Open Stack

Jeremy Stanley wrote:
> On 2016-04-08 14:04:15 +0000 (+0000), gordon chung wrote:
> [...]
>> understandably it's more difficult to manage a large group of
>> cores, but it does seem strange that all the projects have roughly
>> the same core team size even though some projects have a
>> contributor base that is significantly larger than others.
>
> That may be a sign that something like Dunbar's number is at work
> limiting group sizes. Crunching Gerrit approver group membership
> counts across 566 official Git repos (those belonging to a TC
> recognized project team), 90% have 17 or fewer accounts capable of
> approving changes. Below that there's a pretty varied spread (with
> 23% of repos having only 1 approver). This is a pretty raw analysis
> though, so something taking into account how many of those reviewers
> regularly exercised their approval access may yield different
> insights entirely.

Yes, there is a limit here, and it's got to do with trust. To grow a 
core reviewer group efficiently you need to trust that everyone in the 
group would still review and approve with the same rules you would 
apply. It's just difficult to know 20 people enough to trust them to do 
just that. That is the reason why core reviewers groups rarely grow past 
17-18 people, and that is the reason why you need to split larger bodies 
of codes into smaller review areas handled by separate groups.

-- 
Thierry Carrez (ttx)