Open Stack

On 10/08/2013 11:22 AM, Clint Byrum wrote:
> I don't meant to pick on you personally Jiří, but I have singled this
> message out because I feel you have captured the objections to Robert's
> initial email well.

Darn. My pop-up window showed "I don't meant to pick on you personally"
so I rushed to go read the message, and it turns out to be reasonable
and not ranty.

> Excerpts from Jiří Stránský's message of 2013-10-08 04:30:29 -0700:
>> On 8.10.2013 11:44, Martyn Taylor wrote:
>>> Whilst I can see that deciding on who is Core is a difficult task, I do
>>> feel that creating a competitive environment based on no. reviews will
>>> be detrimental to the project.
>>>
>>> I do feel this is going to result in quantity over quality. Personally,
>>> I'd like to see every commit properly reviewed and tested before getting
>>> a vote and I don't think these stats are promoting that.
>>
>> +1. I feel that such metric favors shallow "i like this code"-reviews as 
>> opposed to deep "i verified that it actually does what it 
>> should"-reviews. E.g. i hit one such example just today morning on 
>> tuskarclient. If i just looked at the code as the other reviewer did, 
>> we'd let in code that doesn't do what it should. There's nothing bad on 
>> making a mistake, but I wouldn't like to foster environment of quick 
>> shallow reviews by having such metrics for core team.
>>
> 
> 
> I think you may not have worked long enough with Robert Collins to
> understand what Robert is doing with the stats. While it may seem that
> Robert has simply drawn a line in the sand and is going to sit back and
> wait for everyone to cross it before nominating them, nothing could be
> further from the truth.
> 
> As one gets involved and start -1'ing and +1'ing, one can expect feedback
> from all of us as core reviewers. It is part of the responsibility of
> being a core reviewer to communicate not just with the submitter of
> patches, but also with the other reviewers. If I see shallow +1's from
> people consistently, I'm going to reach out to those people and ask them
> to elaborate on their reviews, and I'm going to be especially critical
> of their -1's.
> 
>> I think it's also important who actually *writes* the code, not just who 
>> does reviews. I find it odd that none of the people who most contributed 
>> to any of the Tuskar projects in the last 3 months would make it onto 
>> the core list [1], [2], [3].

I believe this is consistent with every other OpenStack project. -core
is not a status badge, nor is it a value judgement on the relative
coding skills. -core is PURELY a reviewing job. The only thing is grants
is more weight to the reviews you write in the future, so it makes
perfect sense that it should be judged on the basis of your review work.

It's a mind-shift to make, because in other projects you get 'committer'
access by writing good code. We don't do that in OpenStack. Here, you
get reviewer access by writing good reviews. (this lets the good coders
code and the good reviewers review)

> I think having written a lot of code in a project is indeed a good way
> to get familiar with the code. However, it is actually quite valuable
> to have reviewers on a project who did not write _any_ of the code,
> as their investment in the code itself is not as deep. They will look
> at each change with fresh eyes and bring fewer assumptions.
> 
> Reviewing is a different skill than coding, and thus I think it is o-k
> to measure it differently than coding.
> 
>> This might also suggest that we should be looking at contributions to 
>> the particular projects, not just the whole program in general. We're 
>> such a big program that one's staleness towards some of the components 
>> (or being short on global review count) doesn't necessarily mean the 
>> person is not important contributor/reviewer on some of the other 
>> projects, and i'd also argue this doesn't affect the quality of his work 
>> (e.g. there's no relationship between tuskarclient and say, t-i-e, 
>> whatsoever).
>>
> 
> Indeed, I don't think we would nominate or approve a reviewer if they
> just did reviews, and never came in the IRC channel, participated in
> mailing list discussions, or tried to write patches. It would be pretty
> difficult to hold a dialog in reviews with somebody who is not involved
> with the program as a whole.

For projects inside of openstack-infra (where we have like 30 of them or
something) we've added additional core teams that include infra-core but
have space for additional reviewers. jenkins-job-builder is the first
one we did like this, as it has a fantastically active set of devs and
reviewers who solely focus on that. However, that's the only one we've
done that for so far.

>> So i'd say we should get on with having a greater base of core folks and 
>> count on people using their own good judgement on where will they 
>> exercise their +/-2 powers (i think it's been working very well so far), 
>> or alternatively split tripleo-core into some subteams.
>>
> 
> If we see the review queue get backed up and response times rising, I
> could see a push to grow the core review team early. But we're talking
> about a 30 day sustained review contribution. That means for 30 days
> you're +1'ing instead of +2'ing, and then maybe another 30 days while we
> figure out who wants core powers and hold a vote.
> 
> If this is causing anyone stress, we should definitely address that and
> make a change. However, I feel the opposite. Knowing what is expected
> and being able to track where I sit on some of those expectations is
> extremely comforting. Of course, easy to say up here with my +2/-2. ;)
> 
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>