Sure, if you want to do it once, but Git still has to compute that information (save for a new-ish cache that is just that, a cache). But that is not the point really, the point is that Mercurial's graph Is the same (topologically) everywhere, which is not the case in Git because branches (and thus remotes) have different names. So saying that a branch contains a commit is not the same as a commit being on a branch. There are a bunch of great properties that emerge from this but it's too long for this comment and I should actually properly write this down at some point this year.