[Development] Proposal: Allow contributors to +1 sanity review.
oswald.buddenhagen at digia.com
Fri Aug 16 12:33:11 CEST 2013
i'll answer only the part that andre didn't cover.
On Thu, Aug 15, 2013 at 11:44:31AM -0700, Thiago Macieira wrote:
> On quinta-feira, 15 de agosto de 2013 18:32:09, Oswald Buddenhagen wrote:
> > > and even if it is abused, it cannot override the normal review
> > > process.
> > override, no. but people are easy to lead to wrong conclusions,
> > and tend to trust previous comments without checking themselves. having
> > to the check the box themselves is at least a minor wake-up call.
> So you're saying that a non-approver who did a +1 will keep doing +1 because
> he had done +1 before? In other words, "a reviewer who has bad behaviour will
> have bad behaviour"?
> Or are you saying that someone will override the bot because someone else has
> overridden the bot in a past occasion? If this is the one, I don't buy it.
> It's like saying "Thiago has approved contributions in the past, so I will
> approve too without thinking". People are supposed to use the grey matter
> between their ears, not follow blindly other people's behaviours.
> And if there are people with bad behaviour (whose behaviour may be copied or
> not), we need to nip that in the bud. That's not an argument for the sanity
i'm not talking about setting bad precedents or any other "cultural"
effects that span across reviews. it's all about the scope of one single
it's a simple neurological fact that the human brain can be mislead to
wrong conclusions via "resonance". that's why an external reviewer is
much more effective at finding mistakes than a person one closely worked
with to create a work in the first place.
so the hypothesis is that if a contributor overrides the bot with a
bogus justification, there is a non-trivial probability that the
approver will simply accept it ("it's been dealt with, and it looks
kinda reasonable"). letting the approver (somebody who is supposed to be
familiar with the rules for overriding) do the sanity evaluation "from
scratch" is therefore more likely to lead to a correct conclusion. there
is of course still the possibility that the contributor writes a
justification and the approver simply accepts it and checks the box, but
my expectation is that the need to actually check that box is at least
*a bit* of a trigger to re-think the argument of the contributor.
from that also follows that this is a much less relevant consideration
when the contributor is an approver himself (i.e., the vast majority of
the contributions, in particular the ones where timing is even remotely
though i also think that doing sanity overrides for the own changes is
bad style, because one is always biased.
More information about the Development