No abstract

SummaryColin L. Mallows has had a distinguished career as a research statistician in the telecommunications industry, primarily at Bell Labs, that has spanned more than 50 years.Best known for creating ‘ Cp', the regression model diagnostic procedure that has been widely used and probably benefitted all statisticians who have done regressions, Mallows also has, and continues, to produce influential research on a wide variety of statistical and mathematical topics. He has received the Wilks Memorial Award, the American Statistical Associations' highest award and presented both the Fisher Lecture and the Deming Lecture at the Joint Statistical Meetings. Mallows is one of only two statisticians (the other being George Box) to have received all three of these honours that are among the highest the profession bestows.Born in England in 1930, Mallows received his PhD at the age of 22 from University College London under F.N. David and N.L. Johnson. In 1956, John Tukey invited him to spend a year at Princeton, and in 1960, Mallows joined Bell Labs, the R&D unit of AT&T, which at the time was the regulated monopoly telecommunications provider in the USA (Ma Bell). Mallows served as Department Head of one of the two Bell Labs Statistics Research Departments for many years and remained at Bell Labs until 1995, when he joined the AT&T Labs unit that was created as part of the restructuring of the telecommunications industry. Since his retirement from AT&T Labs in 2000, Mallows continues to consult part‐time in Avaya Labs, which is a sister company in telecommunications that also traces its roots back to AT&T and Bell Labs.This conversation was held in November, 2011, with two long‐time colleagues and collaborators from both Bell Labs and now Avaya Labs. After recording and transcription, the conversation was edited by the participants.

Static program analysis uses many checkers to discover a very large number of programming issues, but with a high false alarm rate. With the aid of dynamic automatic testing, the actual severe defects can be confirmed by failures of test cases. After defects are fixed, similar types of defects tend to reoccur again. In this paper, we propose a SoftWare IMmunization (SWIM) method to combine static analysis and automatic testing results for detecting severe defects and preventing similar defects from reoccurring, i.e. to have the software immunized from the same type of defects. Three industrial trials of the technology demonstrated the feasibility and defect detection accuracy of the SWIM technology.

No abstract

No abstract