I think [stream.types] is trying to say "standard signed integer types", and the footnote should say "standard integer types".

[support.types.layout] p2 and p3 imply that ptrdiff_t and size_t can be extended integer types.
Changing the footnote as requested would make it inconsistent with the normative rule.

@jwakely, I think it's not editorially-obvious whether [stream.types] wants to allow extended integer types or not. It is at least conceivable to have a platform where the largest standard integer type is 32 bits, yet an extended integer type is used to represent 64-bit file sizes.

(I'm reading the existing "basic" as possibly meaning "built-in".)

Could you please make this an LWG issue that makes all of [support.types.layout], [stream.types], and [derivation] consistent?

I think the footnote can be fixed editorially by simple dropping "for basic integral types", since it applies to any typedef.

This is now LWG3696.