Exception handling suffers from a fluke in its evolution. Some time in the 1970's, a researcher called Goodenough introduced the exception handling mechanism, known today as the tcode{try...catch} construct. At about the same time, two fellows called Parnas and Würges published a paper about "undesired events". This paper appears forgotten. The funny thing is, Parnas and Würges effectively described how to emph{use} exception handling mechanisms. There is a need to respond to this undesired event. Exception handling suffers from lack of design guidelines and a number of inconsistencies with the object-oriented paradigm, among other things. The thesis' main contribution is creating a library of exception challenges and the evaluation of safety facades, an approach that introduce an architecture and guidelines for designing exception handling. Through qualitative evaluation, this thesis shows how safety facades form an interesting new approach to exception handling.

When proponents of open source software are asked to explain the success of their movement they typically point to the quality of the software produced, which is in turn attributed to the rather unconventional development model of releasing unfinished versions of the software and having users look over the code and report and fix bugs. This thesis investigates the open source quality assurance model from a knowledge management perspective – based on the assumption that debugging involves a high degree of knowledge work. By doing interpretive case studies of two open source projects – using direct observation, e-mail archives, and bug-trackers as data sources – I present descriptive accounts of the day to day quality practices in open source development. The analysis shows that conceptualizing and classifying bugs is a complex process involving sense-making and subjective considerations; that the peer-review process in open source projects has a lot in common with traditional field-testing; and that communication tools and mediums are used interchangeably, but with certain preferences depending on subject matter. I conclude that perhaps the success of the open source development model is not due to its novelty compared to traditional software engineering, but because open source developers have recognized that debugging is a knowledge-intensive process. Keywords: Open Source, Software Quality, Knowledge Management

As our reliance on software has increased, robustness has become an important subject. Software that is not robust enough may lead to frustration, or loss of time or value. Software architecture forms the main structures of applications. Having focus on the quality of an applications architecture may increase the robustness of the application. This thesis tries to find a suitable architectural process for achieving robustness. I report on the findings from ten interviews with software architects from the software industry, around the theme robustness. Interview results are used in order to form a definition of robustness which is wider than the definitions I found in literature. The thesis's main contribution is a proposed process for designing and analyzing robust software architectures that make use of elements from existing methods. The proposed process is grounded on results from interviews, personal experience, and evaluation of existing methods in literature.

Agile software development have in recent years been widely accepted in industry, as well as being the target of much research. XP has been the main focus, while there exist relatively few studies of other Agile methods such as Scrum. This thesis describe an ethnographic study of a Scrum team in a project. The goal is to give a rich description of the use and application of Scrum. Special attention is given to the implications of differences in implementation from theory. Another focus is to reveal the dynamics of teamwork within the project. The main findings are that Scrum was easy to implement, and worked well, but is challenging when used to increase predictability. With regard to teamwork, Scrum supported a shared mental, communication and adaptability. The leadership function is however complex, and requires good interpersonal skills. Keywords: Agile, Scrum, Teamwork, Implementation, Software development, Software engineering

This master thesis is a case study on software reuse within a subset of systems at the Norwegian Directorate of Taxes, Skattedirektoratet (SKD). The systems chosen for our research are the GLD systems; legacy systems which dates back to the late 1980's and early 1990's. Because of historical reasons, these systems are copied and created over and over again in an annual cycle. There are redundancies in code and data between the annual versions, but also across the different GLD systems. The consequence of this is systems with reduced maintainability and possible inconsistencies in code and data. Our objectives with this case study is to determine both the current level of software reuse within a subset of the GLD systems, and the emphasis on reuse in SKD's development process. After determining the status of as-is, we will continue with an investigation of the potential for software reuse within the context of SKD, and how they can achieve systematic software reuse. The contributions of this thesis can be divided into four main themes: * T1: Review of state-of-the-art literature on software reuse * T2: Investigation of reuse level within selected GLD systems * T3: Investigation of SKD's development process * T4: Investigation of opportunities for systematic reuse in SKD The main contributions are: * C1: Review of literature in the field of software reuse * C2: Measurement of the reuse maturity level within the selected GLD systems. * C3: Survey of the software development process and reuse aspects at SKD * C3.1: Results from SKD * C3.2: Results from SKD combined with results from previous surveys on software developers attitude toward software reuse by NTNU * C4: Process which assures reuse Keywords: Systematic software reuse, Reuse maturity, Software development, Software engineering

This thesis looks into the literature in the fields of informal learning, mobile learning and mobile games. A demonstrator is build to test the effect of informal learning and the use of collaborative games in a learning context.

N/A

The use of indoor positioning systems (IPS) is seen as a promising way to support coordination of activities in hospitals. Despite this, little work has been done when it comes to how the data from indoor positioning systems should be visualized to the users. By conducting a case study of a specific work practice at Aker University Hospital in Oslo we have designed visualizations with the aim of investigating the amount of information richness these should have in order to support coordination work. We have also been interested in investigating whether the approach that we have used in this thesis made it possible for the users to use the visualizations in their everyday work, and if it is possible for the users to give informative assessments of them. As little relevant theory exists on how visualizations should be designed we have used methods from the field of usability engineering, and have designed the visualizations as paper prototypes. These have been user tested by test persons with both direct and indirect knowledge about the coordination of activities at the specific work practice. Our work shows that visualizations that had a high degree of information richness were favored. In addition, the possibility to move between different types of visualizations based on the need for information were seen as especially attractive. The test persons that we recruited were able to give informative assessments of the prototypes, but as these were both tailor-made for a specific work practice and tested using simulated scenarios they had a high learning curve. Based on the results from our work we see usability engineering methods as suitable when designing visualizations, although some of them should be left out or modified to fit in with the project in question.

