Re: Sandbox process needs to evolve to support cross industry collaboation

alexis richardson

Richard how would you formalise this?  The goal, IMO, is to reduce the subjective judgment on entry to sandbox, and increase the quantitative aspects

On Thu, 5 May 2022, 13:38 Richard Hartmann, <richih@...> wrote:
Replying top-level as my thoughts jump across the thread.

I didn't run the numbers, yet I believe that the pace of submissions
has picked up. That alone can increase backlog.

We tried SIGs (now TAGs) doing due diligence for projects. The level
of scrutiny, and the closeness to the guidance material available, was
different across TAGs. In effect, this meant inconsistent processes
which is arguably unfair. And in cases of disagreements, TOC is pulled
in automatically anyway.
A clear delegation from TOC might be possible, yet project advancement
is one of the main tasks of TOC and arguably what votees expect TOC to
do. In any case, it does change any of the underlying desires.

What TAGs could provide is an initial proving ground, though: Projects
could give a presentation and go through questions and feedback in a
more limited scope, allowing them to polish their submittal.

While I know that the current sandbox process is designed to be very
low barrier, I am still not convinced that this is an obviously
desirable design goal. It is true that a neutral playing field is good
and helps some projects grow. It is also true that "CNCF project"
holds immense marketing value and many efforts are ephemeral, in
particular if largely driven by perf & marketing.
Back when sandbox criteria were relaxed, I was of the opinion that
they should remain more stringent. I have come to wonder if four
levels wouldn't be more appropriate: An initial runway on which
projects can be put; but also pruned more aggressively if they do not
show growth/adoption/the usual. E.g. once submitted they have three?
six? twelve? months to show certain progress or are removed outright.
Medium term, this might also allow for a smaller jump towards
Incubating, which is currently significant.

Orthogonally, I believe we can manage expectations better. One
possible approach would be to create dashboards and reports of the
underlying data to help manage expectations and keep ourselves honest.
What are the average and median times a project takes from stage X to
stage Y? How has this changed over time?
Another would be to rework the process & documentation; e.g.
Incubation had distinct requirement docs which TAGs copied together
and deduplicated back during the DD trials.

Having seen things from both sides now, and since CNCF started, I can
understand both the frustrations about some timelines better and also
understand how a few dedicated people are trying to do their best with
the time they have. On all sides.


Join to automatically receive all group messages.