handyfloss

I read in Kriptópolis, via a Basque blog that the companies Coverity and Symantec, along with the Stanford University, have made a study regarding the number of bugs in both free and proprietary software. This study has been funded by the North-American Homeland Security agency.

The study has focused on comparing the number of bugs per line of code of similar free/non-free programs one-to-one. Many previous (non-independent, Microsoft-funded) studies before, simply counted the number of total reported bugs in, say, Windows XP and a given Linux distro. This method is clearly biased against the particular Linux distro studied, because there are many different programs in any Linux distro that perform the same task (being able to choose is important for the FLOSS hippies, you know), and adding up the bugs of all those programs seems unfair.

The results of the study give the FLOSS an appalling victory (surprised?). Firstly, of the 32 program pairs, the free partners showed an average of 0.43 bugs per 1000 lines of code. The non-free ones turned up to have a shameful average of 20 to 30 bugs per 1000 lines (45 times more).

Secondly, not only the number of bugs was lower in FLOSS programs, but also the speed to fix them was found to be much faster. As an example, Amanda (a FLOSS backup program), was found to have 1.22 bugs per 1000 lines of code (the highest of all the FLOSS programs in the study, still much lower than any non-free program in the study). Apparently, the Amanda developers read the study, got ashamed, and one week later they had fixed most of the aforementioned bugs, going from the most bug-ridden FLOSS program of the study to the less bug-ridden one! Apparently pointing out where the errors are is veeery healthy for any FLOSS project.