Process Architecture Study: Documentation
Reviews
Back to the Overview
All elaborated documents were later handed over to two independent
reviewers in order to examine the three solutions. This was done in an
attempt to acheive a more objective evaluation of the produced
solutions.
As described in a previous chapter the documents were reviewed by two
reviewers who both hold a associate professor position at the
university. They were both given half an hour to examine the three
documents and then afterwards they were to give an oral evaluation
which also lasted half an hour. The oral evaluation was recorded with
a dictaphone. The following two sections are transcripts of the
evaluations with the two reviewers.
The First Review
The first evaluation was performed by an associate professor from the
research group of the formal sematics. The general impressions of his
evaluation can be summerized in the following:
- The focus in the object-oriented solution is on data structures.
- The focus in the two other paradigms is on aspects of
synchronization.
- The solutions from the OS and ML paradigms reflect great
complexity.
- All solutions lack a more general design strategy.
First, the object-oriented paradigm is primarly focussing on
data structures described by the objects. According to reviewer R1 the
advantage achieved here is that one can quickly turn from specifying
the design to actual coding the implementation. However, he also
stresses that it would be difficult to verify the requirements stated
in the problem statement with this approach.
The two other paradigms are characterized be the fact that they are
focussing on aspects of synchronization identified in the problem.
Reviewer R1 is of the opinion that this is very useful when avoiding
dangerous situations where the system does not embody the right kind
of behaviour. However, these two solutions also reflect great
complexity and according to reviewer R1, this cannot be avoided even
though you use automated machinery during the design process. In fact,
he is the opinion that the lift control problem would probably
explode in complexity due to the numbers of actions in a final,
complete solution. Hence, it would be very difficult to solve the task
when only using one of these two paradigms.
At the more general level, reviewer R1 stresses that all three
solutions are lacking a more abstract design strategy in order to
outline the requirements to the system. This has not be done in any of
the three solutions. On the contrary, all three solutions reflect much
more specific design where the lower level components have been
identified and specified.
As a final evaluation of the solutions reviewer R1 says that it is
difficult to compare the three solutions due to their very different
approaches. He states, however, that all solutions show a fair degree
of being useful if they should have continued with the design and
implementation phases. So marking the solutions with grades he would
give them the following:
The Object-Oriented Solution: Good
The Operating Systems Solution: Good
The Mathematical-Logical Solution: Good
The Second Review
The second evaluation was performed by an associate professor from the
research group of programming systems. His general impressions can be
listed as follows:
- The operating systems and the mathematical-logical paradigms
embody great complexity.
- All solutions is charactrized by both incompleteness and a very
technical approach.
- The solutions are more implementation than actual design.
Reviewer R2 states that the object-oriented solution at first hand
attracts him at most. This is mainly due to fact that the use of
objects are well-suited for identifying real world entities such as
the elevators, the floors etc. He is aware of that it might be
difficult to verify that the object-oriented solution shows the right
kind of behaviour. He is, however, quite certain that it would be
possible to incorporate some invariants, but he also stresses that it
could not be done in an as elegant maner as the two other paradigms.
He stresses that the operating systems and the mathematical-logical
paradigms are very complex and enormous:
It easily becomes signs and strange actions, where you
cannot associate anything with these widgets. A semantic gap is
introduced between your intuitive understanding of the problem
and the actions in the paradigm.
Generally, he finds all three solutions incomplete. Actually he is of
the opinion that one cannot consider them as being finished. In
addition, he feels that they are enormous. He more considers the
solutions as partial solutions inspired by the rules in the problem
statement. As far as he can see none of the three designers have
verified whether the six rules are consistent or not, or whether there
are some unclarified questions hidden in the problem statement. He is,
however, uncertain how the task could have been approached at a higher
level. In fact, he admits that he would probably have done in the same
way if he had been confronted with this particular task.
On the question of rading the solutions he says that he find all
solution average. He marked the solutions as follows:
The Object-Oriented Solution: Average
The Operating Systems Solution: Just below average
The Mathematical-Logical Solution: Just below average