Connection Minimization in REST API with Random Walks

Abstract

A key constraint of REST API is that all the resources must be reachable by some hyperlink paths from an entry point. However, to apply this constraint without prudence can result in excessive hyperlinks that do not provide new services but increase the dependence between the resources. Excessive hyperlinks are difficult to identify because: 1) a REST API can have dynamic and unbounded paths, and 2) the hyperlinks used to navigate a path are not observable and can be ambiguous. To tackle the first challenge, we propose a REST API model and a random walk algorithm to reduce the paths of a REST API to a small set. To address the second challenge, we develop a client model and a connection minimization algorithm to identify excessive hyperlinks based on given paths. By combining the random walk and the connection minimization algorithms, our method can minimize the connections of a REST API in polynomial time without involving the actual clients. A prototype system has been implemented and the tests show that the method is correct and can converge 90.6% to 99.9% faster than the baseline approach.