Inactive OWNERS cleanup

[Kentaro Hara]

Hi

As of 2022 July, Chromium has 4531 OWNERS files containing 6850 names. These include inactive owners, which are one of the sources of slow code review latency. One year ago, we cleaned up inactive owners and removed ~500 inactive owners. I propose running the clean-up process again to keep the OWNERS files updated.

Specifically, a person is identified as an "inactive" owner iff:

• The person didn't commit or review any CLs in the directory they own while there were 100+ CLs that touched the directory in the past 6 months (as of July 6, 2022).

Last year, I gave the inactive owners an option to flip the decision manually to stay as an owner, but for this cycle, I'm planning to remove the inactive owners unconditionally. The rationale is 1) if the person made no contribution on a very active directory for 6 months, it will be reasonable to say that the person is inactive, and 2) if there is any special reason for it and the person needs to stay as an owner, the person can show evidence that they are meeting the owners expectations and be readded through the standard OWNERS nomination process.

Specifically, people listed in this spreadsheet are identified as inactive owners and will be removed.

I understand this is a tricky proposal. Having your name on OWNERS is an award for your previous amazing contributions, and I understand your feeling about your name being removed. However, I think it's important to keep the OWNERS files updated so that Chromium developers can find active owners and improve the code review latency.

If you have any questions / concerns, please let me know. Thanks!

--

Kentaro Hara, Tokyo

[dan...@chromium.org]

This list includes per-file owners, did the script look for 100 CLs in those files named by the rule when deciding to remove the person?

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CABg10jyArLjDp0ixPu%2BCSZ9NVrn0M1GwNFiJqiPGRE1f0mrbfQ%40mail.gmail.com.

[Ken Russell]

I echo Dana's concern about removing per-file owners and would like to see that policy rethought. Agree with Peter's observations as well.

-Ken

On Wed, Jul 27, 2022 at 9:12 AM Peter Boström <pb...@chromium.org> wrote:

I'm worried that this process excludes/penalizes folks who may be OOO for extended leave (incl long stretches of parental leave, bereavement) and have that in their Gerrit status. This should not be a source of review latency, if it is Gerrit should better surface that they are OOO.

Are any of the inactive owners, who did opt out last time, a source of review latency? I.e. are reviews assigned to them but they don't review them within some SLO window? Otherwise I strongly suggest we let folks decline the OWNERS removal (at other OWNERS' discretion who should probably review removal CLs).

--
--
Chromium Developers mailing list: chromi...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-dev
---
You received this message because you are subscribed to the Google Groups "Chromium-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chromium-dev...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/chromium-dev/CAHtyhaTNC4tgQbqbUq%2BQdFfcORr3aFobjgbeE%2BTaVf7eDgU2Bg%40mail.gmail.com.

--
--
Chromium Developers mailing list: chromi...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-dev
---
You received this message because you are subscribed to the Google Groups "Chromium-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chromium-dev...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/chromium-dev/CAGFX3sFB9G8R2MyHT6rjVtEFRAKMeyCTH6Yu0DYqUOfLPCxCBw%40mail.gmail.com.

[Matt Menke]

Maybe it would make more sense to identify OWNERS who are not active globally in chrome/, instead of owners not active in a particular directory?  How common are OWNERS active in Chrome, but high latency only for specific directories?  I'm asking as someone who was recently inundated by auto-generated removal CLs, the majority of which did not make sense (admittedly, I believe it wasn't based on activity).  The tool even seemed to want to remove all owners from some directories.

[Pavel Feldman]

The data in the table seems off, what is considered a "review": is that a "Code Review +1" or is that any review comment?
I also have an edge case where I'm mostly interested in several files in a folder where other files are being changed more frequently, should I be optimizing OWNERS to list myself as per-file?

[Glen Robertson]

> Having your name on OWNERS is an award for your previous amazing contributions

I'm concerned that being in OWNERS is regarded as a reward, and being removed as a penalty -- it is part of the problem with cleaning up inactive OWNERS. I'd much prefer to have a separate "amazing contributors" file to list people who have made amazing contributions, without this affecting the code review process.

