On Thu, 2008-08-14 at 15:50 +0100, Aidan Skinner wrote: > On Thu, Aug 14, 2008 at 3:21 PM, Andrew Stitcher <[EMAIL PROTECTED]> wrote: > > > On Thu, 2008-08-14 at 15:16 +0100, Aidan Skinner wrote: > >> I've raised an INFRA ticket to get a 'Ready for review' status in > >> between 'In Progress' and 'Resolved'. > >> > >> I'd quite like to move the Java client and broker to a > >> review-then-commit model, with any one committer other than the author > >> being sufficent. > > > > 1. I assume this is intended for changes to released code. > > I was kinda thinking for all the Java stack actually. We don't do much > that is entirely new development.
I do agree that code is always better when it's reviewed, but I think anything like this would have to be enforced (at least somewhat) for this to have value. > > > 2. Without actually commiting the change the reviewer doesn't easily get > > to see the change, and no I don't consider attached patches to be > > adequate with svn. > > Why not? Is it just the rebasing problem? Applying patches, and tracking them is just not well supported by svn. It works well for git, but we can hardly require a reviewer to run a different vcs just to do their task. For example I've been reviewing the Windows and Solaris patches, but doing it from the bare patch isn't adequate as you can't be sure it will apply etc. Also attached patches start to rot after a while and without info as to what version they change they can be hard to apply later. > > Personally I'd prefer small branches that can be reviewed then merged, > but I don't think that's realistic until and unless we move to > $DVCS[1]. > > - Aidan > > [1] For $DVCS == git, as it is clearly superior to bzr and hg in crucial ways Care to elaborate? I've only used git, so have no real idea of the differences. Andrew
