incomplete types for objects with linkage--prior art?

[James Jones]

A posting a while back indicated that the following is strictly-conforming:

	struct woof		arf;
	struct woof {
		int		i, j, k;
	};

It conforms because the semantics in 3.5 only refer to identifiers with no
linkage when constraining the type to be complete by the end of the declarator.

I'm curious--is there prior art for this interpretation?

	James Jones