I like the way Mezzo deals with memory using "ideas from affine type systems, separation logic, and systems based on regions and capabilities," but my enthusiasm is capped for two reasons.
One of the reasons is the same as one of David Barbour's criticisms of typestate.[1] Why should I care about Mezzo's state handling if it only permits reasoning about state metaphors within the program, and not external resources that are inherently stateful?
Two, this blog post focuses on how Mezzo can express programs without particular classes of errors, rather than with particular conveniences or desirable properties. Eliminating failure doesn't imply that success is easier.
