Agreed, I'd rather have a few simple constructs, especially if they compose. Still, there's value in unification and generality, at least when the result is si

Agreed, I'd rather have a few simple constructs, especially if they compose. Still, there's value in unification and generality, at least when the result is simple, even if that covers only 80% of cases. For example you generally run a lot more composable tasks than actual threads and syncs. You get a general, and powerful abstraction for the general case yet threads are still there for more exotic needs.