I've noticed listed problems, they surely do exist. but stepping back a little - proposed solution sounds like: "lets make a brand new version control system for already existing version control system".
Not really. There are many mature tools outside of the GitHub monoculture that do changesets, both on top of Git as well as other VCSs.
I don’t do much reviews these days but I remember the confusion of GitHub PRs. Things just disappear, especially with the rebase workflow which is preferable for improving reviewer burden.
I think Mitchell is mostly right here: a changeset is the most natural data structure for maintaining code. However, is it also the most natural construct for storing code in repositories, ie “should git be replaced in the long term?”. Can the merkle tree of blobs be replaced by an analogous merkle tree of changesets? That I don’t know. But it’s a worthwhile idea.
I don’t do much reviews these days but I remember the confusion of GitHub PRs. Things just disappear, especially with the rebase workflow which is preferable for improving reviewer burden.
I think Mitchell is mostly right here: a changeset is the most natural data structure for maintaining code. However, is it also the most natural construct for storing code in repositories, ie “should git be replaced in the long term?”. Can the merkle tree of blobs be replaced by an analogous merkle tree of changesets? That I don’t know. But it’s a worthwhile idea.