The SCR approach to interface specifications is based on a design approach pioneered by Parnas and others [Britton+ 81]. More information about the other interface examples documented in IDL, HLA, and XML is contained in [OMGIDL 02], [IEEEP1 516 00], and [W3GXML 02].
An excellent foundation paper on exceptions, which lays the groundwork for separating the concern of detecting an exception from the concern of handling an exception, is [ParnasWuerges 76].
Mary Shaw has made the observation that we can't have complete interface specifications, because the cast of stakeholders is too numerous and the range of information they need is too broad. And in the world in which we get our components from other sources and know precious little about them, good interface specifications are even more rare. However, she points out that we can and do accomplish useful work with such incomplete knowledge. This is so because we can assign confidence measures to individual units of information that we pick up about a component from various sources. She calls such a unit a "credential," and assigns it properties such as how we know it and what confidence we have in it [Shaw 96].
Interfaces seen as the set of assumptions that two components are allowed to make about each other dates from early work by Parnas [Parnas 71], echoed in more recent work about architectural mismatch [Garlan+ 95].