A Design Pattern for Reliable HTTP-Based Applications

HTTP is currently being used as the communication protocol for many applications on the web, supporting business and safety-critical services throughout the world. Despite the growing importance, HTTP-based applications are quite exposed to network failures, which can bring in huge losses to the service users and providers, including financial and reputation losses. Several approaches try to achieve reliable communication by using logging and retransmission of whole HTTP messages, which is especially ill-adapted to large messages. Stream-based approaches are more efficient as, upon failure, they transparently resume data transmission from where it stopped. Despite this, designing a stream-based mechanism involves significant challenges, as it is very difficult to know how much data is lost due to failure. In this paper we propose a stream-based mechanism for reliable HTTP communication that is entirely based on HTTP messages and is compatible with existing software. The mechanism is presented as a design pattern and relieves developers from explicitly handling connection failures, providing a standard way for building more reliable applications. Results show that the mechanism is functional, compatible with legacy applications, and that the coding and runtime costs of this design pattern are quite low.