I tend to think of branches in git the same way I think about strings in C or lists in (Common) Lisp. In each of these cases, there's a simpler underlying data

I tend to think of branches in git the same way I think about strings in C or lists in (Common) Lisp. In each of these cases, there's a simpler underlying data structure that's directly overlaid with a set of conventions. It's the conventions on the use of the simpler structure that gives the illusion of some higher order data type. (The distinction I'm making between these and abstraction in general is that here the abstraction is almost intentionally somewhat leaky - presenting both strengths and weaknesses.)