This is intended to improve support for Option #4 in this blog post.
The idea here is to have two parallel data models, and the ability to automatically convert between one and the other allowing archives to be conveniently constructed out of Müsli data models.