Figure 7

The Penumbra’s intersection with other datasets. Of the 55,343 discovered, non-empty repositories, 18% have a first commit hash that can also be found on GitHub (GitHub Duplicate), but 30% of those repositories diverge and contain code not found on GitHub (GitHub Diverged). Likewise, 7% of Penumbra repositories have a first commit archived by Software Heritage [1], and 14% of those contain code not archived by Software Heritage. Finally, 21% of Penumbra repositories share a commit with one or more other Penumbra repositories, and of these, 58% have unique final commits