Open Stack

Sorry for the slow response.

And borrowing the worlds of Blaise Pascal, I apologize for the length
of this letter; I had no time to make it shorter. ;)

One of the values of being offline for a bit is to get to read larger
chunks of the conversation as opposed to incrementally as they come
in. This mail contains comments in response to statements made by
Thierry, Dave, and Anne. This has actually been good in that it has
given me time to figure out why my perception of the situation is
different from the responses from folks here, and caused me to
reconsider my opinions in some ways.

First, let me say what I mean by users. When I talk about users, I
mean people who primarily consume the software and use it. Some of
these people will contribute back, in terms of bug reports, or perhaps
even bug fixes, but their primary scope of work is deployment and
operation of openstack.

Thierry, you're absolutely right; these folks are in the right place
to be able to tell us where documentation is wrong or incomplete,
where openstack breaks in weird (or regular) configurations, etc. This
information is enormously valuable to openstack as a project. I think
that an important question here is how far into the contributors
column these folks will be able to step, in terms of time investment,
effort, etc. This is an impossible question to answer with complete
certainty, and there will be all sorts of caveats having to do with
corporate culture, and a myriad of other factors. Be that as it may,
as a community, it is a reasonable question to ask what the
cost/benefit looks like for leaving that information on the table. The
take home message here is that the incentives aren't aligned between
people who are deeply involved and people who are casually involved.
People who are casually involved have another job to do. Openstack is
a potentially prominent part of that other job, but job #1 is X (site
reliability is a good example, but there are a bunch of others as
well).

Dave, I think that my formulation of the user category is somewhat
orthogonal to the categories that you're talking about; conceptually
it is the user/dev split. Generally, at least in open source projects,
I think of devs as being the people that produce the software, etc.
While there are some people in the user category that contribute to
the project in terms of documentation, infrastructure, etc, bug
reports and help to other users are as far as most people in the user
category go.

One other critical detail; users tend to use the software that has
been released, not the software that is being developed. I think that
this is borne out by the recent spate of comments of users planning
for essex to folsom upgrades on the mailing list. All of the
contribution processes seem tied to release+1, while the focus of many
users is release at best, if not release-1, or release-X.

As a related comment, Thierry, I think that your goal of most
end-users contributing to openstack as a general matter of course is
probably unrealistic. Even with good end user engagement, I'd be
surprised of there wasn't some kind of exponential ratio trend
separating core contributors, from occasional contributors, from rare
contributors, from relatively passive (mainly consumption-oriented)
users in open source projects as a whole. I would argue that Openstack
has the best infrastructure I've seen for contributions when people
are engaged. I think that the mechanisms for interacting with users
(casually involved people in terms of development) is ad-hoc at best.

When thinking about interactions with users (under my definition),
there are a bunch that are worth considering. New users evaluate
openstack, usually though some sort of POC. Once users have decided to
deploy openstack in some form, they need to figure out how best to
deploy it for a production deployment. Once it is deployed, people run
into bugs, performance problems, and the like.

All of these interactions hinge on user support. And user support is a
tricky issue. It is time consuming and expensive, particularly when
the user base is growing so fast, and international.

I don't mean in any way to diminish the contribution or value of the
documentation. The doc has made tremendous strides over the last two
years, and Anne and the rest of the doc contributors deserve credit
for it. At the same time, it seems that there is still a lot of demand
for interactive support, either via email or IRC. I think that this is
unavoidable; because openstack operates across a series of complex
broad areas (virtualization, storage, networking, etc), documenting
openstack in a definitive, comprehensive way is more difficult than
other projects. I personally believe that the need for interactive
support is linked to this fundamental complexity in cloud system
software stacks.

IMO, the only way to address this issue in the long term is to
activate and engage the user community in order to build a relatively
self-sustaining ecosystem. So far, I feel like the openstack community
hasn't managed to catalyze this activity sufficiently. We have several
venues, each of which end up lying fallow much of the time. (I'm
thinking specifically of openstack-operators, but questions are often
left unanswered on IRC as well.)

As a related rhetorical question, I'm left with the feeling that Vish
still spends a lot of time answering user questions. When we first
deployed Bexar, there were a few tricky problems that we would not
have been able to get past without his assistance. At the same time,
as the project has grown, his time/effort has become one of the most
substantial assets we have as a project, as developer contributions
are exploding. This seems to be the case in other openstack projects
as well, where the PTLs seem to be a substantial source of user
support. I'm not suggesting that PTLs should firewall themselves off
from users with questions, but it seems like there is a gap there that
we should be thinking about filling somehow.

I'm also not meaning to suggest that there aren't users that are
stepping into the fray in terms of user support. We all know the
familiar names from the mailing list and IRC of people that step up,
but there seems to be an insufficient number of these people for the
community as it stands, nevermind the community we hope to have
tomorrow.

This leaves the question, how do we effectively build out this part of
the ecosystem? I'm not sure that I have easy answers on this, but I
think that this is the key issue that the user committee needs to
tackle.
 -nld