Interface evolution patterns

Abstract

Remote Application Programming Interfaces (APIs) are technology enablers for distributed systems and cloud-native application development. API providers find it hard to design their remote APIs so that they can be evolved easily; refactoring and extending an API while preserving backward compatibility is particularly challenging. If APIs are evolved poorly, clients are critically impacted; high costs to adapt and compensate for downtimes may result. For instance, if an API provider publishes a new incompatible API version, existing clients might break and not function properly until they are upgraded to support the new version. Hence, applying adequate strategies for evolving service APIs is one of the core problems in API governance, which in turn is a prerequisite for successfully integrating service providers with their clients in the long run. Although many patterns and pattern languages are concerned with API, service design, and related integration technologies, patterns guiding the evolution of APIs are missing to date. Extending our emerging pattern language on Microservice API Patterns (MAP), we introduce a set of patterns focusing on API evolution strategies in this paper: API Description, Version Identifier, Semantic Versioning, Eternal Lifetime Guarantee, Limited Lifetime Guarantee, Two in Production, Aggressive Obsolescence, and Experimental Preview. The patterns were mined from public Web APIs and industry projects the authors had been involved in.