FINE-GRAINED INTEGRATED MANAGEMENT OF SOFTWARE
CONFIGURATIONS AND TRACEABILITY RELATIONS
Pietro Colombo
1
, Vieri del Bianco
1
and Luigi Lavazza
1,2
1
Dipartimento di Informatica e Comunicazione, Università dell’Insubria, Via Mazzini, 5 – 21100 Varese, Italy
2
CEFRIEL, Via Fucini 2 – 20133 Milano, Italy
Keywords: Model versioning, XML.
Abstract: Software Configuration Management is essential to manage the evolution of non trivial software systems.
Requirements for SCM support are continuously growing, demanding for seamless integration with the
traditional development tools and support for management activities like change management, change
impact analysis, etc. This paper presents SCCS-XP, a SCM platform supporting change management and
traceability among fine-grained software artifacts. SCCS-XP exploits a XML-based language to represent
versioned software artifacts. SCCS-XP provides the basic capabilities to build full-fledged SCM
environments featuring traceability management, change management and integrates nicely with
development tools.
1 INTRODUCTION
Software Configuration Management (SCM) is the
discipline of managing the evolution of large and
complex software systems (Tichy, 1988). SCM
provides both functionalities that assist developers in
performing coordinated changes to software
products and change management (CM) services,
like impact analysis. In the case of impact analysis
the traceability relations are used to determine the
consequences of changes in requirements or in other
software artifacts. The level of support to the
configuration and change management activities can
be classified with respect to the following issues:
Abstraction. Tools should be as orthogonal as
possible to the structure and contents of the
managed items, as well as to the relations
involving items.
Granularity. The platform has to be able to deal
with composition hierarchies. Since we do not
know a priori the requirements for the SCM and
CM system, the platform should treat in a uniform
way all the elements of the hierarchy.
Automation. In order to be effective, SCM
activities must be largely supported by automated
functionalities. Besides, since software
development typically exploits a variety of
development tools, it is necessary that the SCM
system exchanges data seamlessly with such tools
.
The proposed platform aims at supporting
version and configuration management, as well as
the typical change management activities. We do not
intend to provide a full-fledged SCM and/or change
management environment, but a platform with the
functionalities needed to build such environment.
We identified the following requirements for the
platform:
Flexibility. The platform has to support the
building of different functionalities, strategies, and
policies. No constraint is expressed on the model,
methodology or process; different tool builders
can build their own tool by implementing the
features needed in each specific environment.
The platform has to be relatively light-weight. It
has to exploit an existing DBMS to focus on the
provision of functionalities specifically oriented to
SCM and CM. Moreover, it has to be based on a
standard DBMS API, in order to exploit any
available implementation, thus increasing
portability and interoperability.
The platform has to support a development
environment where SCM and change management
functions are effectively integrated with the
development tools.
It can be noted that several of the properties
mentioned above call for a representation of artifacts
that is standard, precise, detailed, and supported by
data management systems. From this point of view
159
Colombo P., del Bianco V. and Lavazza L. (2008).
FINE-GRAINED INTEGRATED MANAGEMENT OF SOFTWARE CONFIGURATIONS AND TRACEABILITY RELATIONS.
In Proceedings of the Third International Conference on Software and Data Technologies - SE/GSDCA/MUSE, pages 159-164
DOI: 10.5220/0001883501590164
Copyright
c
SciTePress
the technology is mature to build the desired SCM
platform. XML perfectly fits the requirements since
it enables the representation of information at the
required granularity level. XML was chosen both to
represent and to exchange data with the development
tools. Our platform is thus called SCCS-XP
(Software Change and Configuration System XML
Platform).
The paper is organized as follows: Section 2
presents the conceptual organization and
implementation of SCCS-XP. Section 3 presents a
set of case studies where SCCS-XP is used to build
SCM tools. Section 4 accounts for related work.
Section 5 presents some conclusions and illustrates
future work.