Owners are supposed to be active reviewers for a directory. I'd even expect us to remove people who go on long leave, unless Gerrit can understand that status and avoid suggesting them for reviews (currently it does not do that well). Re-adding an owner is not an arduous process, but adding days to a code review is a significant cost.

I recently tried a similar automated audit of inactive owners - I looked for anyone who hadn't reviewed or authored a CL in 12 months anywhere, regardless of activity in the directory and found (as list, Google internal only) many accounts that no longer exist (or perhaps never did) in OWNERS. It probably has different false positives than the proposed set above. Maybe the intersection of the two sets would be sensible?

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/0a2a01e2-652b-4e31-895c-f020e7b46358n%40chromium.org.

[Kentaro Hara]

Thanks all for the input!

Dana:

This list includes per-file owners, did the script look for 100 CLs in those files named by the rule when deciding to remove the person?

Thanks for pointing this out! I'll exclude per-file owners from the list for now.

Peter:

I'm worried that this process excludes/penalizes folks who may be OOO for extended leave (incl long stretches of parental leave, bereavement) and have that in their Gerrit status. This should not be a source of review latency, if it is Gerrit should better surface that they are OOO.
Are any of the inactive owners, who did opt out last time, a source of review latency? I.e. are reviews assigned to them but they don't review them within some SLO window? Otherwise I strongly suggest we let folks decline the OWNERS removal (at other OWNERS' discretion who should probably review removal CLs).

I think Glen covered this topic very well. As written in this guideline, owners are expected to be an active contributor to the directory ("Have the bandwidth to contribute to reviews in a timely manner. ... Don't try to discourage people from sending reviews, including writing “slow” or “emeritus” after your name."). If you are on an extended leave and removed by this process, you can explain it and re-add yourself through the owner nomination process. Will it work?

Matt:

Maybe it would make more sense to identify OWNERS who are not active globally in chrome/, instead of owners not active in a particular directory?  How common are OWNERS active in Chrome, but high latency only for specific directories?

My personal opinion is that owners who made no contributions globally in the past 6 months *or* owners who made no contribution to the directory they own while there were 100+ commits in the past 6 months can be identified as inactive owners.

Note that this is not an irreversible process. When you have a reason, you can explain it and re-add yourself through the owner nomination process.

 I'm asking as someone who was recently inundated by auto-generated removal CLs, the majority of which did not make sense (admittedly, I believe it wasn't based on activity). The tool even seemed to want to remove all owners from some directories.

Right, the removal tool is not looking at activities, and this proposed process is different from it. FWIW when I removed ~500 inactive owners last year, it ended up removing (only) ~10 OWNERS files. So removing all owners from some directories will be rare.

Pavel:

The data in the table seems off, what is considered a "review": is that a "Code Review +1" or is that any review comment?

"Code Review +1" in the git commit log is considered as "review".

 

I also have an edge case where I'm mostly interested in several files in a folder where other files are being changed more frequently, should I be optimizing OWNERS to list myself as per-file?

This sounds reasonable to me :)

Glen:

I recently tried a similar automated audit of inactive owners - I looked for anyone who hadn't reviewed or authored a CL in 12 months anywhere, regardless of activity in the directory and found (as list, Google internal only) many accounts that no longer exist (or perhaps never did) in OWNERS. It probably has different false positives than the proposed set above. Maybe the intersection of the two sets would be sensible?

I'm happy to tweak the criteria depending on the conclusion of this email thread :)

--

Kentaro Hara, Tokyo

[Ali Juma]

On Thu, Jul 28, 2022 at 6:07 AM Kentaro Hara <har...@chromium.org> wrote:

Thanks all for the input!

Dana:

This list includes per-file owners, did the script look for 100 CLs in those files named by the rule when deciding to remove the person?

Thanks for pointing this out! I'll exclude per-file owners from the list for now.

