It is also possible that Ohloh's tracking is not perfect. They can't track things that haven't been reported to them, and that process can take time. So I'd expect under reporting of recent changes. (Not necessarily as much as we see though.)
Their tracking is probably imperfect, but it seems like a reasonable indicator. I'm still waiting to see if their data has some flaw, or maybe if github has more.
However, that data wasn't the point, just an indicator.
Nope, nor does it need one. Source comments are weakly correlated with code quality :)
Moreoever, you have to be careful about numbers of lines-based statistics for numpy and scipy, because a lot of code is generated or 3rd party libraries (that are integrated because no distribution has them packaged, not even debian). Also, cloc gives me vastly different numbers, with a lot of comments and all (but again, not very meaningful because of generated/integrated code). ohlo stats are not very reliable in my experience, at least for the packages I am involved with.
Sourceforge? Ohloh collects stats and information from all over, they even have a Mongrel2 project page and I've never set it up. I'm pretty sure you're mistaken about the kind and quality of data they collect, but who knows. Their blog seems to say they track git, cvs, svn, others?
That's also why I mentioned github, they could provide other data (if they weren't too busy inflating their numbers by counting every commit, gists, forks, and duplicate projects as a "project").
They can track almost anything, but require somebody to manually create the project and "enlist" it's source repositories. And because often the person who created the Ohloh project has almost nothing to do with original project these data tend to get stale.
Does anyone know how Oholoh gets its numbers? As far as I can tell, projects have to be manually added; it doesn't appear to scrape them from hosting sites.
Actually, around the time of the 'decline' lots of projects started switching to DVCSs like git. So I guess many enlistments are still pointing to the old CVS or SVN repos where obviously nothing is happening.
It is also possible that Ohloh's tracking is not perfect. They can't track things that haven't been reported to them, and that process can take time. So I'd expect under reporting of recent changes. (Not necessarily as much as we see though.)