Re: [BuildStream] Proposal: spreading responsibility and maintainership [Was: 1.4 Release timeline]

[Tristan Van Berkom <tristan vanberkom codethink co uk>]

Hi,

> On Dec 12, 2018, at 7:48 PM, Laurence Urhegyi <laurence urhegyi codethink co uk> wrote:
>
> > On 2018-12-12 10:06, Tristan Van Berkom wrote:
> > Essentially my concern is about accountability; it's very easy for
> > someone to throw out there that they "Approve" something if they are
> > not on the line for merging it, and multiple "Approvals" don't
> > necessarily mean someone has reviewed it thoroughly and really is
> > confident enough to hit "Merge".
> > The way you seem to have it setup in BuildGrid, where only one approval
> > is required, the approval would *be* the review, and the merge is not a
> > consequence of a mob, but of a single reviewer who assumes full
> > accountability for the review, which is better I think.
>
> I understand this concern, of course.
>
> It boils down to trust. By hand picking the list of approvers we trust them to be qualified to do the 
> review and only apply an approval when they really feel the patch is good (I have a patch in BuildStream, 
> but I should not be approving patches for a refactoring of the scheduler, for example :)) And of course 
> people we trust to be a responsible contributor: eg. be involved in a post-mortem of the patch if something 
> transpires to have broken because of it, and help identify where that issue lies, etc etc (not a blame 
> game, of course:)).
>
> I think our list of trusted committers (as per the subversion policy) should be the same the list as our 
> approvers on gitlab.

While it has to do with trust, its not exactly this.

With a single approver things are improved, what I want to avoid is having patches landed with a mob 
mentality, where everyone who approves "shares in the accountability", and consequently nobody is 
specifically responsible.

Cheers,
    -Tristan