Peter:

I'm worried that this process excludes/penalizes folks who may be OOO for extended leave (incl long stretches of parental leave, bereavement) and have that in their Gerrit status. This should not be a source of review latency, if it is Gerrit should better surface that they are OOO.
Are any of the inactive owners, who did opt out last time, a source of review latency? I.e. are reviews assigned to them but they don't review them within some SLO window? Otherwise I strongly suggest we let folks decline the OWNERS removal (at other OWNERS' discretion who should probably review removal CLs).

I think Glen covered this topic very well. As written in this guideline, owners are expected to be an active contributor to the directory ("Have the bandwidth to contribute to reviews in a timely manner. ... Don't try to discourage people from sending reviews, including writing “slow” or “emeritus” after your name."). If you are on an extended leave and removed by this process, you can explain it and re-add yourself through the owner nomination process. Will it work?

The next guideline (on removal of owners) explicitly excludes owners who are on leave. I don't think we should be adding additional friction for folks who go on leave; the default assumption should be that when they return, they are just as capable of being a good owner as when they left, without them having to re-nominate themselves.

 

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CABg10jyCNN1%3DpfT%3DCWPmc4%2Bi9PmGs-%3DbX9e2mUi2bHthF%2B0w-w%40mail.gmail.com.

[Matt Menke]

On Thu, Jul 28, 2022 at 10:29 AM Ali Juma <aj...@chromium.org> wrote:

On Thu, Jul 28, 2022 at 6:07 AM Kentaro Hara <har...@chromium.org> wrote:

Thanks all for the input!

Dana:

This list includes per-file owners, did the script look for 100 CLs in those files named by the rule when deciding to remove the person?

Thanks for pointing this out! I'll exclude per-file owners from the list for now.

Peter:

I'm worried that this process excludes/penalizes folks who may be OOO for extended leave (incl long stretches of parental leave, bereavement) and have that in their Gerrit status. This should not be a source of review latency, if it is Gerrit should better surface that they are OOO.
Are any of the inactive owners, who did opt out last time, a source of review latency? I.e. are reviews assigned to them but they don't review them within some SLO window? Otherwise I strongly suggest we let folks decline the OWNERS removal (at other OWNERS' discretion who should probably review removal CLs).

I think Glen covered this topic very well. As written in this guideline, owners are expected to be an active contributor to the directory ("Have the bandwidth to contribute to reviews in a timely manner. ... Don't try to discourage people from sending reviews, including writing “slow” or “emeritus” after your name."). If you are on an extended leave and removed by this process, you can explain it and re-add yourself through the owner nomination process. Will it work?

The next guideline (on removal of owners) explicitly excludes owners who are on leave. I don't think we should be adding additional friction for folks who go on leave; the default assumption should be that when they return, they are just as capable of being a good owner as when they left, without them having to re-nominate themselves.

The assumption is not that they're no longer good owners, but that people shouldn't have to spend a week waiting on them to reply to a review before giving up and trying someone else.  Note that owners do not need to be nominated - no one is losing their committer status.

 

 

Matt:

Maybe it would make more sense to identify OWNERS who are not active globally in chrome/, instead of owners not active in a particular directory?  How common are OWNERS active in Chrome, but high latency only for specific directories?

My personal opinion is that owners who made no contributions globally in the past 6 months *or* owners who made no contribution to the directory they own while there were 100+ commits in the past 6 months can be identified as inactive owners.

Note that this is not an irreversible process. When you have a reason, you can explain it and re-add yourself through the owner nomination process.

I'm not concerned about the fact that it's not reversible, but wasting time.  I received ~10 emails from the last mass owners purge, and Not LGTMing a bunch of CLs created by a tool following completely opaque rules seemed not a productive use of time.

[K. Moon]

I think our owner-suggesting tooling really needs to be better about deprioritizing owners who are on leave (as well as better load balancing). I'm not sure periodically removing and re-adding owners is the right long-term fix, but it could be acceptable if we're working on better infrastructure in the meanwhile.

I think another good, oft-discussed improvement would be to separate Review+1 from Owners+1. This would make the review load for owners more manageable, as they can just say that they approve of the overall change, without also marking every file they own as reviewed (and all that implies).

