How to Accomplish Successful Software Engineering Research in Universities (position statement)

1
How to Accomplish Successful Software Engineering
Research in Universities (position statement)

• William G. Griswold
• Computer Science Engineering
• UC San Diego

2
SE Research in Universities(SERiU)

• My approach succeed by avoiding failures
• Im going to point out some pitfalls to avoid in
  SERiU
• General university research milieu
• Funding by NSF and DARPA at all-time low
• 5-10 funding rate at NSF
• Short promotion cycles encourage low-risk (and
  low impact) research
• Whats unique to SE research?
• The study of the effective creation of software
• focus on artifacts (programs), processes,
  tools
• Practicality is inherent issues of scale, real
  programmers, ultimately cost/benefit

3
Others Opinions on SERiU Failure Taking on real
risk and managing it

• David Rosenblum (University College, London)
• Lack of connection to real problems
• If they just used my formal method
• Stuck in a research eddy (also D. Jackson)
• Kevin Sullivan (University of Virginia)
• All your eggs in one basket (also D. Jackson)
• Failure to fail fast
• Daniel Jackson (MIT)
• Belief that all real systems have (accessible)
  research problems
• Difficulty in finding exemplars small
  problems that capture essence of real problems
  (e.g. Parnass KWIC)

4
Problems Ive Encountered

• Attracting strong students
• Lack of training
• Lack of appreciation
• Have mostly recruited older students
• Achieving experimental validity realism with
  experimental controls
• Artifacts large, complex, in use, changing
• Tools extension of tools in actual use
• Processes in situ deployment measurement
• Acceptable pace of research
• Building, measurement, and analysis

5
My InsightIts all about scale

• Problems encountered in research are due to
  the fundamental SE challenge of scale
• PhD-bound college grad cant appreciate scale
• 10-15 week class projects little time in
  industry
• Hard to replicate scale in the university
  laboratory
• Addressing scale requires large effort
• More students, time, money (all in short supply)
• Especially hard on young professors

6
My Solutions for Pitfalls of SERiU

• Attracting students
• Work on cool artifacts, use cool tools, etc.
• Train them on the job
• bKeep recruiting older students
• Achieving experimental validity
• Collaborate (try state grant programs like UC
  MICRO)
• Use open artifacts tools (e.g., tools like
  VisualStudio, Phoenix, Eclipse)
• Acceptable pace of research
• Collaborate build a hierarchical team
• Cultivate secret weapons

But then youll need a tool-supported software
process )
7
Researching Outside the Box

• Ask the right questions
• A small project neednt have small impact
• Answer research questions more than build
• What parts can you skip or fake?
• Make lemonade look at your own failures for
  ideas
• E.g., Why doesnt my tool find those bugs?
• Can achieve deeper understanding of the
  phenomena, not just improvement

C.f. Freeman Dysons Infinite in All Directions
8
(No Transcript)