On Fri, Jun 18, 2010 at 12:17:57PM +0200, David Lutterkort wrote: > Hi all, > > I am looking for some guidance/experience on how to best do the import > of the existing Deltacloud code[1] into the incubator svn repository. > > A would very much like to preserve the history of the existing repo, > though I am not sure if that is possible with git-svn or similar, > especially since we unfortunately have some merge commits on the master > branch (which I could probably get rid of with a bit of massaging) > > Besides the possibly mythical ideal solution of importing the current > git master branch into svn, I see two other options: > > 1. Kill history > =============== > > Throw away the history and just commit the current HEAD from the git > repo. I _really_ want to avoid that; there's important info about the > genesis of the code base and subtle bug fixes in the history that really > needs to be preserved. > > I've heard rumors that Apache infrastructure will offer git hosting > soonish - if that is the case we'd have discarded important information > for now really good reason. > > 2. Keep it in git > ================= > > Keep the code in some git repo for the time being and wait/hope for > apache infrastructure to offer git as a SCM. We'd need to advertise that > deviation from Apache customs prominently on the project page. > > OTOH, it would be the least labor-intensive approach, and would people > allow to work with git without the git-svn crutch. > > The biggest downside is that we'd have to manually manage the > committers, all of which would need to get another account at whatever > git hoster we decide to use. Currently, the code is on fedorahosted.org, > which requires that committers have an account in the Fedora Account > System. But I would be just as happy using a different service like > github or gitorious; in both it is possible to set up multiple > committers for one repo. > > what does everybody think ? >
>From my standpoint, if there is any way at all we can avoid getting involved with Subversion, I would take that way. Having said that I'm pretty sure Jim Meyering (cc'd) has done successful git2svn gateways in the past, he may have some advice. --Hugh