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAEK7mvr5vjo0FdbYepkOLBDGJN7KLivmauf4G-HQ1rf67f7pSQ%40mail.gmail.com.

[K. Moon]

On a related topic, one challenge that I have with OOO status is that it's hard to update all the places where you might want to note that (each Gerrit repository, Slack, external Gmail, internal Gmail, Calendar, ...). For Googlers, I have decent tooling for checking whether they're OOO, in my time zone, or whatever (although better integration would save me time), but if I wasn't a Googler (and didn't have access to those tools), or I want to check the status of a non-Googler, it's often not as easy, or impossible.

It'd be great if there were One Tool that (opt-in) tracked OOO status in a centralized directory, and worked whether or not you had a Google or Chromium account.

For parts of the tree that need to balance between a large number of reviewers, something like gwsq probably would be a decent solution. (//ipc/SECURITY_OWNERS is using this now; I think trees like //content would benefit as well.)

On Thu, Jul 28, 2022 at 8:56 AM Peter Boström <pb...@chromium.org> wrote:

I agree with the policy to not remove owners because they are on leave. Owners plus OOO status should inform owner selection and we still need that because we shouldn't remove owners while they are out for a week but they sure shouldn't be used during that week.

If OOO is an efficient signal this should be sufficient without force removing folks on leave. If OOO isn't a sufficient signal and introduces review latency we should look at why that is instead and fix our tooling.

If I may spitball another reason for latency we should maybe also surface how long folks review queues are and perhaps use that to help guide owners selection.

Also because I'm already up on my soapbox, if you're fast and not overloaded and want to help with the review load, please put something like "fast" or "more reviews plz" in your Gerrit status to encourage more reviews your way (me inspired by ellyjones@ here). Encourage others to do the same and reward this as a community contribution.

[Kentaro Hara]

+1 to improving the tool to handle OOO owners.

Let's step back. The problem I'm solving now is: currently OWNERS files contain many long-time inactive, non-OOO owners. The goal is to remove them. Do you agree with the goal?

The "long-time inactive" part is covered by looking into their review / commit activities in the past 6 months.

The "non-OOO" part is tricky because currently there is no way to get the information from the tool. I'm not sure if giving an option to manually flip the decision is a great solution because if they are long-time OOO, they are not likely to be looking at this email and get removed anyway. Then I thought that it's more reasonable to execute the clean-up process unconditionally while explicitly saying that they can explain the situation and re-add themselves when they are back (where they can refer to this email thread).

What do you think?

--

Kentaro Hara, Tokyo

[Glen Robertson]

+1 to needing better tools for handling OOO, and big +1 to using gwsq when there are many reviewers to balance.

Given that long leave is potentially 6 months, but unlikely to be 12 months (citation needed), how about changing the criteria to 12 months? Then it sort of covers both "long-tme inactive" and "non-OOO". Regardless, until we have better tooling for OOO, I think it's reasonable to remove people who are currently inactive and they can be re-added later.

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CABg10jwB528AFfwArSHLKoHWeWt-JOvCS0dnFJNaDycKSc5GDw%40mail.gmail.com.

[Manuel Rego Casasnovas]

On 28/07/2022 16:33, 'Matt Menke' via Chromium-dev wrote:
> The assumption is not that they're no longer good owners, but that
> people shouldn't have to spend a week waiting on them to reply to a
> review before giving up and trying someone else.  Note that owners do
> not need to be nominated - no one is losing their committer status.

Just for completeness, there are some OWNERS files that require a
nomination process. At least third_party/blink/renderer/core/OWNERS
requires that:
https://source.chromium.org/chromium/chromium/src/+/main:third_party/blink/renderer/core/OWNERS

Cheers,
Rego

[Adam Rice]

+1 to removing inactive OWNERS. It should not require insider knowledge to find someone to review your CL.

While it would be great if our tools were better at handling OOO OWNERS, the current state of affairs is that they aren't. It is easy enough for someone to re-add themselves to OWNERS when coming back from a long leave, and I don't think it is unreasonably burdensome.

