It's not a problem for attributes to be unqualified, as long as they are only used on qualified elements, which the atom elements are.
The namespace qualification or lack thereof on attributes has nil to do with parent element qualification and everything to do with XML-reserved attribute names and, when those don't apply, the current default namespace. Should I interpret the above-quoted text as suggesting otherwise?
No. Thinking about it, unqualified elements would work just as well. I was thinking about describing unqualified attributes in a schema, but that would also work with unqualified elements.