Tech Target has an interview/article about Cobbler here, though it contains a few errors that we are trying to get corrected.
The article above made the (wrong) assumption that Cobbler has about 15 installations, because they only looked at Who Uses Cobbler. There are some great folks listed on that page of course, but that’s only a small portion as we obviously never require someone using a project to publically declare they are using it. I guess we have about 1000+ installs.
The particular thing that this brings up is the question — given what data that you have, how do you tell how much an OSS project is being used? You definitely don’t have the pool of people that have paid you (it’s Free) and you haven’t required any scary “phone home” type registration/activation deals for the same reason. While it’s impossible to gauge exact usage of any project, I suspect the user base is instead somewhere in the thousands. So what data do we have?
We have 102 unique mailing list posters in the last 4 months, 45 unique code contributors to cobbler, 14 unique code contributors to koan, and currently 71 folks on IRC. There’s also a huge bunch of corporate users that are absolutely awesome, but I can’t talk about. We have 68 subscribers on Freshmeat, for those that actually have set up Freshmeat accounts. Though none of this really tells us anything.
Either way, the thousands metric for this particular project assumes that about 1/10 of the users of a project are active on a mailing list or IRC. I would suspect that the percentage of users who get involved is actually lower, and I’ve definitely seen a lot of suprise “we’ve been using this for a year” comments from people I’ve been talking to about other things. Lots of folks won’t show up on a mailing list when there are things they can figure out themselves, or they’ll just read the archives and learn what they need to learn.
What is the percentage of users who contribute patches? I suspect that number is still much lower than 1/10, maybe it’s 1/50, maybe it’s 1/100. If it were 1/100, we might have 4400 installs, but that seems too high. Hard to tell.
Another problem in computing a metric is that some projects are easier to patch than users. So the choice of language and domain of the application definitely affects the ratio. Is a project for developers or for users or somewhere in between? How open is the community and how does that play a role? Do people who do “yum search” even look for online communities in most cases, or turn to email or bugzilla when they hit a bug? I suspect a lot of them may not because they are strapped for time or are not aware these things exist or are that widely used.
If there exists a Drake Equation for open source adoption, I don’t know what it is. This may also be another good role for EKG to take on in the future, though I’d really like see a Package Popularity Index type feature in Fedora/EPEL as well — if that just provides one more data point. Even though the numbers of the popularity content themselves are also skewed by the data being public — it’s a popularity contest, after all.
At best, I think we can determine a couple “adoption metrics” but we can never come to compute the actual usage of a given project.
Does anyone have their own guesses what the percentage of users that (A) join lists and (B) submit patches are, for different projects? How might we reverse engineer OSS adoption from the data we have available?