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CABg10jwB528AFfwArSHLKoHWeWt-JOvCS0dnFJNaDycKSc5GDw%40mail.gmail.com.

[Peter Boström]

I'm worried that this process excludes/penalizes folks who may be OOO for extended leave (incl long stretches of parental leave, bereavement) and have that in their Gerrit status. This should not be a source of review latency, if it is Gerrit should better surface that they are OOO.

Are any of the inactive owners, who did opt out last time, a source of review latency? I.e. are reviews assigned to them but they don't review them within some SLO window? Otherwise I strongly suggest we let folks decline the OWNERS removal (at other OWNERS' discretion who should probably review removal CLs).

On Wed, Jul 27, 2022 at 8:08 AM <dan...@chromium.org> wrote:

--

--
Chromium Developers mailing list: chromi...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-dev
---

You received this message because you are subscribed to the Google Groups "Chromium-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chromium-dev...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/chromium-dev/CAHtyhaTNC4tgQbqbUq%2BQdFfcORr3aFobjgbeE%2BTaVf7eDgU2Bg%40mail.gmail.com.

[Peter Boström]

I specifically disagree with the "for this cycle, I'm planning to remove the inactive owners unconditionally" bit. If you suggest this removal but you defer to local OWNERS to make the call I think that's fine to send out removal CLs for. I would encourage you and/or the reviewers to look at the removed owner's calendar if they're internal and confirm that they've not been on long-term leave.

I don't think that 85/6850 OWNERS entries are a large enough problem to bypass local owner approval for this (i.e. land it unconditionally). If some of them are owners in very active folders and have had a lot of reviews assigned to them, sure. This is about 1% of ownership entries if I'm understanding this correctly. How badly are they contributing to review latency in order to override "and they are not on a leave during that time" as a policy? I like this policy, and it's supposed to still be the policy.

If any of these OWNERS entries had a significant amount of reviews assigned to them during this period (i.e. sorta-evidence of latency), you could use that as additional justification in the CLs to get the OWNERS to approve the (possibly temporary) removal.

For fixing review latency which is the real goal presumably, I think handling OOO and load balancing are both more important vectors here. We shouldn't need to get inactive owners down to 0 to solve review latency.

[Peter Boström]

I agree with the policy to not remove owners because they are on leave. Owners plus OOO status should inform owner selection and we still need that because we shouldn't remove owners while they are out for a week but they sure shouldn't be used during that week.

If OOO is an efficient signal this should be sufficient without force removing folks on leave. If OOO isn't a sufficient signal and introduces review latency we should look at why that is instead and fix our tooling.

If I may spitball another reason for latency we should maybe also surface how long folks review queues are and perhaps use that to help guide owners selection.

Also because I'm already up on my soapbox, if you're fast and not overloaded and want to help with the review load, please put something like "fast" or "more reviews plz" in your Gerrit status to encourage more reviews your way (me inspired by ellyjones@ here). Encourage others to do the same and reward this as a community contribution.

On Thu, Jul 28, 2022 at 8:15 AM K. Moon <km...@chromium.org> wrote:

[Kentaro Hara]

Discussed with Peter and Glen offline.

I excluded per-file owners and updated the spreadsheet (based on the data as of July 26 when I announced this).

To mitigate the concern about removing long-time OOO owners and making them feel uncomfortable about re-adding themselves when they are back, we decided to add the following paragraph to the owners guideline (which is under review):

"For the purpose of not slowing down code review latency, Chromium removes inactive owners (e.g., those who made no contributions in the past 6 months) on a regular basis. The script does not take into account personal situations like a long leave and thus can be wrong. If you are removed by the automated process for a wrong reason while you are being a good owner, you can create a CL to re-add yourself to OWNERS and land after getting local owner's approval. You just need to explain the reason (e.g., long leave) and don't need to make a substantial contribution to become an owner again."

Manuel:

Just for completeness, there are some OWNERS files that require a nomination process. At least third_party/blink/renderer/core/OWNERS requires that.

I hope the above process mitigates the concern.

