Suppose then that we have a requirement r = “supply customer with goods upon request” and let s be a system operationalizing r. The “desired way” of the above quote for s is that it always fulfills r, i.e., every time there is a customer request the system meets it successfully. This means that the system somehow manages to deliver its functionality under all circumstances (e.g., even when one of the requested items is not available). Such a requirement can be expressed, roughly, as r1 = “Every instance of requirement r succeeds”. And, of course, an obvious way to operationalize r1 is to add to the architecture of s a feedback loop that monitors if system responses to requests are being met, and takes corrective action if they are not. We can generalize on this: we could require that s succeeds more than 95% of the time over any one-month period, or that the average time it takes to supply a customer over any one week period is no more than 2 days. The common thread in all these examples is that they define requirements about the run-time success/failure/quality-of-service of other requirements. We call these self-awareness requirements.ล