Community Participation Stats
February 13th, 2007 by seth bindernagel
Not sure if anyone else picked up this blog conversation between Asa and Sam Harrelson about online marketing and open source. IMO, Asa does a nice job discussing with Sam the size, vibrancy and contribution of Mozilla Community.
A few months in advance of Asa’s comments in that post, he and I ran a few MySQL queries against the Bugzilla database to get a sense of how much work had been contributed by volunteers leading up to the release of Firefox 2.
What we did?
We looked for those developers who had submitted 50 or more patches to the source code for the six months before Fx2. Why the lower bound of 50? Initially, I was trying to find a list of the key contributors to our software whom I could email to see how or if Mozilla could provide any resources.[1] We needed to draw a line somewhere, so when we first began, we searched for long-standing and key contributors to the project. Initially, we could reach out to them to see if we could help, and then move to those who may have fallen under our lower bound of 50.
Asa and I then cross-referenced those results with a list of Mozilla employees and contractors in order to remove the people who were getting paid to work on Mozilla full-time.
Finally, we added up all the volunteer-submitted patches and divided that number into the total number of patches from the query.
Here is what we found:
- 27% of the patches to Firefox and Gecko and other key projects were submitted by key volunteers. [2]
- Those patches represent 24% of changes made to the source code. [3]
What we learned and what we will consider next time we run the analysis…
1) No lower bound. Next time: let’s look at all patches submitted to the code in Bugzilla for Firefox, even if a volunteer only submitted one patch. This will give a better sense of total community participation and I suspect the percentage of community participation will go up.
2) This analysis did not include data gathered from queries to CVS. To start, we will probably stick to queries on Bugzilla data rather than other databases. But, as we do this analysis more regularly, perhaps we’ll expand to include queries on the version control system.
3) The analysis focused on Gecko, Firefox, Thunderbird, Mozilla Application Suite, NSS, Toolkit and “Other Apps”.
————————
[1] To be clear, we did this query for all the main Mozilla products tracked in Bugzilla, not just Firefox.
[2] Remember, we had a lower bound of 50 patches or more per developer. Admittedly, this was probably high. Next time, we’ll go lower.
[3] This number may not be all that informative because we took the aggregate byte size of just volunteer patches and divided that into the sum of the byte size of all patches submitted to the source.
An interesting question would be: how many people submitted these 27% of the patches?
The number is ~125. But, remember that it is over 50 patches submitted, so the number will be bigger when I re-run the analysis. I suspect it will change pretty dramatically.
does this analysis take into account the *many* developers paid by google / redhat / ibm / novell / etc, who are neither “mozilla employees and contractors”, nor (unpaid) “volunteers”?
We tried to screen out all contributors who are getting paid by a full-time employer. We wanted to see how many unpaid volunteers we have on the project. Next time we run the stats, we will have a lot of options to consider. Anyone have any ideas on the best way to gauge or measure volunteer contributions? I would be happy to take your suggestions and incorporate them next time.
I think you are right in your conclusion to include everyone who submitted a patch because he/she might be the one who sees it that will enhnce the product for the better.