I agree with a lot of what was just said. My only possible issue with the github workflow: I'm not sure how it interacts with having multiple people who have control of the "master" (central) repo. When a pull request comes in, can anyone who has push access to the repo take control of that pull request?
Also, in my vastly finite experience with github, I've had problems with the fact that pull requests seem to be immutable: once a request has been created, it doesn't seem to be easy to add more commits to it, or change the commits. (This may just be my naïvety.) But other than, I think github does have a great model. Is github on the table for a future platform to move to? I think gerrit is definitely a heavy-handed approach to reviews that seems to work better with larger products (like Android), since you do sort of have to adapt to its style of reviews. And I don't know if it is the right tool for Sage (not sure if anything is), but it is certainly worth looking at. --Christopher On Tue, Feb 21, 2012 at 13:52, Fernando Perez <fperez....@gmail.com> wrote: > On Tue, Feb 21, 2012 at 7:38 AM, Jason Grout > <jason-s...@creativetrax.com> wrote: > > I'll have to think more about whether I agree with the "each commit > should > > stand on its own" philosophy mentioned below. It certainly makes > bisection > > easier. But sometimes huge patches make me think that each *branch* > should > > stand on its own, and be merged with no fast-forward. Then it's much > easier > > to tease apart logically separate, but interrelated, sets of changes. > > Many thanks for this super useful discussion! From my own experience > with github's branch-based review model, I lean strongly towards the > view you summarize above. I prefer to let authors make very liberal > use of the commit abilities in git and review with the branch as the > 'atomic unit' I worry about, giving only secondary consideration to > individual commits. > > Obviously if the commit history is a complete mess we may require that > it gets rebased and cleaned up, but in general we tend to be > relatively lax with how the individual commits stack up, as long as > the entire branch diff is sensible and easy to analyze on its own. > > From our own experience in ipython with github and how unbelievably > fluid this has made the process both for new contributors and for > reviewers, I think I really prefer this over the view that gerrit > seems to propose. But it's been very enlightening to read their > rationale, and I certainly see it as a coherent perspective one can > choose to adopt. > > Cheers, > > f > > -- > To post to this group, send an email to sage-devel@googlegroups.com > To unsubscribe from this group, send an email to > sage-devel+unsubscr...@googlegroups.com > For more options, visit this group at > http://groups.google.com/group/sage-devel > URL: http://www.sagemath.org > -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