Oppgaven ser på et område som det i dag er implementert lite støtte for i eksisterende EPJ-systemer (Elektronisk Pasientjournal), knytting av emner til ustrukturert sykepleiedokumentasjon ved hjelp av tradisjonelle søketeknikker og rammeverk. Søking og navigering er dårlig utviklet i mange EPJ-systemer, spesielt for ustrukturert dokumentasjon. Knytning av emner til denne typen dokumentasjon kan derfor lette sykepleiernes hverdag, og metoden som presenteres i denne oppgaven kan være et verktøy i denne sammenhengen. Automatisk klassifisering er knyttet til mange usikkerhetsfaktorer. Disse usikkerhetsfaktorene forsøkes belyst ved vise til sykepleiedokumentasjonens natur og sprikende struktur, noe som vanskeliggjør en regelbasert tilnærming. Oppgaven argumenterer for at man kan tolerere feilkilder så lenge man er klar over at disse eksisterer, og at nytteverdien av å søke på emner fremfor enkelttermer er større enn problemene disse usikkerhetsfaktorene gir. Bruk av eksisterende rammeverk fremfor egendefinerte klassifiseringssystemer er en naturlig innfallsvinkel til en slik problemstilling. Oppgaven forsøker å vise ett prinsipp, mer enn en favorisering av ett spesielt rammeverk. Likevel vil ett rammeverk, Sabaklass versjon 2.0N, bli mer omtalt enn andre på grunn av at det har vært benyttet i implementasjonen av en prototyp i forbindelse med oppgaven. Oppgaven viser hvordan slik emnetilknytning kan foregå ved hjelp av en prototyp som er blitt utviklet for formålet. Prototypen benytter vektorrommodellen (VSM) sammen med en ordliste knyttet til rammeverket for å klassifisere ustrukturert sykepleiedokumentasjon. Valget av vektormodellen fremfor andre, tradisjonelle tekstgjenfinningsteknikker, er gjort ut ifra modellens egenskaper til å rangere delvise treff. De konkrete resultatene av prototypingen kan være diffuse. Dette er i stor grad knyttet til konstruksjonen av ordlisten, og vanskeligheter knyttet til å skaffe en god testsamling. Likevel peker de foreløpige resultatene i favør for en slik søkemetodikk, fordi den kan benyttes sammen med eksisterende søkemetodikker, og den representerer en fordel ved ukjent dokumentasjon.

Rekonfigurerbar maskinvare har i den siste tiden vist seg å være et særdeles nyttig verktøy i forbindelse med akselerasjon av forskjellige algoritmer. Spesielt gjelder dette algoritmer som enkelt lar seg parallellisere. Et av hovedproblemene med bruk av rekonfigurerbar maskinvare er at det tar relativt lang tid å rekonfigurere selve maskinvaren for en gitt algoritme-implementasjon (modul). Hvis en bruker flere forskjellige slike implementasjoner etter hverandre kan dette bli et veldig stort problem. Partiell rekonfigurering har lettet litt på dette problemet. Dette fordi partiell rekonfigurering gjør det enkelt å ha flere moduler lastet inn i maskinvaren samtidig. I tillegg er det mulig å ''skjule'' rekonfigureringstiden ved å rekonfigurere deler av maskinvaren mens resten jobber med en oppgave. Flere moduler på samme brikke betyr også at det er større sjanse for at en ikke trenger å rekonfigurerer maskinvaren før en gitt oppgave, siden det er en sjanse for at modulen allerede ligger konfigurert. I denne oppgaven presenterer vi et system som har mulighet til å ha 4 forskjellige moduler konfigurert til enhver tid. Denne oppgaven ser på muligheten til å benytte temporær lokalitet i bruksmønstret til de forskjellige modulene, til å avgjøre hvor i maskinvaren en ny modul skal lagres. Med dette mener vi å implementere et system som holder rede på hvor ofte de forskjellige modulene blir brukt. Denne informasjonen blir så benyttet til å bestemme hvilke moduler som skal få ligge i maskinvaren og hvilke som skal kastes ut når det blir nødvendig å laste inn en ny modul. Til dette benyttes en enkel LRU- (least recently used) algoritme, og systemets virkemåte har mange likhetstrekk med hvordan hurtig-buffer for minne fungerer på en vanlig PC. Med dette håper en å redusere antall rekonfigureringer og på denne måten redusere den totale rekonfigurasjonsforsinkelsen. Vi ser også på muligheten til å stokke om på rekkefølgen oppgavene blir sendt inn i, for å prøve å minimere antall rekonfigureringer. Systemet består av en programvaredel som holder rede på bruksmønsteret og tar avgjørelse om fordelingen av modulene. I tillegg så inneholder systemet også en maskinvaredel som tar seg av kommunikasjon mellom modulene og programvaren. Modulene ligger på egne datafiler som blir lest inn ved behov av brukerprogrammet. Brukeren sender inn oppgaver til systemet ved å spesifisere hvilken modul han/hun ønsker å benytte og hvilke data som skal sendes inn til den.