scm discussion in fedora
June 8, 2007
I’ve been following the scm discussion with some interest as I’m trying to determine which scm to switch to for yum’s development trees, too. Yum is currently using cvs b/c when we started that was pretty much all their was available. We’ve been slow to change b/c cvs has just worked. More importantly to me is that I’ve been hoping a ‘winner’ or even a few winners would shake out of the scm explosion that’s occurred in the last few years. I was hoping, apparently in vain, that I would not need to learn 10 vaguely different commands to do the same thing. As much fun as scms are I’d rather just get my work done and not learn a new doo-dad to do it. So a lot of the reticence you hear from various folks about changing the workflow is about that.
But there’s more. In fedora we just finished the merge of extras and core. Since you work for red hat it may not have seemed as involved a process as it did from the outside. Additionally, it is still being shaken out. If we were to shift to a new scm right now, as we’re sorting out issues with the merge and the update systems I think we would end up alienating and confusing an enormous number of developers and maintainers., myself included. Therefore I think some deliberation and consideration is warranted for a move of this kind. The last thing I’d want to see happen is us to shift to $SCM only to have development stop on it and have to do it all over again. Now, that may not happen with git or mercurial or svn or whatever; however I’d be much happier taking our time to verify those things and to bring the developer base comfortably along with us. There’s some lovely docs written up for managing volunteer labor in non-profits. They specifically mention that making the processes confusing will absolutely guarantee that no one will contribute, despite their desire.