PROPHETS
plugin makes the variant-richness of the workflows visible to the user: he can
select a solution from a set of candidates, or he can decide to edit the applied
constraints and run the synthesis again in order to refine the set of candidate
solutions before selecting one manually. In fact, this is a striking difference
to the other approaches that have been considered in the scope of this topic
(cf. Chapter 8): the selection of a particular solution is typically left to the
composition algorithms, which simply take the first solution that is found,
or apply more or less sophisticated heuristics for this task. Somehow, this