Hi all,
The OpenStack Technical Committee met for two three hour sessions during the vPTG, covering a handful of topics. This email serves as an extremely high level recap of those topics. If you want full information on the discussions, please review the full etherpad notes[0].
The TC will be tracking long-term projects, related to these discussion items and otherwise, in the 2024.1 TC Tracker[1].
As always; no official TC policy changes can be made via a meeting. PTGs and other synchronous meetings are useful for rapidly gaining context and consensus, which then must be reflected in a resolution and merged when that consensus results in a policy change.
Leaderless Projects[2]
The TC discussed the multiple projects who did not have a PTL elected, and the various reasons that could cause that. A few interesting items were discussed which may need future action by the TC.
First, concerns were raised over the ec2-api PTL candidate not being an Active Contributor in title, despite showing stewardship over a relatively "complete" project. This is an example of how our process of identifying active contributors can be flawed. A workaround in the interim of a PTL in this situation nominating themselves as an extra-AC was identified.
Secondly, the recurring issue of smaller projects missing elections, and generally being less active, was a topic of discussion. The TC decided that as a regular course of action, any project that did not successfully participate in electing a PTL would also have the activity of its project reviewed as part of the PTL appointment process -- with a default towards marking such projects inactive[3] unless they have clear signs of activity.
Follow-ups from 2023.2 TC Tracker[4]
There are some ongoing items which the TC has continued to work on cross cycle. Their status was evaluated and discussed; items which needed additional tracking were pulled forward to 2024.1.
Managing Divergence between documented runtime and reality
This was a session originally inspired by a mailing list post from Herve[5] about how the listed versions of some libraries on the releases page for bobcat[6] were not the versions compatible with bobcat.
After significant discussion, it became apparent that there was confusion and lack of communication around a change that was not intended to merge yet, and did anyway, breaking project compatibility which had not yet done the SQLAlchemy 2.0 migration.
The TC agreed that proposing a new step in the release schedule where we integrate and test libraries with projects before cutting library releases is a potential solution to this problem.
More predictable python minimum supported version
Currently, our documentation indicates we drop support when there is "a solid reason", but recent mailing list discussions[7] make it clear that is not a clear line.
After significant discussion, the consensus was that unit testing the minimum supported version of python does very little harm but provides a benefit. While it was desired to have a general pattern we could follow, the realities of testing on LTS platforms make it so that we should continue to look at python version support on a case-by-case basis, comparing support lifecycles in python upstream, our current tested LTS, and the OpenStack release in question.
Similarly for new python versions, selecting what new ones to support is a technical concern rooted in library, distro, and ecosystem support moreso than what we'd desire OpenStack to support.
Communication
Communication to/from the TC was raised as a topic due to a misunderstanding around language used around a TC action decided on in a meeting. For OpenStack community members who don't frequent TC meetings, or interact with the TC process often, it can be confusing, but we should always be clear: binding decisions are not made in synchronous meetings.
Other items discussed around this communication topic included how more virtual communication methods, such as vPTG, may be limiting our cross-project communication and ability to reach larger groups.
Election Retrospective
This topic also covers many items covered in the communication topic; but essentially, many projects have missed elections -- the TC wants to find a root cause and address it if possible.
During the discussions, it became clear there was a significant cultural disconnect. Many pieces of documentation needed for OpenStack leaders -- including election reminder emails, governance docs, and this summary(!) -- are only available in English. Additionally, there is a concern that there are cultural conflicts that extend past language.
One set of solutions discussed were alternative PTL models, such as DPL. Some attendees indicated a dislike of the DPL model due to it potentially masking inactivity issues by removing the periodic check-in inherent to the elections process. An idea of electing a PTL for multiple cycles was raised as well, but did not gain wide consensus.
Additionally, we discussed how to get new leaders onboarded. A suggestion was made for the TC to send an incoming PTL email, with information and expectations for incoming PTLs. This is a process that traditionally has been handled well, albeit ad-hoc, in larger projects, but is often missed in smaller projects, especially if there is no handoff from one PTL to another.
Finally, there was a discussion about the size of our electorate being artificially small. Despite having 580 contributors in bobcat, OpenStack TC only had 309 eligible voters (AC+Foundation Member), and of those 309, only 91 people had opted-in to receive a ballot. Some suggestions were to revise our definition of who is eligible to vote in elections to include OIF non-members, but still require PTLs or TC candidates to be OIF members.
Shadowing Program
There was some discussion as well about the OpenInfra Diversity working group[8] piloting a shadowing program. Many TC members indicated a willingness to be shadowed, but a lack of people willing to take on leadership roles. (Along these lines; if you're reading this and want to shadow someone; ask!)
There were suggestions that perhaps the prestige of being a TC member or PTL has gone down, making it difficult to take on those responsibilities or justify them to management. We discussed several ways to try and re-engage people into leadership and contribution with OpenStack.
Footnotes:
If you made it this far; thanks!
--
Jay Faulkner
OpenStack TC Chair