This is a direct outflow from chaos theory, it seems.

[headers] lists the C headers in a table, and then [depr.c.headers] lists those exact same headers again.
[Removed my confusion about deprecated C headers vs. deprecated C++ headers.]

@jensmaurer: I'm not sure that's true. [headers] lists <cstddef>, which is not deprecated. [depr.c.headers] lists <stddef.h>, which is deprecated, as stated by [diffs.mods.to.headers].

Moreover we are apparently also deprecating <ccomplex> etc; I have an issue open about that (a pull request, in fact).

What's perhaps confusing is the term "C header", but only mildly so.

@tkoeppe: Ah, you're right. Those lists subtly differ in their .h suffix. :-/

Why does D.4p2 talk about C++ headers, then? [depr.c.headers] is supposed to be about .h -style headers.

Why does C.5.1p4 talk about C++ headers in a section entitled "C standard library", showing incompatibilities vs. C11? Surely, C11 doesn't have those C++ headers.

I think D.4p2 is just because Clark couldn't find a better place for that in P0063.

C.5.1p4 is a note, but I was told that all of Annex C is a note, so it's not marked up. (I recently reworked C.5.1 a fair bit.)

I filed an LWG issue about all this a while ago. I think it would make sense in general to say that "X is deprecated" in the main text, and then describe X in Annex D, but we need to think how that affects things like <cstdalign>. Alternatively, we can provide deprecated facilities entirely in Annex D, such as auto_ptr.

See also #1054.