2 SCCS-XP: CONCEPTS AND
IMPLEMENTATION
In this section we define the architecture of SCCS-
XP.
2.1 Conceptual Organization
The proposed architecture is conceptually organized
in an Abstract layer, featuring an Abstract data
model, a Uniform layer, featuring the Uniform data
model, and a Model layer, featuring Specific data
models (see Figure 1).
Conceptually the Abstract layer is simple, as it
corresponds to a DBMS featuring a standard query
language: the Abstract data model corresponds to
the logic data model of the adopted DBMS. The
functions provided by the Abstract layer do not
make distinction between different types of artifacts.
A relevant requirement for the abstract data layer
is that it must be easy to represent the data belonging
to the Uniform layer by means of the Abstract data
model. Another requirement is that it has to provide
some sort of aggregation mechanism, in order to
represent data at the suitable granularity level.
The Uniform layer supports the evolution
process, with particular reference to versioning,
traceability and change management.
The Uniform data model defines the concept of
artifact on the basis of the Abstract data model; in
particular it exploits the aggregation mechanism to
provide access at different granularity levels. The
Uniform layer can then be regarded as organized
internally in two sub-layers: the first one exploits the
Abstract layer to make artifacts versionable, while
the second sub-layer provides functionalities to
support traceability and change management, which
operate on versioned artifacts (or on their parts).
We decided to build the Versioned DBMS
(VDBMS) exploiting a XML-based DBMS. The
language employed to internally represent versioned
artifacts is based on XML (see Section 2.3). Since
the activities carried out at the upper layers are
typically supported by existing tools it is essential
that they interface with the platform: SCCS-XP
supports loose integration at the data level by means
of dynamically generated Virtual Files (Leblang,
1994). Translators have to be developed to convert
the tools’ specific formats into the format supported
by SCCS-XP. Virtual Files do not exist as files in
the SCCS-XP platform: actual files are produced
upon request by composing the information
available in the platform. This guarantees openness
and scalability of the platform.
The Specific layer includes the tools that are
dedicated to the crafting of the different kinds of
artifacts that are created and managed throughout the
software lifecycle. These tools employ different data
models to represent the data that have to be
exchanged with the kernel of the platform.
Repository
Query engine
DBMS
Versioning
Virtual
files
VDBMS
Change
management
Traceability
management
CASE tool
IDE
. . .
. . .
Abstract layer
Uniform layer
Specific layer
Figure 1: SCCS-XP Conceptual Architecture.
The organization of SCCS-XP requires that the
data managed by the external tools can be converted
(through Virtual Files) into/from data conformant to
the Uniform data model. The conversion must be
possible for every data format supported by the tools
and subject to versioning, and involved in
traceability and control management. When XML is
not directly supported, translators can be used.
2.2 The Versioning Model
The versioning model supported by SCCS-XP is
traditional. According to the classification proposed
by Conradi and Westfechtel (1998), our model
features symmetric deltas, and directly implements
extensional versioning. Intensional versioning can
ICSOFT 2008 - International Conference on Software and Data Technologies
160
also be achieved, by means of the querying
capabilities of the Versioned DBMS.
As already mentioned, SCCS-XP aims at
managing XML artifacts. Since an XML tag could
represent a versioned item, we need to version single
tags. This is achieved by coupling the product model
and the versioning model.
Custom relations can be defined by means of
suitable plug-ins (see Section 2.4). Specialized
relations among versioned artifacts are defined by
means of XML elements belonging to the Uniform
data model.
Figure 2: Artifact composition and fine-grained
versioning.
2.3 Internals of the Versioned DBMS
In order to make the implementation of the platform
practical, we needed to make the data models as
uniform as possible. For this purpose we defined a
meta-model that can be used to define the various
data models used in the platform. In this way, all the
data models share a common layer; moreover their
definition, correspondences, transformations, etc. are
precisely defined in the meta-model. In particular,
the meta-model has to support data definition, data
exchange procedures, data persistency and querying.
