Engineering families of protocols using a formal language: successes and failures

So called formal methods are notoriously weak on the "method" side. Refinement is arguably one of the best established method supported by mathematical notations and tools. We explore the use of refinement to specify retransmission mechanisms of communication protocols. Starting from a very abstract channel, we present a partial hierarchy of protocols ending with a fairly complex variation on sliding windows. While we were able to specify such a hierarchy, we have found that the benefits of using refinement to specify such protocols in a practical setting are limited.