‘Requirements’ are misleading and undermine good web design
Bas van Essen,
Friday, March 9, 2018
In-depth research into software development. Jexia provides you with the latest scientific coding insights.
In many software engineering projects the ‘requirements’ would not be simply flawed, but entirely fictitious.
It’s just one of the claims in the paper ‘The Dangerous Dogmas over Software Engineering’, written by two British scientists. Their first sentence is hard and fast: “To legitimize itself as a scientific discipline, the software engineering academic community must let go of its non-empirical dogmas.”
At most 1.8% of all Github repositories containing code from Stack Overflow, applies code in a way compatible with Stack Overflow’s license ‘CC BY-SA 3.0’.
One of the involved researchers, Sebastian Baltes, wrote a blog about it and concluded that (1) developers rather refer to the questions than to individual answers in source code comments; (2) at most one quarter of the copies of Stack Overflow snippets are attributed using a link to the post as required by the license; (3) most developers are not aware of the licensing of content from Stack Overflow and its implications; (4) about 10% of all projects on GitHub could be affected by licensing issues due to the usage of non-trivial code snippets from Stack Overflow.
Results suggest the concept wastes a feedback channel from the IDE to the coder.
An even more worrisome conclusion: “The data provided no evidence that syntax highlighting improves novices’ ability to comprehend source code.”
Three Chinese scientists emphasize that ‘widely used defect predictors such as code metrics and process metrics can not well describe how software modules change over the project evolution, which we believe is important for defect prediction.’
They got satisfying results after testing their new bug prediction approach. And now they would like to extend their approach to more projects. Anybody help them?
To find out, three Finnish and one Canadian researcher investigated more than 700,000 version control commits from 87 software projects.
Join our monthly newsletter ‘Coding Insights from Scientists’.
Also published on Medium.
This post was written by Bas van Essen