toggle quoted messageShow quoted text
The issue we are facing here is how to make projects, and hence the cncf, sustainable. As Quinton pointed out this requires incentives. Marketing buzz is short lived and sponsors will move on to the next thing, if there isn't an economic system to help the ecosystem.
We have seen this before in various forms, including asf, esf, and openstack. Ultimately these foundations need to make it possible for enough actors to make money or it is really hard to keep going. Openstack made it too hard to differentiate. Eclipse got bogged down in tools. Apache is overrun by bureaucracy and placepeople, and is overprotective of its marks in the name of "open source".
When we formed cncf we wanted something better that learned from all the prior art. CFF was quite interesting and had created terrific end user interest, but lacked diverse projects. It was basically CF plus libs and addons.
CNCF is not Kubernetes plus addons. Yes k8s is our sun. But many use cases don't use or need it, and this also helps users. In addition ISVs like Hashicorp can see value for integration of their suite with multiple points in CNCF.
If you look away from the bright light of k8s, you see a diverse range of projects. Many come from ISVs, and some from end users.
Can i suggest that we listen to these ISVs about what works *for them* and without assuming that they want to game the system or corrupt the ideals of this foundation.
On Thu, 9 Jul 2020, 18:48 Jaice Singer DuMars, <jdumars@...
As an observer, I think this is a very solid analysis and recommendation. The risk of a "rubber stamp" steering committee is a serious concern, and clearly doesn't resolve the underlying challenge of meaningful corporate diversity among maintainers.
That said, I do believe a standardized steering committee template has value though, especially for projects reaching the natural governance evolution where one makes sense. It's easy to forget that the Kubernetes steering committee was a response to community distress as documented in the contributor survey at the time. Having a vetted and well thought-out charter to start from can limit the time needed to resolve governance friction.
Is this working group going to suggest an alternative?
On Thu, Jul 9, 2020 at 9:59 AM Josh Berkus <jberkus@...
As promised, we had an in-depth discussion of the Steering Committee
proposal drafted by Alexis Richardson for the July 7th TOC meeting
during the Governance WG meeting. Six members were present and
participated in the discussion; notes are available here:
Regarding the specific proposal of using steering committees (SC) as a
workaround for the requirement to have maintainers from multiple
organizations for projects to move to the Graduated level, the
Governance WG recommends that the TOC not adopt it.
Steering Committees are frequently important governance tools for large
and distributed projects, and more projects should consider having one
as a channel for end-user, collaborator, and diverse audience
representation. We valued Alexis’ writeup, and would like to
incorporate it into our handbooks in progress for CNCF projects on how
to develop governance.
Projects that would need the "SC workaround" are projects that have been
unable to attract a single maintainer from outside the original
sponsoring organization, which is usually a sign of serious issues
within the project. The Governance WG feels that the CNCF ecosystem is
ill-served by moving projects with such problems to the Graduated level,
and the proposal will not have the desired outcomes.
Our decision was based on what we view as the inability of a
non-technical Steering Committee to ensure that a project with
maintainers* exclusively employed by the same organization treat
submissions, roadmap items, and maintainer candidates from other
organizations fairly. Even diligent SC members would find it difficult
to understand enough about technical architecture decisions to
differentiate between bias and legitimate objections in reviews.
Further, unlike code and docs maintainers, it would be challenging for
the TOC to monitor activity and involvement levels of SC members, as
that would not create the same kind of contribution trail.
For this proposal, we considered specifically projects that are having
problems attracting contributors, because only such projects would need
this mechanism. It certainly takes time to bring code reviewers up to
speed, but the current requirement is a low bar; even a single dedicated
documentation leader from an end-user company would technically satisfy it.
While many folks have cited the Kubernetes project as an example,
Kubernetes has a diversity of maintainers all the way down to the SIG
level, so it would qualify for a maintainer multi-org requirement even
without a Steering Committee. At this point, the Governance WG does not
know of a good example of a project that successfully has used an SC to
moderate the influence of development being dominated by a single
company, so doing so would be experimental. As an experiment, we might
adopt it for one specific project, but we'd want to see the outcome of
that before we adopt it as general policy.
Even Alexis’s document suggests that in problem cases it would be up to
the TOC or their delegates to intervene to resolve project problems.
Given this, it’s unclear what advantage having an SC would offer over
Alexis’s original suggestion of having designated TOC monitors.
Overall, our judgement was that adopting the SC workaround would be, in
essence, removing the maintainer multi-organization requirement, and
that it would be better to simply remove the requirement instead if that
is the direction the TOC wishes to go.
Drafted by Governance WG:
(* by “maintainers” we mean involved, leading contributors with the
authority to merge code, docs, and/or community materials into any of
the key repos belonging to the project. Such contributors may be called
"maintainers", "committers", or other titles. It does not refer to the
official CNCF maintainer list for voting purposes, as that list contains
many people who do not have merge authority in their individual projects.)
Red Hat OSPO