The keynote on day 3 was done by Professor Sir David King, Director of the Smith School of Enterprise and the Environment at Oxford University. His talked was (obviously) not Software Engineering related. He talked about the challenges that we as a world face in this century. He identified the following challenges: climate change, ecosystems, health & development, energy security & supply, water resources, conflict & terrorism, food production and minerals. He argued that one should look at holistic solutions. If not addressing one challenge might worsen another challenge. In general an “awareness” talk, very well executed, but not saying much that I have not heard of before.
And then the days pick in presentations from me…
Bacchelli et al. Linking e-Mails and Source Code Artifacts.
This research tries to match source code artifacts (classes, modules, etc) to the email history that have accompanied the creation/changing of the artifact. To do this they studies 6 projects and their email mailing list histories – six coders analyzing by hand the e-mails and linking it to source code artifacts manually. This led to the creation of a benchmark on which the effectiveness of automated methods can be tested. They then continued analyzing the application of some machine learning techniques on the emails and show various levels of success. Some of these techniques took a long time – running times of 3 days were mentioned. They therefore set off to develop a lightweight technique for analyzing the emails based on good old regular expressions and found that they got quite good results. What really impressed me though were the actual presentation. The presenter managed to take a topic which many would think will result in Death by Powerpoint and present it an a relatively simpele way. The key to his success: simple slides, diagrams that’s not overly complex and careful use of animation. Well done!
Fritz et al. The Degree-of-Knowledge Model to capture Source Code Familiarity
Like the previous presentation this presentation was conducted in a very understandable format, taking a fairly abstract model and making it really practical to the audience. The presentation was done by three presenters, one of which were in Canada and provided voice-over slides. The basic introduction was done by one, then the voice-over slides explained the basic idea. What was interesting to note here is that they really connected to the audience by explaining the Degree-of-Knowledge (DoK) concept from the perspective of a multi-author paper submission to a conference, not source code. They first presented the how it worked and then gave a perspective from an independent observer (an owl) perspective.
The basic concept is: If I write code, I have a high DoK of that code, everybody else have nothing. As somebody else reads the code they gain some DoK, but I don’t loose any; however, once the other party edits the code, their DoK goes up and the original author’s DoK goes somewhat down. They then used this to try and answer specific questions, like who should one ask about the code? Their DoK indicator corresponded quite well with the opinion of the two teams the evaluated there results with. It turned out, however, that their DoK is not very successful predictor of what a new person should know about. The reason probably related to the fact that the DoK metric pretty much looks at details, whereas the coders would suggest studying some of the interfaces first – the level of abstaction is clearly different. While I think their is still a lot of refinement ahead, it is a very interesting idea. What’s more it was presented excellently – a case maybe of “selling your work” as opposed to “telling me about your work”!
ISCE2010 was a very enjoyable and informative conference. ICSE2011 is in Hawaii and ICSE2012 in Zurich, Switzerland. [If you are into Software Engineering start writing! this is a high-class event.