In many cases, groups of computer scientists do the right thing when
designing a new system. The soundness of the protocols have been
proved. The efficiency and reliability of the algorithms have been
analyzed. The APIs have been designed and widely reviewed. This was
the case with Network Admission Control (NAC). Yet the adoption of NAC
was not as smooth as one might have hoped.
In this talk, I will discuss NAC, describe some of the adoption hiccups
in the last four years, and show how a greater understanding of the
customers' deployment requirements might have avoided these problems.