SVN-Git migration plan
jamonation at gmail.com
Mon Jan 31 15:09:52 UTC 2011
On 01/30/2011 04:20 PM, Antranig Basman wrote:
> Thanks for all this careful analysis and presentation of alternatives, Jamon. To me, option 2 sounds fine. An important part of our responsibility in keeping a source history is keeping track of IP - however, empty revisions could could carry no IP. I think it is important to get our work into git as soon as is practicable and option 2 seems the fastest approach which will yield an acceptable result.
> One issue might be that we have some implementation which depended on the existence of a particular empty directory, and that a git checkout of this history would produce a non-working image. I believe there has only ever been one instance of this issue, in the mid-period of Engage... but as I understand it, none of the other options would help with this in any case, since empty directories cannot be represented in git.
> That's my opinion - Colin, what are your thoughts?
Creating an empty .gitignore in an otherwise empty directory (and
committing) is a workaround/hack for that particular issue. Same for
.hgignore files in Mercurial.
Sounds like option 2 is the most reasonable - Justin and I will continue
working on the import using that approach.
More information about the fluid-work