Looked over bugs filed about v6 mode and did some triage and analysis. smudge has the details.

This led to changing what's done by git add and git commit file when annex.largefiles is not configured. Rather than behaving like git annex add and always storing the file in the annex, it will store it in the annex if the old version was annexed, and in git if the old version was stored in git. This avoids accidental conversions.

It might make sense to have git annex add also do this, even in v5 repositories, but I want to concentrate on v6 for now, and also don't think that git add and git annex add necessarily need to behave identically in v6 mode. While using git commit -a doesn't imply anything about whether you want the file in git or the annex, using git-annex add seems to imply they you want it in the annex, unless you've gone out of your way to configure otherwise.

Also did some design work on supporting versioned S3 buckets with git-annex export.

This work is supported by the NSF-funded DataLad project.