Software Paradigms
The Algorithmic Paradigm (AP)
 This is classical design automation.
 An application is viewed as a program that maps
an initial input to a final output.
 It is characterized as the execution of a domainspecific algorithm which, given a set of
requirements, generates a design satisfying them
in a finite number of steps.
 It views computing as an algorithmic style and is
essentially language-based.
 Significantly, the regularity of this paradigm does
guarantee the avoidance of errors.
The Analysis-Synthesis-Evaluation Paradigm (ASE)
 Methodology: given a set of requirements, the
software design process principally involves:
 Analysis: Analysis of requirements on
order to determine the different
components of the system
 Synthesis: Driven by the decomposition
in the Analysis step, combine the different
components to create your solution.
 Evaluation: evaluate the solution of the
Synthesis step.
 The ASE paradigm leads the designer to assume
that his problem is well structured.
 Example: Waterfall Model of the Software
Lifecycle. It derives from the desire to make
software development "scientific"; essentially to
define an engineering Method.
 The ASE paradigm, leading to the classical design
approach of decomposition, is not widely applicable
to software development: Decomposition essentially
identifies possible choices of components
considered independently.
 In general, the ASE paradigm is too rigid for all of
our needs although useful for some aspects of the
system.
 Also, requirements may be inconsistent and
Problems are initially incomplete
The Formal Design Paradigm (FD)
 Due to the limitation of decomposition,
abstraction with subsequent refinement was
introduced.
 Abstraction provides us with a more flexible
tool.
 It is in essence a refinement of the ASE
paradigm.
 Mathematics provides a tool for abstraction
and further as Hoare reasoned in 1986 at his
inaugural lecture at Oxford:
 Computers are mathematical machines
(behavior is mathematically defined)
 Programs are mathematical expressions
(describe precisely what)
 A programming language is a
mathematical theory (a formal system for
programming; notations, concepts,
definitions, etc.)
 Programming is a mathematical activity
(calculation, proofs, etc.)
 Thus software design becomes a mathematical
proposition or theory that solves the problem
as represented by the specification of the
requirements.
 Hence, in the FD paradigm, the design process
is an activity that exploits the traditional
methods of mathematical reasoning.
 However, like the ASE paradigm it assumes
that requirements are known and that
essentially the problem is well-structured.
The Artificial Intelligence Paradigm (AI)
 It assumes that software design problems are
not well-structured. Thus, rather than define a
rigid set of requirements, the system should
absorb domain-specific knowledge to an extent
that it can provide a set of possible answers
which satisfy the constraints implied by that
knowledge.
 The more the problem solving system relies on
domain specific knowledge the "stronger" is
the problem solving method itself.
 The paradigm is explicitly founded on the
concepts of search, knowledge and heuristics,
in that the search is determined by knowledge
of the problem domain and by a collation of
general heuristics.
 The design process involves: a symbolic
representation of the problem (the problem
space) structured in terms of:
 Initial problem state
 Goal or desired problem state
 All other states being reached or
considered in attempting to reach the
goal state from the initial state
transitions from one state to another to
search for the solution through the
problem space.
 "Expert Systems" are instances of such
"strong" methods.
 Example: 8-puzzle
 Cost function:
^
C = g(x) +h(x)
where
h(x) = the number of misplaced tiles
and g(x) = the number of moves so far
 Assumption: move one tile in any direction cost 1.
Initial State
1
5
7
2
6
8
Final State
3
1
5
4
1
5
7
^
C 1 4  5
1
5
7
2
6
8
1
5
7
3
4
2
8
4
3
4
^
C 1 4  5
2
8
3
6
4
2
6
8
2
8
3
6
4
1
2
5
8
7
1
5
2
8
7
3
4
^
C235
3
6
4
^
C303
3
6
1
5
7
^
C235
^
C3 25
1
5
7
3
6
4
^
C 1 2  3
^
C  2 1 3
1
5
7
2
6
8
2
8
7
3
6
4
Note: In case of tie, choose the leftmost node.
1
5
7
2
8
3
6
4
The Theory of Evolutionary Design Paradigm (TED)
 The adequacy of the design is solely determined
according to whether it meets the requirements
prevailing at that stage of the design process.
 In this paradigm, the software design is viewed as
an evolutionary process:
The requirement is to create a design framework
that accommodates change (and failure) and
utilizes one of the previous paradigms at the
appropriate places. Basically it is a collection of
tentative hypotheses such that:
One can attempt to provide evidence in their
favor to establish the "plausibility" of the design
believe in the design's plausibility may have to be
revised
 TED is based on the view of design as an "empirical
scientific" activity in contrast to the Formal
Paradigm which views design as a "mathematical
modeling" activity.
 TED views the act of software design as an
evolutionary process and the design itself, at any
stage of its development, as a tentative solution to
the problem posed.
 In other words, according to the evolutionary
model, a design at any stage of its development
(including the final stage) is:


an evolutionary offspring of an earlier
design form
likely to evolve further in the future
 Essentially TED maintains that design in mixed
paradigm systems can consist only of establishing
constraints on the implementation. Collectively
these constraints represent the design.
 The approach means, in practice, that designers are
forced to resort to satisfactory rather than optimal
designs, thus recognizing both the evolutionary
and error-prone nature of software development.
 Thus the design is not only tentative at intermediate
stages of its development but also when the
designers (or their managers!) see fit to terminate or
freeze the design.
 The software design problem solving is a special
case of the process of scientific discovery and
suggests that we recognize in our design
management systems:

the strong link between natural and artificial
sciences
the use of testable hypotheses as a method (use
of prototyping and simulation)
The nature of evolutionary systems and hence
the need for an incremental development
approach (a working product (or subset) which
evolves on a daily basis).
 Reference: “The Software Paradigm”, By Brian Warboys.


 Summary:

5 Types
1. AP: Algorithmic Paradigm



2.
ASE: Analysis-Synthesis-Evaluation



3.


FD assumes problem is well structured
It is a refinement of ASE, but uses abstraction
problem becomes a mathematical proposition
AI: Artificial Intelligence


5.
assumes problem is well structured
like Waterfall Model
It uses problem decomposition
FD: Formal Design

4.
AP assumes problem is well structured
application = program to map input to output
finite number of steps, no errors
It is explicitly founded on the concepts of
search, knowledge and heuristics
It does not assume problem is well structured
TED: Theory of Evolutionary Design



It uses the fast that software design is an
evolutionary process
It does not assume problem is well structured
Nothing is fixed, everything can evolve.