I think in most cases you want the dependencies between commits that darcs removes. For instance, let’s say you have three patches:

1. Create hello.sh with echo Hello World
2. Create lib.sh with hello() {echo Hello World}
3. Edit hello.sh to source lib.sh; hello

Of course, in this case it would be easy to manually specify the dependency (3 depends on 2) and unlike git 2 wouldn’t have to depend on 1. But for a larger project I’m imagining always having to look at the current path set to figure out dependencies and the git model of “every commit depends on the patch before” seems better for large codebases.

But I could be wrong. I’d be curious to see what a large darcs repo looks like.