Axis 1 - Viewpoint engineering

Software Engineering has been adopting the viewpoint notion since the late 80s in order to represent heterogeneous entities. The upstream phases, known as requirements engineering, has been carrying out this notion to get rid of the contradictions and inconsistencies which raise from the confrontation of needs issued from different users of a shared system. The database field had already studied the questions in the framework of sharing heterogeneous data sources and partially solved the problems with a very close diversity respect principle. Object-oriented models and languages, designing methods, formal logics are the main fields which already contributed significantly to the notion of viewpoint.

But the diversity of representations is not the only aspect of a system complexity. The notion of viewpoint gives means to study the representation among other features in the global context of the complex relationship which arises within the actors in a system, during the different phases of its development, from design to implementation and maintenance. The expected profit is first, to improve the system appropriateness to all the actors: users, developers, managers; then to improve the general quality of the system: performance, development safety, evolution, operation, usages. Viewpoint engineering aims at developing designing methods, or parts of methods, which are viewpoint centered and computer aided platforms which carry these methods.

Among the questions arisen form viewpoint engineering, we have so far pointed out three of them which put directly at stake the individual and collective competences of ISYCOM research team.

Representation of viewpoints and relations between viewpoints

Version Française

The relations between the actors of a system are mediated by means of languages which depend on the position of these actors within the system, on their own activity (designer, user, developer…). The study of these relationships under the viewpoint angle allows analyzing them globally, putting forward concerns related to human computer interface. Natural language is studied using statistic text processing techniques which aim at classifying information useful to visualize shared knowledge.

Reusable multi-view objects components

Version Française

L'objectif est ici de développer une méthodologie visant à produire et réutiliser des composants d'information intégrant l'approche objet et la notion de point de vue des différents acteurs interagissant avec le système à concevoir. De tels composants peuvent être élaborés dans la conception/réalisation de systèmes d'information.

Cestravaux s'inscrivent en continuité avec le projet VBOOM réalisé il y a quelques années à l'IRIT-Enseeiht en collaboration avec l'ENSIAS de Rabat. Nous travaillons actuellement sur la définition du formalisme VUML (View based UML) qui peut être considéré comme un profil UML permettant de construire des composants multivues. Un composant multivues est une entité composée d'un classe appelée base et de vues étendant cette base selon les besoins et droits d'accès (points de vue) des acteurs du système. Nous sommes également en train d'élaborer une démarche d'analyse/conception fondée sur l'utilisation de VUML.

Un autre aspect de ce travail concerne la réutilisation de composants multi-vues sous forme de patrons. Nous avons ainsi pour objectif de définir une bibliothèque de composants multivues réutilisables.

Ce travail sera concrétisé à moyen terme par la réalisation d'un outil supportant le profil VUML et la démarche définie.

Il est mené en collaboration avec les partenaires du réseau STIC franco-marocain en Génie Logiciel.

Viewpoint centered requirements engineering

Version Française

The upstream phase of a system designing project is the first opportunity when multiple viewpoints which may be contradictory, inconsistent, must be studied to feed the next specification phase.

We are concerned with multi-viewpoints requirements elicitation process methods and tools in software engineering and system engineering. Requirements elicitation process engineering aims at describing, validating and controlling the upstream phases of software design process. Computer aided requirements capture, analysis, formalization relies upon modeling shared knowledge, and the study of their possible inconsistencies and incompleteness. We point out 3 features:

1. integration of the multi viewpoint requirements elicitation process to the modeling of software development process and further system development process in the framework of software engineering methods and system engineering methods;
2. representation of viewpoints and the relations between viewpoints which are first expressed in natural language and then translated in suitable languages, object or agent centered;
3. management of the requirements elicitation process, by means of a semiotic analysis, by means of several tools: (1) visualization tools issued form statistical and text process engineering; (2) singularity retrieval from knowledge based systems; (3) reasoning formalization.

Thus our research leans upon other disciplines than computer science, in particular human and social sciences: the real requirements are often expressed in natural language by individuals or groups of people whose discourses are not homogeneous. Linguistics, semiotics, sociology of organizations bring up methodologies to study these questions. So far, we have been lead to identify common features between several levels, a priori not reducible to each other, where the viewpoint notion can be pointed out, and to elaborate computerized models. At present, we must identify the elements which remain irreducible between these different levels, and elaborate a methodological support in order to manage this irreducibility.

1. CORINTHE project joins ISYCOM research team and LEREPS research laboratory in economics. It aims at developing a strategic watching system for a given field of economy. It relies on information mining techniques issued form heterogeneous electronic sources. The system will be elaborated starting from the viewpoint notion. It will also rely on the expertise of LEREPS laboratory and the SISIFE database which gather strategic information about energy in Europe.
   
   
2. VOLYNE project with the Japanese Iwate University and Computer Based Software Methodology Institute of Tokyo offered in 2002 2003 an industrial field which allow to elaborate an engineering viewpoint oriented method of the upstream phase of the designing project of a software system. This action aimed at adding a computer aided multi viewpoint engineering method to Lyee software generation method. Lyee project joined together universities from Japan, Canada, and several European countries.
   
   
3. The Academic-Industry working group on requirements engineering of the CNRT-AE (National Centre of Research and Technology Aeronautics and Space). In 2002 2003, the group aimed at identifying industrial needs in the domain of aerospace in the requirements engineering field and the ability of research laboratories to solve these needs. Research axis issued from the group will be used French and European calls for projects.
   
   
4. The ACI (Incitation Joint Action) interdisciplinary work in human and social sciences. This action is conducted by University Toulouse 1 and it is supported by the Ministry of Research and is granted with a PhD scholarship. For the computer science field, this action is related to the auto-organization in multi-agent systems. It provides a practical field for experiments. It will give the opportunity to an operational definition of the concept of organization, and to the elaboration of a multi-agent architecture adapted to the social system simulation using agents.
   
   
5. Des expériences réalisables in vitro avec des populations d'étudiants à propos d'études de cas seront organisées à partir de la rentrée 2004 dans des enseignements de Génie Logiciel. Le point de départ sera le cours Modèles d'ingénierie du processus de développement du DEA Informatique/Master recherche SCLP de Toulouse.