The Future of Software Engineering: Tools Michael Ernst MIT Lab for Computer Science http://sdg.lcs.mit.edu/~mernst/ [Be excited! Have a twinkle in my eye! This is good research; it is exciting; share the fun I had and enthusiasm I feel.] [Smile when I make a joke -- don’t be too dry, don’t sound mean.] [Don’t go too fast through first part of talk.] [Make eye contact.] This is a technique for helping programmers to cope with inadequately documented programs. Invite questions and comments. Daikon: an Asian radish (not the Asian radish) Perhaps don’t say this, but this is but a summary: I can’t talk about every aspect of this research, but welcome questions either now or in individual meetings.

Tools matter The most productive people are the most effective tool users Tools are a change agent

Reproducibility Replication is key to the scientific method Papers may gloss over details Insist on published systems Permits additional evaluation Stop accepting papers without supporting data and implementations Central repository for such information Need support for infrastructure-building

Counterargument to publication Responsibility to have commercial impact Can have commercial impact without IP If it's a good idea, it will be adopted Secrecy is antithetical to science & education

Evaluation of implementation Tools should work for users other than implementers when run on production systems Correct the impression that software engineers do not build real systems

Evaluation of tools Humans complicate systems Theorems much easier to understand! Case studies, not experiments Devise rewards for reproducing experiments Special publication venues? Do not raise the bar for new ideas Industry can assist academia

Ignore the average programmer Practitioner/researcher gap What is the job of researchers? Researchers must assume best practices, in order to have long-term impact

Static and dynamic analysis How to obtain information: Programmer-supplied Static analysis: examine the program text properties are guaranteed to be true pointers are intractable in practice Dynamic analysis: run the program efficient, precise complementary to static techniques

Combining techniques Use static to help dynamic and vice versa Transfer from one domain to the other Example: Purify and LCLint Combine in a principled way Select the desired efficiency/soundness

Lightweight tools No need to produce exact results Useful results that people can check

The return of formal specifications Full formal specifications do not work Partial formal specifications: Specify only certain aspects of behavior Generated automatically Other examples from languages and compilers Example: functional programming

Summary Tools are key Publication must include tools Case studies Assume best practices Static and dynamic analysis Lightweight tools