Matt:

Maybe it would make more sense to identify OWNERS who are not active globally in chrome/, instead of owners not active in a particular directory?  How common are OWNERS active in Chrome, but high latency only for specific directories?

FWIW I created a list of OWNERS who made no commits / reviews in the past 6 months (as of July 26). We could take an intersection between this list and my proposed list but I'd prefer simply going with my proposed list with the above mitigation process.

Peter:

I don't think that 85/6850 OWNERS entries are a large enough problem to bypass local owner approval for this (i.e. land it unconditionally). If some of them are owners in very active folders and have had a lot of reviews assigned to them, sure. This is about 1% of ownership entries if I'm understanding this correctly. How badly are they contributing to review latency in order to override "and they are not on a leave during that time" as a policy? I like this policy, and it's supposed to still be the policy.

We are removing (only) 85 owners (47 owners after per-file owners are excluded) because we removed ~500 owners one year ago. I think it's important to run this cleanup process periodically to keep the codebase up-dated. Our Chrome infra team is now brainstorming an idea to automate the process of removing inactive owners.

I'm planning to execute the cleanup next week. If you have any questions, please let me know. Thanks!

--

Kentaro Hara, Tokyo

[K. Moon]

Given the amount of discussion this topic has generated, I almost wonder if the, "Let's just run this script and fix any problems later," approach is actually taking more time in total than something more nuanced.

Also, is there an AI for any team to follow up on some of the suggestions about surfacing OOO status, or other sources of high review latency?

On Mon, Aug 1, 2022 at 8:22 AM Christian Dullweber <dull...@google.com> wrote:

I looked into the first 10 entries manually and there are multiple cases where I don't think it is correct to remove them.

There are people who are on long term leave and are indicating this in their status message. 

I would be quite annoyed if I had to figure out all the files I was owner for and restore this when I come back from a leave. 

Could you check for each person you want to remove if the status message has an indication that the OWNER is only temporarily inactive?

There are also otherwise active owners who have a comment that limits their ownership to a subset of files.

I wonder why it would be useful to remove people who are otherwise active just because they haven't reviewed files in a certain directory?

There are also per-file owners but I guess the list just isn't updated yet?

As an alternative to implementing the suggestions above:

How about sending CLs with autosubmit enabled for each removal to all owners of the corresponding file and see if they agree with the removal or not? 

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/chromium-dev/CABg10jzMeS604sff1NNK4BK9E3A8A7rn3wxPiqB0G95%3D3xupgA%40mail.gmail.com.

[Christian Dullweber]

I looked into the first 10 entries manually and there are multiple cases where I don't think it is correct to remove them.

There are people who are on long term leave and are indicating this in their status message. 

I would be quite annoyed if I had to figure out all the files I was owner for and restore this when I come back from a leave. 

Could you check for each person you want to remove if the status message has an indication that the OWNER is only temporarily inactive?

There are also otherwise active owners who have a comment that limits their ownership to a subset of files.

I wonder why it would be useful to remove people who are otherwise active just because they haven't reviewed files in a certain directory?

There are also per-file owners but I guess the list just isn't updated yet?

As an alternative to implementing the suggestions above:

How about sending CLs with autosubmit enabled for each removal to all owners of the corresponding file and see if they agree with the removal or not? 

On Mon, 1 Aug 2022 at 15:56, Kentaro Hara <har...@chromium.org> wrote:

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/chromium-dev/CABg10jzMeS604sff1NNK4BK9E3A8A7rn3wxPiqB0G95%3D3xupgA%40mail.gmail.com.

[Jayson Adams]

This policy, despite the proposed disclaimer paragraph, still feels like it penalizes people who are on leave. Whether it's parental or disability or any other kind of leave, no one should have to jump through hoops upon their return to restart their job (or I guess why not also remove committer status, in the interest of making things tidy, while you're at it?).

Given that your updated list only contains 47 owners, it seems like checking each one to see if they're on leave (and removing them if so) would not be that big of an issue.

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/chromium-dev/CABg10jzMeS604sff1NNK4BK9E3A8A7rn3wxPiqB0G95%3D3xupgA%40mail.gmail.com.