XML provides the most comprehensive support for
the aforementioned features, thus it was employed
for the Abstract layer.
Having adopted XML for the Abstract layer, we
had to make XML data versionable. For this
purpose, we defined a XML-based language named
VerML (Versioned XML), which supports the
representation of the versioning information
contained in the models illustrated in Section 2.2.
2.4 Implementation
The implementation of the Abstract layer of SCCS-
XP is based on an XML DBMS. Versioning of XML
tags was implemented according to the models
presented in Section 2.2. For this purpose we
adopted the following strategy:
XML data are stored exploiting the functionality
of the DBMS.
It is possible to decompose a piece of XML data
into individual XML tags that are stored
separately, thus implementing a fine-grained
repository.
Information concerning the composition of
documents is contained in VerML data, which is
stored in the DBMS as well.
Different versions of the same artifact, possibly
belonging to a composite artifact, are individually
stored.
Versioning information is also stored in VerML
data. For instance, consider the following
situation: class C contains attribute A and method
M, and there are two versions of method M. The
database contains artifacts representing attribute
A, version 1 of method M, version 2 of M, version
1 of C and version 2 of C. Each version of C
contains a VerML tag pointing to the correct
version of M.
The Abstract layer of SCCS-XP implements the
operations of the VDBMS (see Figure 1): creation of
a new versioned artifact, insertion of a new version,
extraction of a given version of a (possibly
composed) artifact, etc. Basic check-in and check-
out services were implemented as part of the
Abstract layer. They are accessible via a simple
command line interface. The operations can involve
any XML tag individually stored; therefore it is
possible to extract pieces of a document. In this way
the Abstract layer provides the Uniform layer with
the flexibility needed to build the Uniform data
model required by the user.
Locking is currently not implemented; SCCS-XP
does not manage safe concurrent access to the
versioned data.
Items stored in the database are manipulated by
means of queries. The querying capabilities were
implemented using XPath (Clark and DeRose,
1999); the more powerful XQuery (Boag et al.,
2007) will be considered for future improvements of
SCCS-XP XML query engine. The implementation
of the Uniform layer is based on the modular plug-in
architecture described in Figure 3.
The Uniform data model is obtained by means of
suitable plug-ins that define data and operations on
the basis of the abstract data layer. For instance, we
developed the Java plug-in, which is able to parse
Java files and translate them into XML data, which
can then be versioned and stored by means of the
services provided by the Abstract layer. The Java
v1 1
v1 2
v1 3
v1 4
v1 5
v21
v22
v1 1
v1 2
v1 3
v1 4
v1 5
v21
v22
v1 1
v1 2
v1 3
v1 4
v21
v3 1
v11
v12
Artifact X
Artifact Y
Artifact Z
Vers io n v12
of artifact X
Y
Z
W
T
Y
Z
W
T
Y
Z
W
T
Vers ion v14
o f a rt ifa c t X
U
v1 1 Artifact U
Versioned item Non versioned item
fo r k
merge
branc h
FINE-GRAINED INTEGRATED MANAGEMENT OF SOFTWARE CONFIGURATIONS AND TRACEABILITY
RELATIONS
161
plug-in can also retrieve a given version of a Java
XML artifact and un-parse it, generating a Java text
file that can be read by any suitable development
tool. In order to manage requirements expressed in
natural language, we developed a plug-in, named
SimpleReq, which parses LaTeX files and builds a
XML hierarchical representation of the requirements
contained in LaTeX files.
XML database
Abstract layer
Event bus
Plug-in 1Plug-in 1
Plug-in 2
SCM system CLI
Data file
Data file
Development
tools
Development
tools
Development
tools
Figure 3: Architecture of SCCS-XP based SCM tools.
A difficulty in managing data exchange with
external tools consists in understanding which parts
of a given document correspond to elements already
stored in the platform and which parts are new, thus
requiring the creation of new sub-element versions.
For this purpose it is possible to insert annotations in
the files exported from SCCS-XP. For instance the
Java plug-in exports a source code file containing
annotations, so that when editing it, the user can
recognize the annotations and preserves them; when
the plug-in re-imports the modified file, it can
recognize annotations as marking parts
corresponding to existing artifacts. A plug-in could
also exploit the XML diff techniques (Mouat, 2002).
Besides managing the conversion of data
between the format understood by external tools and
the format supported by the Abstract layer, the plug-
in can be employed for more complex goals. The
StructureX plug-in introduces new artifact types, and
uses them to correlate the versioned artifacts with an
external hierarchical file system (see Section 3). The
new artifacts representing files, directories, etc. are
managed by the Abstract layer in the usual way, but
are not made visible to the user, who interacts with
these objects only through a suitable interface.
The more complex plug-in named RelationX (see
Section 3) reacts to artifact changes in order to
maintain referential integrity. For this purpose the
plug-in is connected to the event bus: it receives the
relevant events from the Abstract layer and triggers
operations that update the links between artifacts
that are defined by the plug-in itself.
Every plug-in can add or modify the commands
provided by the command line interface (CLI)
originally defined by the Abstract layer. During the
usage of the system, the user can declare a context
corresponding to a plug-in: this determines which
commands are executed in case of conflicts.
3 VALIDATION
In order to assess the validity of SCCS-XP we
employed it in three different applications, each
conceived to exercise a specific set of features of the
platform.
The first application consisted in the
development of a basic single user SCM tool “a la
CVS” (XVS). The goal was to assess the ability of
SCCS-XP with respect to effective and efficient
management of data of realistic size. XVS features
the usual functionality: check-in, check-out,
branching, etc. XVS uses the Java plug-in to version
fragments of source code; in particular, class and
methods are versioned as separate artifacts.
SCCS-XP was then employed to develop a
second application, with the purpose of showing that
the platform can be used to build a proper
configuration management tool. The main goal of
this tool, called XCM, was to provide the user with
the possibility of managing configurations consisting
of coherent sets of file versions. For this purpose the
concept of structured project was introduced. Such
concept is implemented by means of a new artifact,
which is treated by the Abstract layer as any other
artifact. We developed a specific plug-in (called
StructureX) to deal with the specificity of the
concept of project: new versions of a project are
created whenever files or directories are added,
changed or removed from the project. StructureX
interprets project artifacts and correlates them to the
other artifacts and the underlying file system.
Thanks to StructureX, the concept of artifact evolves
independently from the hierarchical structure (the
file system) which it belongs to. StructureX can be
reused in different contexts, for instance it can be
used to enhance XVS, providing the user with the
possibility of versioning both Java files and projects
containing such files. In fact, we actually added
StructureX to XVS (obtaining XCM). Consider that
StructureX employs the Abstract layer in order to
define an enhancement of the Uniform layer data
model, in which the concepts of project, directory
and files are defined (see Figure 1).
Finally, a third plug-in named RelationX, was
developed to manage the semantic relations among
artifacts. In particular, we considered the typical
traceability relations that link requirements
specification documents (managed by the
SimpleReq plug-in) to the source code, and the
relations that link the code with unit test cases. The
ICSOFT 2008 - International Conference on Software and Data Technologies
162
goal of the application was to perform impact
analysis on the basis of the known relations among
artifacts. The application handles three types of
relations:
Relations already present in the data managed at
the level of the Abstract layer. For instance, the
composition relationships and the derivation and
branch relationship.
Relations that have to be specified explicitly by
the user: for instance, the relations that link a
requirement to the classes that implement it. These
relations are represented extensionally.
Finally, some relations can be inferred from the
information normally present in the repository.
For instance, the dependence of a method M from
a given requirement R may be inferred by the fact
that M belongs to class C, and class C implements
R. In order to ease the management of relations,
they are expressed intensionally by rules that the
plug-in is able to interpret, thus freeing the user
from explicitly entering them.
