I had a chance to talk to some people from the user community at the
conference, and a lot of them seemed a bit skittish about getting
involved in development. The model I talked about with them is creating
nuclea about product areas like LAMP, dynamic language, media players.
Let users talk about what they're doing with the various pieces of
technology we're packaging or that they want packaged. Start with a very
loose consistency model: have them share blogs, and let them create
special interest group mailing lists to build communities around focused
interests. For some people, looking at the broad problems of developing
a distro isn't a way in. Talking about production, documenting what
they do, and writing code to test product stability around the
operational side they know is something they're comfortable doing. Build
them up from that. Once they've done it, they realize they can have
confidence building and packaging stuff, knowing that they've got a test
rig to show them the confidence is well-founded.<br>
<br>
My surmise is that we've got a lot of users who are--I don't want to say
conservative but production-centric. We need to help them make the
transition from being Sun customers to OI contributors. They come
from an operational background, so grow them into developers using the
devops paradigm: focus them on defining quality, then measuring quality,
and then moving the software forward.<br>
<br>
What I'd like to propose to get us there is three things: one is doing a
survey of the user community to figure out where they are, what their
interests are, and what they expect from the project. My guess is as
good as anyone else's, but we're still guessing. Let's form some
hypotheses, write a survey that gets us data to evaluate, and get
ourselves something we can analyze more conclusively. And let's do that
cross-distro for illumos, if possible: OI, illumian, SmartOS, StormOS,
etc.<br>
<br>
Second, let's try to organize a pan-illumos userland hackathon that's
also a bootcamp. Let's target OI users and FOSS developers. We teach a
package of skills: how to build stuff with OI, how to measure stuff with
DTrace, how to test stuff with the unit testing rig Delphix is putting
together (I'm assuming we just want to steal this, but maybe I'm wrong),
and how to automate the boring stuff with Jenkins. (As much as possible, downplay packaging in the interest of cross-distro unity: here's the common data you have to understand for all of userland, so you know what to grok if gmake publish fails.) If possible, fly in a DTrace rock star. Teach those skills
in a day or less, and spend the rest of the weekend using them so that
there's immediate practical reinforcement. Let people sign up ahead to
work on packages so that they come away from the event with something
read to commit and thus a sense of accomplishment on which to build.<br><br>The
Nexenta OSS Europe conference may be the right event to work off of, as
it's just far enough in the future that we can make it our goal now and
keep ourselves focused making it happen between now and then. We've got
people in Belgium, the Netherlands (I got contact info at FOSDEM for the
person who used to run the Solaris users group for Holland), Germany,
the UK, Denmark, and Central Europe who could our usual suspects. Tell
them to bring friends this time. Help developers see us as a growth
platform where it's worth their time to engage with us to make sure
their software runs well our platform, which helpfully makes that a
measurable question.<br><br>
Third, let's designate some people to organize more locally to pull
people in. Give those people
a private mailing list and let them talk about what's working and what
not. There are a number of places we can go to find organizers, and FOSDEM provided some contacts for people willing to do this.
Again, let's go pan-illumos on this to get some scale. Encourage them to cross-pollinate their meetings between software
users and software developers. Help them focus on DTrace in particular
because it intersects so neatly: illumos is the reference platform, but
DTrace is more widely available. I talked to a Python developer at the
conference who mostly wanted to rave about how great DTrace was, based on his experience of it on OS X.<br><br> Our hook is that we make quality something you can hold in your hands. That's what our community wants, that's what makes our community valuable to engage.<br>