As suggested by @albahnsen I am introducing this issue to discuss the topic of the providing of a compatible format across datasets. This won't be implemented now but could be relevant in the near future.
From my point of view providing the datasets in a compatible format can be done in two ways:
Storing them locally which can raise storage space and license issues. We would have to write code to process the data.
Dispatching some import libraries, which consumes no storage, requires also to write importers and may induce a coupling with the importers language. Though this coupling could be reduce by writing small connecters for different languages.
As suggested by @albahnsen I am introducing this issue to discuss the topic of the providing of a compatible format across datasets. This won't be implemented now but could be relevant in the near future.
From my point of view providing the datasets in a compatible format can be done in two ways: