<p>Most software is now developed by professional programmers, and the millions of users of “shrink-wrapped” applications never meet the developers. The skills of professional programmers are obviously required in many software development projects. However, there are also reasons, ranging from practical and economical to philosophical and educational, why people without this background may want to develop their own applications or tailor applications to their own use.</p><p>The general purpose programming languages and development tools used by professional programmers are not suitable for people outside this profession. Instead, people outside this profession need more high-level tools that allow them to express solutions using domain and task specific terms. Despite the accumulated experience from such environments within the end-user programming community, creating a new application development environment for a domain or task remains a large development task, and can be too time-consuming and costly to be found worthwhile.</p><p>This thesis presents an approach to reducing the effort needed in developing extensible application composition environments for end users. We use the term “end user” in the same way as the end-user programming community, to denote a person who wants to develop an application or tailor an application to a specific use, without needing professional programming skills. The work in the thesis was initiated based on experience from development of proof-of-concept implementations of such environments in two European research projects. With this background, a set of user and developer roles are identified and organised into a value chain for development of such environments. Further, the approach combines the research areas of component-based software engineering and end-user programming to propose an overall architecture, component frameworks and tools suitable to support development of such environments. The validity of the results is established through partial prototyping of the frameworks and tools, combined with qualitative analyses of how these can be fully implemented and of their suitability for their intended purpose.</p><p>The main contributions of this thesis are:</p><p>• a value chain for development of application composition environments, with identification and description of tasks of each of the identified developer roles;</p><p>• an overall architecture for developing extensible application composition environments based on component frameworks;</p><p>• architecture of two component frameworks defining mechanisms and rules of behaviour for components ensuring extensibility at runtime and edit time;</p><p>• definition and description of a set of UML stereotypes for modelling domain frameworks based on the component frameworks, and a mapping to Java/JavaBeans allowing a code generator tool to produce part of the implementation;</p><p>• definition and description of tools which use the model of a domain framework as input and partially transform the work of creating composition environments and editors for domain objects from a programming task to a configuration task.</p>

Most software is now developed by professional programmers, and the millions of users of “shrink-wrapped” applications never meet the developers. The skills of professional programmers are obviously required in many software development projects. However, there are also reasons, ranging from practical and economical to philosophical and educational, why people without this background may want to develop their own applications or tailor applications to their own use. The general purpose programming languages and development tools used by professional programmers are not suitable for people outside this profession. Instead, people outside this profession need more high-level tools that allow them to express solutions using domain and task specific terms. Despite the accumulated experience from such environments within the end-user programming community, creating a new application development environment for a domain or task remains a large development task, and can be too time-consuming and costly to be found worthwhile. This thesis presents an approach to reducing the effort needed in developing extensible application composition environments for end users. We use the term “end user” in the same way as the end-user programming community, to denote a person who wants to develop an application or tailor an application to a specific use, without needing professional programming skills. The work in the thesis was initiated based on experience from development of proof-of-concept implementations of such environments in two European research projects. With this background, a set of user and developer roles are identified and organised into a value chain for development of such environments. Further, the approach combines the research areas of component-based software engineering and end-user programming to propose an overall architecture, component frameworks and tools suitable to support development of such environments. The validity of the results is established through partial prototyping of the frameworks and tools, combined with qualitative analyses of how these can be fully implemented and of their suitability for their intended purpose. The main contributions of this thesis are: • a value chain for development of application composition environments, with identification and description of tasks of each of the identified developer roles; • an overall architecture for developing extensible application composition environments based on component frameworks; • architecture of two component frameworks defining mechanisms and rules of behaviour for components ensuring extensibility at runtime and edit time; • definition and description of a set of UML stereotypes for modelling domain frameworks based on the component frameworks, and a mapping to Java/JavaBeans allowing a code generator tool to produce part of the implementation; • definition and description of tools which use the model of a domain framework as input and partially transform the work of creating composition environments and editors for domain objects from a programming task to a configuration task.

Force est de constater que composer des applications existantes afin d’en réutiliser tout ou une partie est une tâche complexe. Pourtant avec l’apparition quotidienne d’applications, les éditeurs d’applications ont de plus en plus besoin d’effectuer de telles compositions pour répondre à la demande croissante des utilisateurs. Les travaux existants ne traitent généralement que d’un seul point de vue : celui du "Noyau Fonctionnel" dans le domaine du Génie Logiciel, celui des Tâches" ou celui de l’"Interface Graphique" dans le domaine des Interactions Homme-Machine (IHM). Cette thèse propose une nouvelle approche basée sur un modèle d’application complet (fonctionnel, tâche et interface graphique). Elle permet à un utilisateur de naviguer entre ces différents modèles pour sélectionner des ensembles cohérents pouvant être composé par substitution. Une implémentation de cette approche a permis d’effectuer des tests utilisateurs confortant les bienfaits d’une modélisation complète. / One has to note that composing existing applications by completely or partly reusing them is a complex task. Nevertheless with the daily appearance of new available applications on any media, the application editors need to perform such compositions more and more to answer the increasing users’ requests. Modeling an application for composition or just determining by which point of view on applications make this composition is not easy. Works exist, but generally deal or ensue from only a single point of view : the "Functional Core" point of view in Software Engineering field, the "Task" one or "User Interface" one in Human Computer Interaction (HCI) field. This thesis defines a new approach based on a complete application model (functionnal, task and user interface). It enables an user to navigate between those different models in order to select consistent sets. These last ones are composable by substitution. An implementation of this approach was used to perform user tests whose results consolidate benefits of a complete model.

Abstract Application composition refers to the creation of applications by using Web Services and resources such as mobile devices, displays, and various augmented everyday objects as building blocks. This approach is especially useful for ubiquitous applications which focus on supporting users' needs and everyday activities. This thesis proposes to compose these applications by choosing the appropriate set of resources and services and their configuration as dictated by the users themselves, together with their needs and other contexts. This thesis studies automated and interactive application composition. The first approach enables the system to act upon users' needs and intentions, while the second enables users themselves to specify their needs and achieve their goals. The research was carried out by designing, implementing and evaluating eight prototypes for automated and interactive application composition. The evaluation methods used included performance analysis and user experiments. The main results include (1) the design and implementation of automated composition mechanisms which rely on evolutionary computing and genetic algorithms, (2) a detailed performance analysis of these mechanisms using synthesized datasets and in a real networking environment; (3) the design and implementation of interactive application composition prototypes which rely on graphical and physical (i.e. touch-based) user interfaces for mobile devices and utilize various contexts; (4) an evaluation of these prototypes in a series of user experiments. The evaluation studied the following issues: users' attitudes towards an automated composition system which makes decisions on their behalf; users' attitudes towards a context-aware composition interface; and the issue of balancing user control and system autonomy. / Tiivistelmä Sovellusten kokoamisella tarkoitetaan sovellusten luomista käyttäen rakennuselementteinä Web Services -ohjelmistojärjestelmää, sekä muita resursseja, kuten mobiililaitteita, näyttölaitteita ja monenlaisia lisävarusteltuja tavanomaisia esineitä. Tämä menetelmä on erityisen käyttökelpoinen sellaisten jokapaikan tietoteknisten sovellusten luomisessa, joiden tarkoitus on tukea käyttäjän tarpeita ja päivittäisiä toimia. Tutkimuksessa esitetään, että tällaisia sovelluksia voidaan koota valitsemalla sopivia palvelu- ja resurssipaketin sekä niiden konfiguraation käyttäjien vaatimusten mukaan, ottaen huomioon heidän tarpeensa sekä muut kontekstit. Tässä väitöskirjassa tutkitaan automatisoitua ja vuorovaikutteista sovellusten kokoamista. Automatisointi mahdollistaa sen, että järjestelmä toimii käyttäjän tarpeiden ja päämäärien mukaisesti, kun taas vuorovaikutteisuuden ansiosta käyttäjä voi määritellä tarpeensa ja pääsee näin haluamaansa lopputulokseen. Tutkimus toteutettiin suunnittelemalla, toteuttamalla ja arvioimalla kahdeksan automatisoidun ja vuorovaikutteisen sovelluksen kokoonpanon prototyyppiä. Arviointimetodeina käytettiin suorituskykyanalyysiä ja käyttäjäkokeita. Tutkimuksen tärkeimpiä tuloksia ovat (1) evoluutiolaskentaan ja geneettisiin algoritmeihin perustuvien automatisoitujen kokoamismekanismien suunnittelu ja toteutus; (2) näiden mekanismien yksityiskohtainen suorituskykyanalyysi syntetisoitujen tietoaineistojen ja todellisten verkkoympäristöjen avulla; (3) graafisille ja fyysisille (l. kosketus) mobiililaitteiden käyttöliittymille perustuvien ja monenlaisia konteksteja hyödyntävien vuorovaikutteisten kokoamissovellusten prototyyppien suunnittelu ja toteutus; (4) prototyyppien arviointi erilaisin käyttäjäkokein. Arvioinnissa tutkittiin käyttäjien asenteita käyttäjän puolesta päätöksiä tekevää automatisoitua kokoamisjärjestelmää kohtaan, käyttäjien asenteita kontekstitietoista kokoamiskäyttöliittymää kohtaan, sekä käyttäjälle jäävän kontrollin ja järjestelmän autonomian välistä tasapainoa.