Yet files are bigger now and more numerous than ever. We definitely want rich indexes of files sometimes but files can be so much more than rich data for users

Yet files are bigger now and more numerous than ever. We definitely want rich indexes of files sometimes but files can be so much more than rich data for users — they are also cheap book-keeping for programs for users. Imagine trying to implement a performant database on top of this database. No amount of technology can make O(log N) a substitute for O(1).