~. .

HEWLETT
a:~ PACKARD

Software Reuse at Hewlett-Packard

Martin L. Griss
Software and Systems Laboratory
HPL-91-38
March, 1991

software reuse, In this position paper, several software reuse

software process, related activities at Hewlett-Packard are
software described, focusing on those in which the author
construction, has been involved over the past 6 years. Recent
corporate reuse activities include an investigation (started in
program, hypertext, November 1989) of the opportunities for a
libraries, application systematic, HP Corporate Software Reuse
frameworks program that led to a new HP Corporate
Engineering project to define and execute such a
program. Planned work involves a combination
of consulting, training, methods development,
the writing of a reuse handbook, and several
pilot projects. Also related are two research
projects at HP Laboratories, one involving the
development of a prototype hypertext-based
reuse library management system (Kiosk), and
the other (FAB) exploring component-based
software construction using application
development frameworks for distributed
applications.

Internal Accession Date Only

Copyright Hewlett-Packard Company 1991

1 Introduction
It is widely believed that a systematic application of reuse to prototyping, development,
maintenance and software process is one of the most effective ways to significantly
improve the software development process[BE89, Tra88]. Software reuse simultaneously
offers improved time-to-market, software quality and application consistency, and
reduced development and maintenance costs[BB91].
New work in software reuse such as domain analysis[PD90], object-oriented methods,
library technology[FG90] and architectural frameworks promises to lead to a consistent
methodology that will be useful for domain specific reuse. It is also important to jointly
consider software process and software reuse[CHSW90], and how open-architectured
software engineering environments are built and tuned to support domain-specific reuse
processes[Boe90].
However, a cursory investigation of the status of reuse in many companies, such as
Hewlett-Packard (HP), indicates that while several groups are making progress with
some aspect of reuse, very few software practitioners see reuse as a major or systematic
part of their software process. Recently, several companies have started to report
significant success with their reuse programs. HP also is involved in understanding and
applying the state of the art and practice to remain competitive.
I have been directing software technology and software reuse projects at Hewlett-Packard
Laboratories (HP Labs) for over 6 years, with a recent focus on reuse-oriented research in
HP Labs. Since November 1989, as consultant to HP Corporate Engineering's Software
Initiative, I have studied the status of software reuse in HP, and exploring opportunities
to make software reuse a significant part of HP's software process for the 1990's. We
believe that a fairly broad, well coordinated software reuse program involving
management, process and technology is needed to make significant progress.
Our reuse-oriented work in Corporate Engineering and HP Labs is directed at two major
goals:

Making systematic software reuse a significant part of HP software development

processes.
Research in the area of reuse library systems and reuse-based applications
development.

My activities supporting these goals involve:

Consulting with HP Corporate Engineering on software reuse and several other

software process issues, to help define a coherent set of software initiatives aimed at
improving HP software productivity and quality, and at reducing time to market
and maintenance costs.
Investigating reuse inside and outside HP via an extensive e-mail survey, attending
several workshops and tutorials, and investigating reuse activities in several
companies and universities. The results were presented in talks and tutorials to HP
upper management and engineers during 1990.

2
 Defining and (since October '90) helping execute a Corporate Software Reuse
program, which will involve the investigation of impediments to effective reuse, the
development and acquisition of reuse methodology technology, the ongoing
assessment of HP's progress towards systematic reuse, and the discovery and
communication of best practices.
Leading a small HP laboratories project team to develop a hypertext-based reuse
library prototype for software libraries. In October 1990 we released the first
prototype to internal HP users.
Leading a small HP laboratories project doing research on methods of rapid
application development based on reuse and user programming, building upon
application- and reuse-oriented architectures and frameworks.

More details on several of these activities are given in the following sections.

2 HP's Experience with Software Reuse

Hewlett-Packard has an extensive history with software reuse, and currently has several
active projects in its divisions and laboratories. The following gives the flavor and scope
of HP's projects. Several have developed or use existing DBMS systems to store and
distribute software components.

