A very long time ago, I wrote an XML library (Gnosis Utilities xml_objectify) that had this same issue, and adopted the "duality" approach (where possible, with both dictionary and other styles also available). JSON is sort of the new XML, and it feels like the same concern. FWIW, LXML explicitly adopted my approach for an optional (mostly) compatible mode, and unlike me, they maintained that thing during the last 15 years since mine was touched.
I think the way to think about parsing JSON needn't be "this is just a Python dictionary once we read it", but rather that we have a custom dict-like object that borrows JSON/Javascript semantics. I've mentioned that I think a new name emphasizes this. And living in `collections` rather than `types` also fits this purpose better.
However we get there, writing `obj.element.sub.another` is SO MUCH nicer.
--
Keeping medicines from the bloodstreams of the sick; food
from the bellies of the hungry; books from the hands of the
uneducated; technology from the underdeveloped; and putting
advocates of freedom in prisons. Intellectual property is
to the 21st century what the slave trade was to the 16th.