Hey Jakob, and all, On 3/19/13 10:32 AM, "Jakob Homan" <[email protected]> wrote:
>+1 to standard RTC with reviews before all commits excepting those >necessary to fix a broken build. > >On Tue, Mar 19, 2013 at 10:13 AM, Henry Saputra ><[email protected]>wrote: > >> I like the RTC but I believe we should not be too strict about it in the >> early phase of incubating project. >> >It's not about being strict, it's about establishing a culture and >community norms. Whether it's strict or not-strict isn't uber important. At Apache, those doing the work decide. If those doing the work really want to take advantage of patch reviewing, before committing stuff to version control, by all means. I'm a big fan of Review Board, but I never like projects that *require* anything, least of all patch review. These are really community and social norms we're talking about here, not technical. If there is a desire by a minority to perhaps commit and not hold stuff in patches, and that minority has a ton of great work and thoughts, and discusses them on list, I would encourage Tajo to tell that minority (PPMC member, let's say): 1. Create a branch 2. Go wild 3. Merge into *pristine* area selectively, with consensus, (LOL I'm not a Git expert, but in my SVN mind, let's say "trunk") and that merge may need to be reviewed by patch review, if most of the Tajo peeps like RTC and are working in that pristine area. We're using version control here, and branches are cheap, so I would encourage the above flow. That being said, "play nice" is the advice I'd give :) >[..snip..] >> >> I would say we use RTC with common sense. If you are in doubt fire JIRA >>+ >> review. Especially when half of the team are split between day and >>night =) > >Geographically dispersed teams is a better reason to use RTC as it gives >those still asleep a chance to look at the code before it goes in, >avoiding >the cost of reverting. I don't think definitively geographically distributed teams means RTC or CTR are better. Or at least I'll say I haven't read any software engineering research on either. > >Henry, it looks like there's pretty strong consensus to standard RTC with >6x+1 for it and Chris' -0 (is that a good characterization, Chris). Heh, thanks for trying to characterize it Jakob. Yeah I guess I'm ambivalent. I think people should be able to operate in both modes per my 1-3 above. I tend to do that in OODT, Nutch, Tika, Solr/Lucene (when I was on those projects), SIS, Gora, etc. That is, sometimes I really desire something that touches "controversial" or "load bearing wall" code to be reviewed by the other peeps to make sure I didn't screw stuff up. In that case, I'll create an RB, and request RTC. In other cases, when I'm working on parts of the code that are new, or that no one else in the community cares about, I've been known to go into CTR mode. I've seen others do this as well. YMMV :) Cheers, Chris
