Daniel James wrote:
On 1 March 2012 13:06, Dave Abrahams <dave@boostpro.com> wrote:
on Thu Mar 01 2012, Daniel James <dnljms-AT-gmail.com> wrote:
OK then, a modularised boost is an expensive precondition for moving to git.
It isn't a precondition for moving to Git. We can move to Git and then do the modularization step; it's not a problem. However, it does mean two transitions.
Which can be a good thing. Breaking a process down into smaller stages can make it easier. It seems to me that we've discussed git several times, and it's always part of a grand scheme.
If by "part of a grand scheme" you mean CMake, modularization or the Ryppl model in its entirety, I agree; but if you mean gitflow branching or the crude test image model that emerged from our discussion with Thomas Heller, I don't. I think both of the latter two ideas would involve only small, gradual changes which are best implemented (shortly) after the conversion to git is made. These are just a matter of "adapting to new opportunities". In fact, it seems that https://github.com/ryppl/boost-svn almost conforms to the gitflow branching model already.
If git is considered desirable enough, then it might be best to just switch to it, keeping everything else the same.