launchpad-dev team mailing list archive

[Robert Collins <robertc@xxxxxxxxxxxxxxxxx>]

This had stalled, so I spoke with Francis about it.

On Fri, Oct 22, 2010 at 4:02 PM, Robert Collins
<robert.collins@xxxxxxxxxxxxx> wrote:
> Specific proposed changes to the experiment:
>  - allow anyone to have optional reviews
>   This could be interesting, but I think it takes a while to gain enough
>   knowledge of the system to know just how much you don't* know.
>   Maybe in a future iteration.

We agreed this is a future thing to examine.

>  - allow reviewers to have optional reviews
>  I proposed 3 months because I think in that time period one will
>  learn enough to know when one knows and when one doesn't: that's the
>  key metric that matters. I think that someone who isn't a reviewer but
>  has been landing regular branches for 3 months is certainly past this
>  level of knowledge. I'd *like* us to use 3 months, but if it makes the
>  difference between doing/not doing the experiment, then 'is a full
>  reviewer' would be tolerable.

We agreed it will be reviewers only that participate in this experiment.

>  - length
>  until the epic sounds fine
>  - post-landing evaluations per month.
>  I suggested 1 per reviewer per month on the basis that:
>  - there are ~ 20 active review/devs
>  - we land 200 branches a month (on devel)
>  - we'd need a 10% unreviewed rate to even reach one post-landing
>    review per month per reviewer.
>  So I suggest staying with 1 per month and if the unreviewed rate is over 30%:
>  reviewing 1 branch per reviewer per month would review 1/3 of the unreviewed
>  branches - more than enough to form an assessment

Staying with 1/reviewer per month.

> I do like the idea of catching problems early, but let's do that
> explicitly: I will review *all* unreviewed landings (other than ones I
> do) in the two weeks, and if I land anything unreviewied, someone else
> - say jml - can review those.

I will be doing this.

>  - javascript reviews
>  - sure, let's leave them out of scope

Javascript is staying out of scope.

I will update the wiki tomorrow (today is perf day!) - but the
experiment is now active as far as Francis and I are concerned.

-Rob