I haven't looked at working with GraphQL too closely yet, but this seems to partly answer some questions I wondered about. Because REST leverages HTTP, our infrastructure--our client software, our frameworks and programming languages--already have some knowledge of what to do with it, how to parse URLs and GET and POST requests, and so on. But the discussions I've seen of GraphQL suggest that in order to make it work, additional layers of code must be added, such as Relay and Apollo, or some equivalent. Is this really the case? Perhaps this will change in the future--as GraphQL becomes more widespread, it may be integrated more natively into the infrastructure.