Vaseline
03-08-2004, 10:06 AM
Stephen R. Schach
Department of Electrical Engineering and Computer Science
Vanderbilt University, Nashville, TN, U.S.A.
We present the results of three research projects in empirical open-source software engineering.
First, we consider Linus’s Law (“Given enough eyeballs, all bugs are shallow”), named in honor of Linus Torvalds, the creator of the Linux operating system. We show that Linus’s Law does not hold for Mozilla, Gnome, or Apache. On the contrary, between 70 and 90 percent of faults in the versions we examined were corrected by the members of the small core group (the inner circle of software developers), rather than by the “eyeballs” of the hundreds of thousands of worldwide users who have downloaded the software.
Second, we give a new categorization of common coupling within the context of software product lines, and use it to show that Linux will become extremely hard to maintain in the future.
Third, in 1978, Lientz, Swanson, and Tomkins (“LST”) published data that seemed to demonstrate that less than 20% of maintenance is performed in order to correct a defect. However, when we examined 60 versions of the Linux kernel and 15 versions of GCC, we found that over 50% of the maintenance was corrective in nature. We also query the validity of the original LST data.
An advertisement for a talk on my university campus.
Department of Electrical Engineering and Computer Science
Vanderbilt University, Nashville, TN, U.S.A.
We present the results of three research projects in empirical open-source software engineering.
First, we consider Linus’s Law (“Given enough eyeballs, all bugs are shallow”), named in honor of Linus Torvalds, the creator of the Linux operating system. We show that Linus’s Law does not hold for Mozilla, Gnome, or Apache. On the contrary, between 70 and 90 percent of faults in the versions we examined were corrected by the members of the small core group (the inner circle of software developers), rather than by the “eyeballs” of the hundreds of thousands of worldwide users who have downloaded the software.
Second, we give a new categorization of common coupling within the context of software product lines, and use it to show that Linux will become extremely hard to maintain in the future.
Third, in 1978, Lientz, Swanson, and Tomkins (“LST”) published data that seemed to demonstrate that less than 20% of maintenance is performed in order to correct a defect. However, when we examined 60 versions of the Linux kernel and 15 versions of GCC, we found that over 50% of the maintenance was corrective in nature. We also query the validity of the original LST data.
An advertisement for a talk on my university campus.