Does anyone have anything further to add here? Looking forward to a few more opinions.
On Fri, Dec 11, 2015 at 3:52 PM, Lars Francke <lars.fran...@gmail.com> wrote: > Hi, > > thanks Larry. > > Lars provided the Knox community with some feedback into our development >> practices and JIRA usage [1]. >> >> I wanted to bring up a DISCUSS thread on how our CTR policy may or may >> not relate to a couple points made in his feedback. In particular: >> >> 1. Whether a CTR based policy should require actual patches attached to >> every JIRA or does the git link to a commit provide the same ability to >> review post commit >> > > I think having a patch attached before commit is a good thing because: > * It allows feedback before a change goes in. In my experience that is > easier to change than committed code > * It allows looking at the exact change set in a standardised form (diff > format) without having to use whatever web frontend (pure Git, Github, ...) > is currently being used (so a patch file is useful even when only attached > after committing)[1] > * Should the Git web interface change (or be down) at some point in the > future all those links might go stale (say Apache switches to Github or > Gitlab or whatever) > > The downsides I can come up with is the extra work required to attach the > patch (before or after commit). > > >> 2. Whether a cool off period of 24 hrs would encourage more external >> contributions and if so, how >> > > This one is probably a matter of weighing off between fast iteration and > possible feedback on patches. I assume with a project the size of Knox > there is not much feedback coming in for each patch so the benefits of > immediate commits probably outweigh the (assumed) benefits of waiting. > > My personal opinion though is that a wait period is a good thing. I've > been bitten (and frustrated) in the past by reviews that have been ignored > by certain communities/members/companies where it was clear that a release > schedule had to be met. Ignoring reviews is easier with code that's already > been committed. Again this is my personal (bad) experience and I have no > reason to believe that the Knox community behaves the same. A cool off > period doesn't mean that reviews are mandatory, it just invites/allows > feedback in my opinion. > > One "real" benefit for pre-commit reviews is that there's tools available > and in use at Apache for that (Reviewboard and JIRA to a degree) but > there's currently no tooling support for post-commit reviews. > > Caveat to all of this: I'm only here for a few days a year probably and > won't contribute much if at all so you should decide carefully whether you > want to change working practices for an unknown benefit. > > Cheers, > Lars > > [1] I know that the current web interface has an option to download the > patch but it's a different process than most other "Hadoop related" > projects. >
