Results of Community Research

[Max Kanat-Alexander]

On 02/12/2010 02:37 PM, Max Kanat-Alexander wrote:
>> Or perhaps look at attached patches in Bugzilla as well as
>> checkin data?
> 
> 	Yeah, actually, I think that would be definitely worthwhile, at least
> to get a sense of whether or not there was some problem with people
> attaching patches but them never getting in, if nothing else.

	Wow, so this analysis produces some very interesting data!

	I see some definite trends:

	1. From 2002-05 to 2004-01, the number of contributors decreases steadily.
	2. From 2004-01 to about 2005-12, it increases significantly and regularly.
	3. From 2006-01 to about 2007-12, it dives significantly and continuously.
	4. From 2007-12 to 2008-05 it makes a comeback.
	5. From 2008-05 to 2009-07 (which is where my data basically ends) it
stays stable, perhaps a slight rise.

	These trends are actually very similar to the checkin trends, except
that they more clearly show a very specific rise to a peak at the *end*
of 2005, whereas the previous data looked roughly level for most of 2005.

	I then did an analysis of the review flag, excluding review requests
and only checking denied and granted flags. The trends of how many
reviewers we have active at any given time correlate nearly-perfectly to
the curve of how many active contributors we have at any given time.
Also, interestingly, roughly 50% or more of all review flags set are
review+, and the percentage of r+ reviews granted in a month doesn't
correlate to the number of active contributors by any curve. I think it
doesn't matter whether we grant r+ or r-, as long as we grant it quickly
either way.

	Here's the raw data for this analysis:

	http://spreadsheets.google.com/ccc?key=0Ar05y8vgF7BhdDBlb3hJU1BKYWMySDFYdE1oOXlKTWc&hl=en

	The next thing to check would be "average review delay" -- that is, the
average amount of time in a particular month that a patch attacher has
to wait for a review, including standard deviation. That would confirm
my hypothesis from above, though there were probably other factors
involved in how many contributors we had at any given time.

	-Max
-- 
http://www.everythingsolved.com/
Competent, Friendly Bugzilla and Perl Services. Everything Else, too.