the 3rd slide shows the total number of commits per language on popular code hosting sites. To my eyes, there is a distinct split between 1st tier languages (C++, Java, Python, C, Javascript and Ruby) and 2nd tier languages (C#, Perl and PHP).
As a Drupal developer, I'm well aware of the enormous sample bias inherent in this study: Drupal folks don't commit PHP to code forges because they commit it to Drupal.org, which built its own code forge back in the mid-2000s and has a community culture that strongly encourages its use.
Similarly, Wordpress runs a plugin repository where the Wordpress folks probably keep most of their code.
Meanwhile, this study also doesn't count CPAN as a code forge, so pitchfork-wielding Perl programmers will be arriving on this thread in 3...2...1 seconds to make this argument again, only in a much stronger form. Only a fraction of PHP programming involves Drupal or Wordpress, but my impression is that nearly everything that gets done in Perl is eligible for CPAN, which has been around longer than any of Github, SourceForge, Google Code or CodePlex.
But that's part of the point, isn't it? The very fact that Drupal folks prefer to live in their own isolated code community could indicate a degree of provincialism.
The study also left out Bitbucket, which I'm sure still attracts a decent amount of Python projects (Mercurial is written in Python and was blessed by Guido as the preferred SCM). Yet that doesn't seem to have prevented Python from coming out strong in the comparison.
Lots of constructive feedback on the slides. To the questions of the forges included in the study, we were going off of the snapshot of data that Black Duck had available. Because it isn't randomly sampled, it can't be considered representative of the industry as a whole. This is why you don't see any, for example, over-representation of the underperformance of PHP relative to competitive runtimes: this isn't broadly representative, and therefore those type of conclusions aren't justified.
Instead, we're looking at specific trends within the forges studied from an observational study perspective. Which tells us something about diversification as well as forge performance relative to one another. Nothing more, nothing less.
All of that said, I certainly hope we can expand the number of forges studied in future. Bitbucket in particular is of interest to me.
5 comments
[ 3.8 ms ] story [ 25.1 ms ] threadSimilarly, Wordpress runs a plugin repository where the Wordpress folks probably keep most of their code.
Meanwhile, this study also doesn't count CPAN as a code forge, so pitchfork-wielding Perl programmers will be arriving on this thread in 3...2...1 seconds to make this argument again, only in a much stronger form. Only a fraction of PHP programming involves Drupal or Wordpress, but my impression is that nearly everything that gets done in Perl is eligible for CPAN, which has been around longer than any of Github, SourceForge, Google Code or CodePlex.
Lots of constructive feedback on the slides. To the questions of the forges included in the study, we were going off of the snapshot of data that Black Duck had available. Because it isn't randomly sampled, it can't be considered representative of the industry as a whole. This is why you don't see any, for example, over-representation of the underperformance of PHP relative to competitive runtimes: this isn't broadly representative, and therefore those type of conclusions aren't justified.
Instead, we're looking at specific trends within the forges studied from an observational study perspective. Which tells us something about diversification as well as forge performance relative to one another. Nothing more, nothing less.
All of that said, I certainly hope we can expand the number of forges studied in future. Bitbucket in particular is of interest to me.