[Gabriel Charette]

On Mon, Aug 1, 2022 at 11:22 AM 'Christian Dullweber' via Chromium-dev <chromi...@chromium.org> wrote:

I looked into the first 10 entries manually and there are multiple cases where I don't think it is correct to remove them.

There are people who are on long term leave and are indicating this in their status message. 

I would be quite annoyed if I had to figure out all the files I was owner for and restore this when I come back from a leave. 

Could you check for each person you want to remove if the status message has an indication that the OWNER is only temporarily inactive?

There are also otherwise active owners who have a comment that limits their ownership to a subset of files.

I wonder why it would be useful to remove people who are otherwise active just because they haven't reviewed files in a certain directory?

There are also per-file owners but I guess the list just isn't updated yet?

As an alternative to implementing the suggestions above:

How about sending CLs with autosubmit enabled for each removal to all owners of the corresponding file and see if they agree with the removal or not? 

We used to have documentation that required that all large-scale changes be `git cl split` unless 100% mechanical (Owners-Override), I can't find it anymore (lost in Style Guide / LSC documentation move to .MD?). It sounds like this isn't 100% mechanical and it thus makes a lot of sense to me that it'd at least require a local owner's approval.

 

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/chromium-dev/CAKAfQi2EPsAZ0FNGE%3DOTqd1ubradPHO%2Bi8q8S9hbrJzSZaD4MA%40mail.gmail.com.

[Andy Perelson]

Hi folks,

A fair bit of the feedback on Kentaro's proposal has focused on improving Owners suggestions and better handling OOO Owners. We do believe there's a lot of room to improve owners suggestions to help improve review time. OOO is one aspect, but you could also imagine such a system taking into account time zones, review loads, etc. It's an area we'd like to improve on but also a complicated one. I can't promise specific features or a timeline at this stage, but know that we hear and agree with the feedback that this would be a useful area to improve that could positively help review times. And one we've been thinking about for a while.

Thanks,

Andy

On behalf of the Chrome Ops Source team

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/chromium-dev/CA%2B4SKEKJsKET9S7LgyvhEoCBV85ynahAy6WeuB5pM7DEG_nB5w%40mail.gmail.com.

[Kentaro Hara]

Hi

Thanks all for the thoughts! I discussed with some of the people on this thread and Eng Review offline.

It's important to keep OWNERS files up-dated. OWNERS should include people who are meeting the owner's expectations ("Have committed or reviewed substantial work to the affected directory within the last ninety days", "Have the bandwidth to contribute to reviews in a timely manner"). The consumers of OWNERS shouldn't need to care what's happening with individuals on that list, and should be able to get useful, fast turnaround time from them. I get the argument that we shouldn't penalize folks on leaves, however, we also have to make sure that the system doesn't punish everybody by having a number of absent owners, so I'm still in favor of removing absent owners and we'd just add them back in when they're back and active again.

On the other hand, I agree with the feedback that removing inactive owners unconditionally is too aggressive (thanks for the feedback!). So I propose going with the following plan:

1) Add this policy to code_reviews.md. The policy says that if you are removed while you were on a leave, you can re-add yourself when you are back.

2) Use the (already existing) automated owner removal tool and let it create a CL to remove owners who made 0 commits / reviews to Chromium in the past 12 months (2021/7/26 - 2022/7/26). This will be a strong signal to indicate inactive owners. There are 320 owners.

3) The CL adds existing owners of the directory (including the owner to be removed) to a reviewer list. The CL description clearly asks the reviewers to check that the removed owner is not on a leave. The CL is committed after getting an LGTM from one of the reviewers.

4) If an owner is mistakenly removed while they are on a leave (which is not likely to happen due to 3)), they can re-add themselves according to the policy.

It's possible that some owners made 0 commits / reviews to Chromium in the past 12 months because their directories are under maintenance mode and didn't have many CLs. In this case, they (or other owners in the directory) can say Code-Review -1 in 3).

If you have any concerns, please let me know.

Thanks!

--

Kentaro Hara, Tokyo