Early work involved the development and networked distribution of a family of

instrument modules (in Instrument Basic).
Subsequent work used Objective-C to develop class libraries, tools and metrics.
Human-Interface classes and common instrument sub-systems have been widely
distributed within the company, and some have been provided outside. HP has also
developed and distributes Motif 'widget' libraries.
More recent work involves several multi-division domain analyses and common
architecture and library projects for several HP groups, such as embedded software
for instruments, and systems architecture for chemical and medical systems. Some
of the work involves generic application frameworks and major components to be
used in several products.'
Several other divisions are involved with various aspects of the reuse process,
setting up libraries, establishing reuse goals (such as deposition ratios, or reuse
levels), and establishing and running reuse councils or component projects.
Several divisions have been developing C++ extensions, tools and class libraries for
instrument systems and general-purpose use.
At HP laboratories, we are researching object-oriented analysis and design, reuse
tools, domain-oriented frameworks, user-programmable systems and components,
and various kinds of object-oriented systems and tools.
lOur Corporate Reuse program has begun to consult on domain-analysis, architecture and object-oriented methods to some
of these projects

3
3 Corporate Reuse Program
We have designed a Corporate Reuse program that will consist of a core team of software
reuse experts in Corporate Engineering, with several additional people working on
assignment with a selection of pilot projects in the divisions. We do not plan to build a
single Corporate reuse library (though we might provide a single library mechanism,
perhaps with a few key components "pre-loaded"). This focus seems most appropriate to
HP's culture and state of reuse and process maturity.
The core team will be involved in:

The development and execution of a reuse assessment (questionnaire and metrics),

to determine how different groups are progressing, and to diagnose various people,
management, process, technology and readiness issues.
The design and execution of several "pilot projects" with divisional partners to
develop and test methods and tools for Domain Analysis, Application Architectures,
Design for Reuse, Development with Reuse, and Library Management.
The collection and dissemination of reuse best practices, processes and guidelines in
a handbook, customizable to the process and needs of different divisional reuse
efforts. We will test and refine these guidelines in the context of the pilot projects.
Consulting as a team of "resident experts" to the pilot and other reuse projects in
HP.
Helping develop and coordinate training and other technology transfer activities
with the assistance of a Corporate Engineering technology transfer unit, and other
management and engineer software training, workshop and publication programs.
Coupling and coordinating the reuse program to other Corporate Engineering
Software Initiative programs (maintenance, configuration management,
prototyping, process and metrics).

This program is just getting underway: we are staffing the core team, we have started
assessing several divisional reuse programs, and we held a first Reuse Practitioner's
Workshop for 60 HP engineers and managers in February 1991. We expect to select the
first of several pilot projects in April, and will refine our plans as we proceed.

4 Hypertext-based Software Reuse Tool

Much of our work on reuse-based toolsets at HP Labs is guided by a view that
hypermedia-based systems are the most appropriate frameworks to integrate tools,
maintain links between all software workproducts (such as specifications, requirements,
source code, header files, documentation, test files, build scripts,
etc..)[Big89, FHR91, Car90, LJ88]. Appropriately combined with a configuration
management system, hypertext also offers a structure to manage the evolution of a
developing system and its supporting component libraries[LFB89, GS90]. Thus we are
interested in integrated tools for both reuse and total software lifecycle support.

4
Several tools for C++ programming, particularly browsers and graphical aids, were
originally developed in our laboratory, and further developed by partner divisions. Of
some interest is a tool (now available as an HP product, Soft Bench C++
Developer[AB90]) that keeps several levels of abstraction compatible, allowing a library
of C++ code, header files and partially completed software to be managed consistently.
The Reuse Tools project team in HP labs is developing a workstation based reuse toolset,
using C++ and InterViews[LVC89] as a base. The initial goal of this prototype (called
Kiosk[CFG91]) is to provide a hypertext framework for manipulating (object-oriented)
libraries, such as InterViews, or other C and C++ libraries. The prototype is well
integrated with the Unix file system and Unix tools, providing us with a flexible base for
experimenting with several kinds of reuse-oriented software development and software
management environments.
Kiosk includes an import tool (called Cost++) which uses a simple declarative language
to describe and mechanically build links and nodes to help classify, catalog and structure
the relationships between several software "workproducts" (software lifecycle artifacts).
We have used Cost++ to build several different kinds of reuse library structures, on top
of InterViews and an internal HP library. We have also used Cost++ to develop a
hypertext interface to the Kiosk documentation, producing a rudimentary on-line help
system.
These classification and catalog nodes, and the original workproducts are then accessed
from a browser/editor, which provides both graphical and textual access and
editing/annotation capabilities. We have also developed a simple query language, which
is integrated with the browsing mechanism to produce a filtering/finding environment -
queries produce hypertexts which can then be further browsed, combined, pruned, or
otherwise manipulated.
We have also explored some simple IR-based mechanisms to extract keywords, to
automatically create links between interesting places in several workproducts, and to
experiment with simple automatic classification. This has also been applied to the
on-line Kiosk help/documentation browser.
User and visitor comments on the role of hypertext for component library management
and our initial Kiosk implementation are favorable. We have received several interesting
suggestions about increasing the scope of Kiosk application, such as using it to help in
managing the terminology and artifacts produced during domain analysis, or combining
Kiosk with a faceted approach.
Kiosk 1.0 was first released to HP internal users in October 1990. We have made several
improvements to the system based on user feedback, and released an updated version in
January 1991. At this point some 40 internal users have installed Kiosk. Several of the
users are exploring how Kiosk might manage workproducts other than code, such as
network management interfaces, or design fragment and decisions.
Nevertheless, we have yet to show that a hypertext-based library does, in fact, help make
reuse easier. Over the next several months, we plan to evaluate the prototype in use by
groups doing library-based reuse.

5
5 :Framework-based Application Construction
Another project team (called FAB) is exploring a "kit" building technology to produce a
series of compatible, application-oriented kits. These kits (reuse library, glue language,
application framework and supporting environment) will be used to produce distributed
applications that are easy to customize and extend, and which can be usefully and easily
integrated together (by an application developer or end user) to produce
application-oriented environments. As a first step, we have begun to explore the
structure and use of "software bus" architecturesjliyast'], distributed applications based
on a broadcast message server (SoftBench BMS[Cag90]), and several object-oriented
application frameworks.
A key feature is a common application architecture and compatible or common extension
language(s), with support tools that can be easily and effectively used by application
developers and end-users. We believe that effective kits need both an execution
environment supporting application use and customization, and a development
environment that is oriented to the reuse/glue paradigm (with visual builders, library
browsers, component interconnectors, etc.).
We believe that these execution and development environments can (and must) share
some common technology, mechanisms and philosophy. For example, systems such as
Hypercard, AutoCAD and others actually combine both environments, and use the same
scripting language (visual and textual) as both extension language and application
development language. This allows a spectrum of usage, ranging from "simple end user"
through "experienced customizer" to "advanced application package writer".
Issues explored in these projects include:

Design - domain analysis and object-oriented methods; architectural reference

models; generic framework and architectures; distribution mechanisms; broadcast
message and software bus models.
Reuse - component design and framework API's; library structure and
taxonomy(s); component interconnection mechanisms; parameter transmission;
extension mechanisms; generic "kit" technology.
User Programming - compatibility of textual and visual paradigms; "glue"
languages; generic customization languages; metaphors for how library and
extension structure reveal themselves (e.g. cut-and-paste from palette, versus
declarative 4GL code, etc).
Environments - library browsers and code-writers; hypertext linkages between
different workproducts; debugging support for glue/scripting languages; version
management.

