[lozenge]

You wouldn't want to let people filter by an unindexed field, sort huge large result sets, or make your server store a huge result set because they stop making requests after getting the first few pages. GraphQL lets you express what is permitted, without allowing more than your underlying data store can support.

For example, from the GitHub GraphQL API you can in a single request get all open pull requests on a repository and the last 10 comments on each one. You can't get the last 10 comments on all of GitHub or the last 10 comments made by a particular user, because they haven't exposed or even stored the data that way.

In a REST API that would be dozens of separate requests, or the endpoint would just not exist because they have never had the need for that exact combination of records to be queryable.