> It seems David places more value on the idea of the concrete mapping "pointing forwards" with respect to the abstract directed graph, while it seems Tim places more value on the idea of the abstract mapping direction corresponding to the final static order. These two goals are in conflict, assuming we don't want to change the behavior.

Yes, that's right. But the good news is that if you're willing to rewrite all of the documentation you probably can explain this in a clear and simple way, and without breaking compatibility.

I say *you* because I'm not going to argue with you all about this anymore, especially with Tim being...

---

Tim: you're conflating the words "predecessors" and "dependency".
In some contexts they can be synonymous, but they are not the same thing.
 * Predecessor refers to one of the sides of a directed edge.
 * Dependency refers to a semantic relationship between two of the users' things.



> The only possible topsort [...] For which see absolutely any text defining the terms.

From wiki: "Precisely, a topological sort is a graph traversal in which each node v is visited only after all its dependencies are visited."

This definition doesn't say anything about the "predecessors" or how the graph is stored,
or anything about "edge direction". I like this definition.



> that's not a matter of preference, it's just plain wrong

I know that there are many different ways to represent a graph, but your graph format *is just plain wrong.*

Goodbye