[Ksummit-discuss] [CORE TOPIC] Recruitment (Reviewers, Testers, Maintainers, Hobbyists)

[Chris Mason]

On Mon, Jul 20, 2015 at 10:40:55AM -0700, Tim Bird wrote:
> 
> 
> On 07/17/2015 01:02 PM, Chris Mason wrote:
> > The first patch really doesn't seem to be a problem.  At least from
> > the stats I've seen so far.  How do we get the 10..100th patches,
> > hopefully without 90% of them being whitespace fixes?
> > 
> > We're not going to be able to answer these without
> > actual data.  This means surveys and talking with
> > new developers that we really hope to turn into
> > long term members of the community.
> 
> I may have a data set that is relevant here.
> Last year I surveyed developers, targeting those who
> 1) had made a change to the kernel that shipped in a commercial product,
> (defined as "qualified" in the survey analysis)
> and
> 2) did not make contributions to mainline
> 
> There were 93 out of 278 "qualified" developers who never submitted a
> patch.  That's about 33% of the developers who responded to the survey.
> I suspect that the no-submission rate for developers who either did
> not see the survey or did not respond to it is much, much higher.
> This survey was focused on developers in the consumer electronics field.
> 
> 25% of all survey respondents (not just the "qualified" ones) reported
> they did not know how to contribute a patch.
> 
> The survey was not detailed enough to determine what parts
> of the submission process caused the most reluctance to contribute.
> 
> I would guess (not very scientifically) that although we have thousands
> of developers contributing to mainline, the pool of developers paid by their
> companies to make products is in the range of 10s of thousands, and
> our rate of conversion to contributors is under 10%.  (I'm not sure
> if that's good or bad.)

I love hard data on this topic, and I really like the idea of trying to find
the pool of actual kernel developers vs the pool of people visible on
the mailing list. 10:1 makes me sad, but I'm not that surprised.

Still, I'll channel Greg for a minute and google "how to send a patch to
the linux kernel".  I'd definitely believe people don't know how to get
their employer to prioritize (allow?) sending the patches in, but I
can't stretch to they don't know how to do it at all.

-chris