Hi Pedro, Thanks I have also worked with code owners, in the past. I guess it would be up to committers as to whether this is an approach that would work for them – I guess we are talking about introducing a CODEOWNERs file into the core Flink repo. Is there an appetite for this and potential subsequent automation from the community,
Kind regards, David. From: Pedro Mázala <pedroh.maz...@gmail.com> Date: Wednesday, 19 March 2025 at 10:38 To: dev@flink.apache.org <dev@flink.apache.org> Subject: [EXTERNAL] Re: Community review process improvement proposal Any time. More than happy to help in any way I can. On the subscription idea, I tackled this on companies I worked with before by using codeowners. We could have certain "groups" where the committers could join and then get notified/assigned. But rather than directly assigning, we could notify the committer about it. This could even be associated with the tag `community-reviewed-LGTM`, and the committer of a certain component gets notified only then. Perhaps it's worth checking how other communities are handling such cases. I'll try to gather some of this information and be present at the next CHI meeting. Att, Pedro Mázala Be awesome On Wed, 19 Mar 2025 at 10:03, David Radley <david_rad...@uk.ibm.com> wrote: > Hi Pedro, > Thank you very much for you feedback on this. > > - About the tag `community-reviewed-required-deep-review`, would it be > manually added or something @FlinkBot would be capable of doing? This would > be added via the Flink bot command ,as non committers do not have the > authority to add a label. > - How the committers/PMCs could rely on a `community-reviewed-LGTM` tag > without needing to double-check the attribution? Yes they may want to check > the attribution, I see this as a hint that there are community members > thinking this is ready to merge, they could be wrong, so the committer > still needs to be confident to merge. I would like to think the majority > would just be merged quickly, especially smaller changes. Bigger changes > are likely to require a deep review, but in my experience a lot of feedback > can be requested by the community reviewer – questions that the committer > does not need then to ask. > - Committers/PMCs could have a way to subscribe to paths that they'd like > to review and get notified. Nice idea – it would be nice for committer (or > other) to be able to subscribe to a set of components they are interested > in. I am not sure how this could be implemented – we are thinking of Git > actions and flink bot as implementation options – but I am not sure where > the appropriate place to store emails would be. Open to suggestions. > - How could the community also help with creating consensus when needed? > Is this something the CHI wants to tackle in the future? Existing Apache > processes around voting and dev list discussion should be able to resolve > these situations. If there was an issue with consensus I suspect the PMC > would be involved. If someone wants to talk about a specific consensus > issue in the CHI meeting – that would be fine, I would be encouraging them > to use the usual Apache processes- though maybe we could nudge individuals > to get open discussions going. > > Kind regards, David > > From: Pedro Mázala <pedroh.maz...@gmail.com> > Date: Tuesday, 18 March 2025 at 12:13 > To: dev@flink.apache.org <dev@flink.apache.org> > Subject: [EXTERNAL] RE: Community review process improvement proposal > I think this is a great way of keeping things tidy. Thank you for the work > being done here. > > I have some thoughts to share: > - About the tag `community-reviewed-required-deep-review`, would it be > manually added or something @FlinkBot would be capable of doing? > - How the committers/PMCs could rely on a `community-reviewed-LGTM` tag > without needing to double-check the attribution? > - Committers/PMCs could have a way to subscribe to paths that they'd like > to review and get notified. > - How could the community also help with creating consensus when needed? > Is this something the CHI wants to tackle in the future? > > > > Att, > Pedro Mázala > > On 2025/03/17 17:13:31 David Radley wrote: > > Hi , > > In the last Community Health Initiative< > https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=345377343><https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=345377343%3e> > we talked of improving the Flink process to involve development review. > > @Robert Metzger<ma...@apache.org> suggested I raise a Flip for this. > Prior to raising the Flip I would like to have a conversation on the dev > list to see what people thought and assess the support for this proposed > change. > > > > Motivation for flip > > There are many Flink contributors and fewer committers. As a community > we want to drive down our technical debt. The committers can be time > strapped, so may not have enough time to review every PR that comes in. > There are members of the community that are willing to review PRs and > appropriately share the review burden. This activity has been occurring > under the Community Health Initiative, where every new PR since the start > of the group has been reviewed / triaged. This proposal is to formally > create a process where the community (non-committers) can review PRs and > ease the load on the committers. The benefits of this approach are: > > > > * we formally encourage the community to review PRs - making this > one of the things we do as a community. > > * encouraging the community to review will get more eyes on code > changes. > > * easing the burden of review for committers > > * for straight forward PRs that the community approves of > > * identify PRs if they need expert assessment > > * community reviewing then becomes a way to contribute to Flink on > the road to becoming a committer. So, it is in the contributor’s interest > to review > > * measuring community review activity gives us metrics to show its > impact and whether the process is working. > > Proposal for Flip > > Use Flinkbot commands to add new labels to indicate that the community > has reviewed a PR. > > Suggested new labels > > > > * community-reviewed-LGTM > > > > This can be set if there are 2 committer approves > > > > * community-reviewed-required-deep-review > > This could be set if 2 community members agree that a deep review is > required. > > > > * Community-health-initiative-reviewed > > > > A tag to indicate that the Community Health Initiative has reviewed the > PR > > The by-produce of this process is that sanity checks (does the Jira have > a title – look right, does it have unit tests, code logic tests, does the > Jira have a decent description on what is being changed and why etc) will > occur on PRs generating lots of prompt feedback to the submitters of the > PRs. Also it is easier for committers to identify what PRs to review and > merge easy changes . > > In the Community Health Initiative, we hope to move as much of the > sanity checking as possible to the Flink bot, where it can be automated. > > One concern about this approach is that someone could unethically put > the community review labels onto PRs without having reviewed the PR. If we > see this is occurring, then we would use usual Apache processes to deal > with members of the community not behaving well. Any thoughts on this? > > I am interested in what the community thinks about this idea, and will > raise a Flip to formally discuss and vote on if there is a support for it, > > Kind regards, David. > > > > > > > > > > > > > > Unless otherwise stated above: > > > > IBM United Kingdom Limited > > Registered in England and Wales with number 741598 > > Registered office: Building C, IBM Hursley Office, Hursley Park Road, > Winchester, Hampshire SO21 2JN > > > > Unless otherwise stated above: > > IBM United Kingdom Limited > Registered in England and Wales with number 741598 > Registered office: Building C, IBM Hursley Office, Hursley Park Road, > Winchester, Hampshire SO21 2JN > Unless otherwise stated above: IBM United Kingdom Limited Registered in England and Wales with number 741598 Registered office: Building C, IBM Hursley Office, Hursley Park Road, Winchester, Hampshire SO21 2JN
