REST’s limit
Typical REST APIs have two problems for mobile/SPA clients:
- Over-fetching — endpoint returns more data than needed
- Under-fetching — an operation requires calls to N endpoints
Facebook internally (2012) develops GraphQL to feed the iOS/Android mobile app with heterogeneous data (profile, feed, notifications) in a single optimised call.
The release
GraphQL spec is released as open source by Facebook on 14 July 2015. Leads: Lee Byron, Dan Schafer, Nick Schrock. The first stable version arrives in 2016. MIT licence; spec governed by GraphQL Foundation (Linux Foundation) from 2018.
Concepts
- Schema — strongly typed type system with scalar, object, interfaces, unions, enums
- Query — read operation, client specifies data shape
- Mutation — write operation
- Subscription — real-time stream (typically over WebSocket)
- Resolver — server functions that resolve each field
type User {
id: ID!
name: String!
posts: [Post!]!
}
query {
user(id: "123") {
name
posts { title }
}
}Advantages
- Single endpoint
/graphqlfor everything - Client-specified data — no over/under-fetching
- Typed schema — introspection, tooling
- Versionless — evolution via deprecation
- Strong tooling — GraphiQL, Apollo Studio, Insomnia, Postman
Disadvantages
- N+1 queries — naive resolvers devastating; requires DataLoader
- Caching — more complex than REST/HTTP cache
- Complexity attacks — deep queries expensive; needs depth limit, cost analysis
- Over-engineering for simple APIs
Implementations
- graphql-js — reference, Node.js
- Apollo Server/Client — most widespread ecosystem
- Relay — Meta client for React
- Hasura — GraphQL auto-generated from Postgres
- PostGraphile — Postgres → GraphQL
- Strawberry (Python), graphql-ruby, sangria (Scala), graphql-java, HotChocolate (.NET)
Federation and Gateway
- Apollo Federation (2019, v1) / (2021, v2) — sub-graph composition into super-graph
- Schema stitching — earlier approach
- GraphQL Mesh — multi-source gateway
Relative decline 2022-2026
After the 2016-2020 enthusiasm, many companies return to REST, tRPC, OpenAPI-first. Reasons:
- Apollo becoming more commercial
- Operational complexity
- Harder HTTP caching
- Netflix, Twitter deprecate internal use
- tRPC offers end-to-end type safety for fullstack TypeScript stacks
GraphQL remains dominant where multi-source aggregation is needed (BFF, API composition, public dev portals).
In the Italian context
GraphQL is used in:
- Magento/Shopify/Commercetools (headless) e-commerce
- Media publishing (Mediaset, GEDI)
- Italian mobile apps with GraphQL BFF
- Internal admin tools for data composition
References: GraphQL open source (14 July 2015). Facebook/Meta. Lee Byron, Dan Schafer, Nick Schrock. GraphQL Foundation (2018). Apollo Server/Client ecosystem. Apollo Federation v2 (2021).