You sound a bit like those people, “git is just directed acyclic graph of objects, it’s all that simple”.

“A monad is a monoid in the category of endofunctors, what's the problem?”

http://james-iry.blogspot.com/2009/05/brief-incomplete-and-m...

The difference is that that formulation is exactly correct whereas the other ones are not.

Is git necessarily acyclic? Or is it just that nobody has bothered to contrive a git repo with cycles?

It seems possible in principle: via a kind of quining, a git commit can be created whose parent hash is its own.

It involves a bit of a tough search.

It’s very likely specified as an acyclic graph (because, without checking, I see no reason why it would not be). So if you made a cyclic graph it would be wrong. It’s possible, likely even, that a lot of git’s implementation and associated tools don’t check for that, because it’s very hard[1] to achieve, but that just means stuff would break in more non-obvious ways.

[1] Constructing a cyclic git graph means finding a hash collision, and git uses SHA1 which is known to have weaknesses. With another hash algorithm it might be “practically impossible” instead of “very hard”.

You can use the "grafting" feature to emulate rewriting commit parents (but not true rewrite--as in the object itself is not actually modified and some tools do not respect it), so maybe give it a shot

I know it's oversimplified, but as a poor man's explanation I couldn't really think of a better way to articulate it at the moment