The RelationX plug-in enhances the Uniform
layer data model with the aforementioned types of
relations. This plug-in is able, given a (versioned)
artifact, to exploit the knowledge of the relations for
finding the set of related (versioned) artifacts. This
ability can be used to reason about dependencies
among artifacts.
4 RELATED WORK
The Software Configuration Roadmap (Estublier,
2000) reports several problems that need to be
addressed when developing Configuration
Management tools. SCCS-XP fully addresses some
of the proposed issues since:
It provides a flexible data model that can be
used to represent complex data as well as data
involving any user-defined relationship.
It guarantees the interoperability with SE tools
by supporting XML based data exchange.
Scalability and efficiency are provided to the
extent supported by the underlying database.
It supports the definition of any relevant
attribute as an XML tag; then appropriate
queries can be used to extract and combine data
according to the attributes’ values.
It provides mechanisms to extract the data and
put them in the required format, place, etc.
Actually this is done by means of the plug-in
mechanism mentioned in Section 2.4.
It is an extensible platform that can be enhanced
in several respects. For instance, it is possible to
plug-in the definition of new data models, to
add support for impact analysis or for
exchanging data with external tools, or for
consistency check, and so on. SCCS-XP is
actually open to incremental customization.
Other research efforts comply, to some extent,
with the indications provided by the roadmap.
Among the many initiatives, UVM (Westfechtel,
Munch and Conradi, 2001) is probably one of the
most conceptually advanced. UVM features uniform
version model and support architecture for SCM.
UVM provides a base model that can be employed
to express specific version models. The underlying
layered architecture is completely orthogonal to the
data model used for representing software artifacts
and their relationships. In UVM, version rules are
placed at the bottom of the layered architecture and
can be used for expressing different version models.
With respect to UVM, our approach is more
lightweight. SCCS-XP is not perfectly orthogonal
with respect to the data model (which is XML),
nevertheless, it employs XML as a meta language to
define Specific data models. The result is that SCCS-
XP is largely independent from the data model used
for representing software artifacts.
Our approach shares some characteristics with
NUCM (van der Hoek et al., 2002): both provide a
quite generic repository, upon which several
different SCM policies can be built. This is
fundamental to free users form the responsibility of
managing issues that can be dealt with at the
platform level. MCCM (van der Lingen and van der
Hoek, 2004) is a platform featuring a repository to
which pluggable components can be added. MCCM,
like NUCM, manages data at the level of files or file
collections and features a distributed repository. On
the contrary, SCCS-XP manages data at finer
granularities, but does not manage distribution issues
at all. Another relevant difference is that MCCM and
NUCM feature a very powerful repository, so that a
concrete SCM tool is obtained essentially by
constraining the way data are managed by the
repository itself. SCCS-XP allows plug-ins to extend
the data model and the functionality provided by a
lightweight repository.
Other systems, like Coven (Chu-Carroll and
Sprenkle, 2000), address the granularity of artifacts.
However, these are specialized systems that exploit
document structuring for their very specific
objectives. Coven manages fragments of source code
corresponding to project elements in order to support
collaboration in an environment integrating SCM
and collaborative development.
EPOS (Conradi et al., 1995) is one of the best
known process support environment. In order to
effectively support the development process, EPOS
consistently manages the software artifacts, thus
featuring integrated process and product
FINE-GRAINED INTEGRATED MANAGEMENT OF SOFTWARE CONFIGURATIONS AND TRACEABILITY
RELATIONS
163
management. The EPOS repository supports
versioning and long transactions. The principles that
inspired the creation of the EPOS repository are
similar to the principle underlying SCCS-XP, in that
both approaches deal with the management of
complex structured artifacts, and tend to support a
smooth and effective integration with the
development environment. There are also important
differences: EPOS uses a different versioning model
(Change Oriented Versioning), and does not support
fine-grained versioning. SCCS-XP does not provide
native support for the software process, although in
principle it would be possible to exploit the event-
based architecture of SCCS-XP to build a process
support plug-in.
