Context : AcademicJournals
Got into a discussion wrt this Quora question.
An automated system that recursively downloaded cited papers would be awesome, of course.
In fact, if academic publishing worked properly (providing a useful service), that's what it would provide : "the tree of papers that led to this one", datasets from the tree of papers that led to this one. Etc. etc.
Even if it wasn't open. If the academic publishers really wanted to "add value" to academia, why not do it like this? Eg. don't just sell me individual papers expensively, or journals even more expensively. Get together with other journals and figure out a way to sell me "the papers relevant to my research" in a single bundle.