6
6 Summary
We have several reuse related projects underway at Hewlett-Packard. We believe it
important to make progress on several areas simultaneously. In order to make systematic
reuse a significant part of the way we develop and maintain software at HP, we are
defining a comprehensive and coherently managed program, involving methodology and
technology development and transfer, assessment and training, and planned change
management. We have good management support for our program and are making
significant progress.

7 References
[AB90] Michael Armistead and John Burnham. HP C++/Softbench: A development
environment for C++. Journal of Object-Oriented Programming, 3(4):82-85,
November 1990.
[BB91] Bruce Barnes and Terry B. Bollinger. Making reuse cost-effective. IEEE
Software, 8(1):13-24, January 1991.
[BE89] Ted Biggerstaff and Alan Pedis (Eds). Software Reusability, volume 1 & 2.
ACM Press, NY, 1989.
[Big89] Ted J. Biggerstaff. Design recovery for maintenance and reuse. IEEE
Computer, pages 36-49, July 1989.
[Boe90] Barry Boehm. Trends in US environments. In Richard N. Taylor, editor,
Proceedings of Fourth ACM SIGSOFT Symposium on Software Development
Environments, Dec 3-5, 1990, Irvine, CA. ACM, ACM Press, December
1990. (Keynote address, not printed in proceedings.).
[Cag90] Martin R. Cagan. The HP Softbench environment: An architecture for a new
generation of software tools. Hewlett-Packard Journal, pages 36-47, June
1990. (See other papers in this issue.).
[Car90] Patricia Carando. Hyperbole: A retrieval-by-reformulation interface that
promotes software visibility. In Doug Lea, editor, Third Annual Workshop:
Methods and Tools for Reuse. CASE Center, Syracuse University, June 1990.
[CFG91] Michael L. Creech, Dennis F. Freeze, and Marlin L. Griss. Kiosk: A
hypertext-based software reuse tool. Technical Report SSL-TM-91-03,
Hewlett-Packard Laboratories, March 1991. (Preliminary version).
[CHSW90] Joachim Cramer, Heike Hanekens, Wilhelm Schafer, and Stefan Wolf. A
process-oriented approach to the reuse of software components. Memo Nr. 43,
University of Dortmund, Helenenbergweg 19, D-4600 Dortmund 50, FRG,
March 1990.
[FG90] William B. Frakes and P.B. Gandel. Representing reusable software.
Information and Software Technology, 32(10):653-664, December 1990.

7
[FHR91] Gerhard Fischer, Scott Henninger, and David Redmiles. Cognitive tools for
locating and comprehending software objects for reuse. Technical report,
University of Colorado, Boulder, May 1991.
[GS90] Pankaj K. Garg and W. Scacchi. A Hypertext System to Manage Software
Life Cycle Documents. IEEE Software, pages 90-98, May 1990.
[LFB89] E. Lippe, G. H. Florijn, and E. G. J. Bogaart. Architecture of a distributed
version control system. Technical Report RP/dvm-89/4, SERC, P.O. Box
424, 3500 AK Ultrecht, The Netherlands, Apri11989.
[LJ88] L. Latour and E. Johnson. SEER: A graphical retrieval system for reusable
Ada software modules. Third International IEEE Conference on Ada
Applications and Environments (Cat. No.87CH2470-9), pages 105-113, May
1988.
[LVC89] Mark A. Linton, John M. Vlissides, and Paul R. Calder. Composing user
interfaces with InterViews. IEEE Computer, pages 8-22, February 1989.
[PD90] Ruben Prieto-Diaz. Domain analysis: An introduction. Software Engineering
Notes, 15(2):47-54, April 1990.
[Rya90] Doris Ryan. RAPID/NM, Reusable Architectures for Transaction Processing
and Network Management Applications. AT&T, 1990.
['Ira88] Will Tracz. Tutorial: Software Reuse: Emerging Technology. Number IEEE
Catalog Number EH0278-2. IEEE Computer Society Press, 1988.