Finally, the COOP/Orm project developed a
collaborative SCM system integrated with a
programming environment. COOP/Orm features a
fine-grained versioning model (Magnusson and
Asklund, 1996) similar to the one featured by SCCS-
XP. COOP/Orm’s approach to SCM suffers by some
constraints: COOP/Orm comes with a built-in
optimistic check-out mechanism with synchronous
updates, it obliges the usage of the integrated editor,
and it does not exchange data easily with external
development tools.
5 CONCLUSIONS
SCCS-XP is a light-weight, XML-based platform
providing basic SCM functionalities. SCCS-XP
supports a data model which is actually a sort of
meta-model that can be used to build support for
several different types of software artifacts;
integration, at the data level, with external tools is
easy. A traditional, extensional versioning model is
provided, but SCCS-XP provides the possibility to
extend this model, as well as to introduce additional
functionality. SCCS-XP was employed to realize
prototypes of SCM tools ranging form a simple
CVS-like tool to a sophisticated tool that can
manage semantic relations, and perform different
types of impact analysis.
In conclusion, we believe that SCCS-XP can be
employed to build customized SCM environments
suitable to support modern software development
practices.
Future work will concern: 1) the development of
a plug-in for UML, supporting the XMI data model;
2) the implementation of a locking mechanism 3) the
implementation of a plug-in that allows the user to
define new specific relations, and to define rules that
can be triggered by specific events and invoke
suitable actions; 4) the development of a plug-in that
exploits the virtual files mechanism to supports
different views of a single project.
ACKNOWLEDGEMENTS
We would like to thank Luca Ridolfi and Riccardo
Serafin for their contribution in the design and
implementation of the tool.
REFERENCES
Boag, S., Chamberlin, D., Fernández, M. F., Florescu, D.,
Robie, J., Siméon, J., 2007. XQuery 1.0: An XML
Query Language. W3C Recommendation.
Chu-Carroll M.C. and Sprenkle S., 2002. Coven: Brewing
Better Collaboration through Software Configuration
Management. In FSE 2000, San Diego.
Clark, J., DeRose, S., 1999. XML Path Language (XPath)
Version 1.0. W3C Recommendation.
Conradi R., Westfechtel B., 1998. Version Models for
Software Configuration Management. In ACM
Computing Surveys, Vol. 30, N. 2, pp. 232-282.
Conradi, R., Larsen, J., Nguyen, M.N., Munch, B.P.,
Westby, P.H., 1995. Integrated Product and Process
Management in EPOS. Journal of Integrated CAE.
Estublier, J., 2000. Software configuration management: a
roadmap. In ICSE - Future of SE Track.
Leblang, D. B., 1994. The CM challenge: Configuration
management that works. In Configuration
Management, Vol. 2 of Trends in Software, Wiley,
pp. 1-37.
Magnusson, B., Asklund, U., 1996. Fine grained version
control of configurations in COOP/Orm. In ICSE
1996, SCM-6 Workshop.
Mouat, A., 2002. XML Diff and Patch Utilities. CS4
Dissertation, Heriot-Watt University,.
Tichy, W. F., 1988. Tools for software configuration
management. In Proc. of the Int. Workshop on
Software Version and Configuration Control,
Grassau, Germany, Teubner Verlag.
van der Hoek, A., Carzaniga, A., Heimbigner, D., Wolf,
A. L., 2002. A Testbed for Configuration
Management Policy Programming. IEEE TSE,
Volume 28, Issue 1, pp. 79 - 99.
van der Lingen, R., van der Hoek, A., 2004. An
Experimental, Pluggable Infrastructure for Modular
Configuration Management Policy Composition. In
ICSE 2004, Edinburgh.
Westfechtel B., Munch B. P., Conradi R., 2001. A
Layered Architecture for Uniform Version
Management. IEEE TSE. Volume 27, Issue 12.
ICSOFT 2008 - International Conference on Software and Data Technologies
164