home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.umcs.maine.edu
/
2015-02-07.ftp.umcs.maine.edu.tar
/
ftp.umcs.maine.edu
/
pub
/
WISR
/
wisr4
/
proceedings
/
detex
/
solderitsch_2.detex
< prev
next >
Wrap
Text File
|
1992-04-05
|
10KB
|
242 lines
[12pt] article
An Organon: Intelligent Reuse of Software Assets And
Domain Knowledge
James Solderitsch
Valley Forge Laboratories
Unisys Defense Systems Inc.
Paoli, PA 19301
E-mail: jjs@prc.unisys.com
Methods and techniques to produce large
systems can be viewed on a continuum ranging from pure manual
construction with no tool-based support to a highly
evolved environment that greatly aids in the engineering of these
systems. An organon represents one view of such an environment.
The paper describes work underway at Unisys, partially supported by the
DARPA/SISTO STARS Program, to advance the state-of-practice on the
path to such an organon.
0.3in
Keywords: reuse, software engineering, knowledge-based systems,
construction, generation, domain model
* Introduction
Unisys and its affiliates participating
in the STARS Unisys is supported in STARS
under contract number: F19628-88-D-0031.
(Software Technology for Adaptable, Reliable Systems)
program are keenly interested
in understanding, developing (or acquiring) and applying technology to
support the development of complex software systems based on a
reuse perspective.
This support is currently reflected in joint activities being pursued
by all of the STARS Prime contractors (Boeing, IBM and Unisys) to produce
a STARS Reuse Concept of Operations (CONOPS) and an Asset Library
Open Architecture Framework which together
will guide future STARS efforts in the
area of software reuse.
The content of this paper
reflects an approach to reuse that is in tune with
the emerging STARS CONOPS. This approach emphasizes
domain life-cycle activities which utilize a variety of techniques
and methods. Too often,
the word reuse is interpreted strictly to mean the reuse
of code components, whether informally through ad hoc reuse by
a programmer remembering and reusing a previously composed code
fragment, or more formally by finding and retrieving code from a
code library.
Reuse should be understood in a larger
context to mean the reuse of knowledge about the complex application areas
(or domains) to be served by software systems for these areas.
Engineering models that capture knowledge about these application domains
can provide a vital basis for technology to support reuse.
Such domain models themselves require their own enabling technology
so that they can effectively infuse the application development process.
* Vision
The mode in which domain models drive the development
of systems can be viewed on a continuum ranging
from the production of a system via
a custom design and handwritten code to the
automatic generation of a system
from a high-level specification in a domain-specific language.
In the past, knowledge gained from system development
and deployment has been left in the heads
of expert software developers, and has not been extracted and collected for
corporate reuse. Thus there has been little movement on the maturity
continuum. The development and application of domain models can
accelerate this movement and thereby boost system quality and system
development productivity.
Unisys believes that a knowledge-based approach to the creation of
domain models can provide a successful strategy for moving the
engineering of software systems further along this important maturity
scale.
An Organon
The dictionary defines organon as ``...an instrument
for acquiring knowledge; specifically, a body of methodological
doctrine comprising principles for scientific and philosophical
procedure and investigation''. The ancient library at Alexandria
was known as an organon.
represents a vision of a knowledge-based environment able
to support large-scale reuse based on the development and use of
domain models. An organon will:
be an interactive and evolving public storehouse of expertise and componentry
serving particular application domains;
effectively support wide-spectrum reuse including
requirements, design and test cases;
support a number of different points-of-view on system
engineering including those of construction and generation; and,
be a central repository of domain expertise
that effectively combines people, plus emerging and maturing methods, plus
supporting technology.
As domain models mature,
the component-based construction of systems by human hands can evolve
to a computer assisted, and eventually computer controlled, production
of these systems from available components
based on a domain-specific software architecture.
When domain knowledge is sufficiently complete, work at the component
level will give way to design and specification at the subsystem level
with the corresponding system components produced automatically
from templates and other library components.
In the end, an organon can support complete generation of
software subsystems from high-level, domain-derived specification
languages. Such generation is
possible for domains that are mature enough to admit the
expression of parametric information in sufficient detail to allow the
automatic production of code based on this parametric information.
This vision represents an advanced instantiation
of the STARS Reuse CONOPS that is currently
under development. The CONOPS proposes a Reuse Process Framework
that incorporates families of related reuse supporting activities:
reuse planning;
asset creation;
asset management; and,
asset utilization.
An organon will enable and facilitate the production, management and
application of software assets which will run the gamut from source
code modules and test cases to domain-specific requirements to
software generation subsystems.
* Reality
Unisys is currently working to lay a
foundation to move the state-of-practice further
along the model-based maturity scale described earlier.
Unisys is acquiring and producing a technology base to provide
a machine readable and processable representation of domain models
in a form which domain-specific tools may directly utilize and manipulate.
The Reusability Library Framework (RLF), a step along
the path leading to an organon, is a set of Ada knowledge-based
tools (semantic network and rule-based systems that can be used
in concert) to support the definition
and manipulation of domain models.
The RLF is currently targeted to the production
of domain-specific, knowledge-based
software library systems where the
engineer is supported in his or her production of software systems using
software assets contained in the library. Domain models are captured
in sophisticated semantic networks which capture the crucial objects,
operations on objects and relationships between objects within the domain.
Library assets are catalogued and stored according to this domain
model. In addition, the human engineer is aided by a rule-based inference
system which provide guidance in exploring and understanding the
domain model and the assets within the library. Domains addressed to date
include Anti-Submarine Warfare (ASW), User Interfaces (in particular
the use of Ada/Xt) and Ada benchmarks.
The RLF has also been used to develop
a model-based tool utilization assistant (TUA) for the domain of
document preparation. An early version
of the RLF was used to produce an Ada Unit Test Assistant
(Gadfly) which contained
a model of test heuristics and generated test plans based on parsing
of Ada units and interaction with a human test engineer.
Broad objectives of the RLF project include:
develop knowledge-based interfaces to repository (object) management systems;
investigate the mapping between application domain and reuse technology
(part selection, part composition, part generation);
go beyond supporting retrieval of static parts to include
program generation, system/software configuration, system/software testing
and even system/software design and requirements analysis;
support the basic integration of reuse technologies (knowledge-based
and generation techniques); and,
perform some applied research in domain analysis.
Recently, an X Windows-based graphical browser interface was developed
for librarian applications built on top of the RLF.
* Conclusion
Currently, the RLF attempts to support the middle of the continuum with
the human engineer being guided by an RLF-based environment. In order
to move farther along, additional support for generative approaches must
be provided. The Fourth Annual Workshop on Reuse
presents a stimulating opportunity to share the Organon
vision with other researchers and practioners of reuse.
Through attending and presenting at the workshop,
it is hoped that the vision can be compared to current reality
and be influenced by the recent work of others.
In turn, the lessons learned and experiences of the RLF project and the
STARS Reuse joint activities
can help fellow attendees in their current work in support of reuse.
alpha
About the Author
Jim Solderitsch is technical lead for the portion of the Unisys STARS Reuse
task being performed at Valley Forge Laboratories, Unisys Defense Systems.
He previously served as chief programmer for the Reusability Library Framework
(RLF) project. Before coming to Unisys in 1986, he was an assistant professor
in the Mathematical Sciences department of Villanova University. He received
his Ph. D. in Mathematics in 1977. His current research interests include
knowledge-based approaches to software reuse and program generation
techniques. He is a member of the ACM.