Existing Wording: Best practice is to explicitly include all known properties on contained resources, even if redundant with properties on the containing resource. E.g. If the containing resource and contained resource both have a "subject" element, both elements should be present, even if they reference the same resouce. Applications that choose to omit redundant elements on contained resources should not expect other applications to infer (or correctly infer) the intended meaning in this case. Applications cannot safely presume that omitted elements in contained resources can safely be inferred to be the same as the containing resource.

Comment:

I commented on this in the last ballot. I still don't understand why this can't be made less "iffy." A contained resource cannot have an independent existence outside of the resource that contains it, according to 1.13.0.1: so why can't the subject, if blank, be always safely inferred to be the subject of the containing resource? What other subject could it possibly be, if this instance is its only existence? Yes its subject can be overridden, but I would rather that absence of subject always have the same predictable meaning of "same subject as container." Leaving it as a "best practice" that cannot be guaranteed leads to ambiguity which I think is unnecessary. I understand why an external reference cannot make the same assumption about a blank subject, but think it is safer and simpler to say that a contained resource has the same subject as its container, except when overridden.