As discussed in PTG, we will try the DPL model first for the leaderless projects or if PTL assignment is needed in between of the cycle. - https://etherpad.opendev.org/p/tc-zed-ptg#L109 Change-Id: Ie65553720247b585be94d53811a2ade20353f9cc
9.0 KiB
House rules for governance changes approval
While most of the governance changes follow the rules described in
the charter-motions-section
section (or charter-amendment-section
section) of the Charter and call for a formal discussion and vote by the
Technical Committee membership, we also have a number of exceptions to
that general rule, in order to speed up the processing of smaller
changes. This document lists those "house rules" for reference.
The goal of the TC is always to operate by consensus where possible. The best decisions are always the ones where all TC members can commit to the outcome, even if they voted against it. These guidelines and the ones in the charter are designed to provide the chair with an objective standard for when they may consider consensus to have been achieved. However, if it is clear to the chair from other factors that consensus has not been achieved (for example because there are multiple incompatible proposals with sufficient votes to merge), the chair should use their discretion to keep debate open and seek rough consensus by other means.
Typo fixes
- Gerrit topic
-
typo-fix
When the change fixes content that is obviously wrong (updates a PTL email address, fixes a typo...) then any TC member (who is not the proposer) can directly approve them.
Community-wide goal proposals
- Gerrit topic
-
goal-proposal
The process for choosing community goals has two stages relevant to Gerrit changes: defining goal proposals and selecting goals for a cycle. For changes that propose a new goal, or that iterate on an existing proposal, apply the normal code review rules (with RollCall votes being considered +-2): change will be approved once 2 RollCall+1 (other than the change owner) are posted (and no RollCall-1). Any TC member (who is not the proposer) can approve them at this point.
Selecting community wide goals for a cycle should be proposed as a single change that moves selected goals from goals/proposed/ to goals/selected/. This way the collective feasibility of the goals can be clearly evaluated. Such a change requires a formal-vote.
Code changes
- Gerrit topic
-
code-change
The openstack/governance repository also contains code to build and publish pages on the governance.openstack.org website. For those we apply the normal code review rules (with RollCall votes being considered +-2): change will be approved once 2 RollCall+1 (other than the change owner) are posted (and no RollCall-1). Any TC member (who is not the proposer) can approve them at this point.
Documentation changes
- Gerrit topic
-
documentation-change
The openstack/governance repository also contains documentation related to internal operations of the TC but that does not represent formal policy. Changes to these documents, as well as minor updates to policies to improve the document without changing the policy itself (formatting, extra headings, wording changes that are not substantive, etc.), follow the normal code review rules (with RollCall votes being considered +-2): change will be approved once 2 RollCall+1 (other than the change owner) are posted (and no RollCall-1).
Election Results
- Gerrit topic
-
election-results
The results of elections are documented in the openstack/governance repository, but are not subject to "review" or "approval" by the TC, other than to confirm that they accurately reflect the outcome of the election. Patches to update those results for TC and PTL elections should be reviewed and confirmed by the election officials, and then approved and merged following the normal code review rules (with RollCall votes being considered +-2): change will be approved once 2 RollCall+1 (other than the change owner) are posted (and no RollCall-1).
Delegated metadata
- Gerrit topic
-
release-management
The release-management
setting for a deliverable is
delegated to the PTL of the Release Management team. When proposed or
approved by the PTL, changes can be directly approved by the chair.
Other project team updates
- Gerrit topic
-
project-update
This topic is used for other changes within an existing project team, like addition of a new git repository or retirement.
If there is no objection posted after the addition or retirement of a project is proposed (assuming the correct retirement process has been followed), then the change can be approved by the chair if it has at least two 2 roll call votes (which can include the vote of the chair and excludes the change owner) and a code review vote from the PTL of the project (or their delegate). If the change is from the PTL or their delegate, this will count as an acknowledgment code review vote.
One exception to this would be significant team mission statement changes, which should be approved by a formal vote after discussion on the mailing list or the gerrit change.
If a technical committee member disagrees with the addition or
retirement of a project, they can propose a revert which would then be
discussed by our usual formal-vote
rules.
Goal Updates from PTLs
- Gerrit topic
-
goal-update
PTLs will acknowledge community-wide goals at the start of each cycle by providing links to artifacts for tracking the work, or an explanation of why work is not going to be done. They will also provide references to completion artifacts at the end of each cycle for goals where work was needed. These patches should be reviewed by TC members for completeness and clarity (is enough information provided for interested parties to track the work, or is the justification for not doing work clear enough).
The patches to add team acknowledgments, planning artifacts, and completion artifacts to the goal document do not need to go through the formal vote process, and so lazy consensus rules will be used. If there is no objection posted one week after the most recent version of a change is proposed, then the change can be approved by the chair.
If a TC member feels that one of these responses needs to be discussed by the entire TC, they should bring it up on the mailing list and the change should not be approved until after the discussion is completed.
Appointing Project Leaders/Liaisons
In a resolution regarding leaderless programs
, the TC was granted authority to
appoint the leader to any official project where the election
process failed to produce a leader. The TC preference will be to move
such project to Distributed Project Leadership model (see /resolutions/20200803-distributed-project-leadership
)
than appointing the PTL. When this happens,
reference/projects.yaml
in the governance
repository should be updated to indicate the DPL model leadership
liaisons or the new PTL and their appointment by adding their name and
contact details and updating an appointed
key with the
cycle during which they will be the PTL. If the appointed
key is already present, add the cycle to the list. If the key is not
present, add it and set the cycle as a single member of a list. This
format is used for two reasons: to track all the cycles for which there
has been an appointment and to require a comprehensible change for
review by the TC. The appointed
key should only be changed
when the PTL was not chosen by the election process.
In the case of a project using, or moving to, Distributed Project
Leadership model (see /resolutions/20200803-distributed-project-leadership
),
the TC will make sure all the mandatory liaisons in the DPL model are
assigned.
These changes are subject to the standard review and approval guidelines.
Rolling back fast-tracked changes
As a safety net, if any member disagrees with any change that was fast-tracked under one of those house rules, that member can propose a revert of the change. Such revert should be directly approved by any TC member (who is not the proposer) and the change be discussed on the mailing list or on the re-proposed change in gerrit.
Voting on Changes in openstack/governance
TC member should use their RollCall-Vote permissions on all patches. Code-Review votes are ignored for the purposes of tallying votes, regardless of the content of the patch.
In the course of evaluating alternatives for complex proposals, we often ask one TC member to write several patches that might be mutually exclusive so that the committee can compare them and select one by voting on them independently. Because of this, we need to ensure that it is clear which patch the author of the patches prefers, and so we usually ask all TC members to cast a vote on all patches, even those they write.