The ISR Research Forum held at UCI in late May of this year was highlighted by two keynote addresses, one by Dr. Eric Dashofy from The Aerospace Corporation and the other by Dr. Marija Mikic from Google, Inc. Though titled differently and addressing somewhat different concerns the two talks both fundamentally addressed how software can be, and is being, built in industry today. Software architecture was a common theme, but with two very different takes: Dashofy highlighted the limits imposed on principle-driven design by the ever-encroaching maw of frameworks; Mikic highlighted how up-front, clear, documented design coupled with an admirably comprehensive development process rules the day within Google. The ensuing discussion centered around what corporate and/or financial characteristics need to be in place before a “doing it right” approach is both feasible and cost-effective. One answer posited was that it takes an organization with the resources of Google to “do it right.” But other experience belies this. I am familiar, for example, with a small software shop overseas that indeed “does it right” and does so without access to a lot of financial or personnel resources. The issue, rather, seems to be the constraints an organization labors under. When constrained by particular kinds of interoperability demands, procurement policies, or lock-in to some vendor’s products, life becomes much more difficult, and doing it right becomes less of an achievable goal—even though there would be many benefits. This statement is too facile, of course, and could be used to exonerate plain-old poor engineering practice. Teasing out the issues and understanding the core underlying matters is what makes research in this area so interesting. The Forum just scratched the surface, but I think it gave our audience a lot to contemplate.
The degree to which academic research can contribute to practical software development practices is a matter of much debate. Arguably, system design that proceeds methodically from the application of solid reasoning based on analyzed, codified experience is ultimately the fruit of good academic work. I recently attended a UCI-ICS/Engineering alumni event in New York City during which several attendees related stories validating the utility of what they had learned during their time at UCI. While that is always gratifying to hear, my view is that the research field today is, in large measure, moving most unfortunately towards the useless kind of academic work that makes for cynical comments. More than a few recently published papers in prestigious conferences, for instance, give well-packaged experimental answers to questions that no one is the least bit interested in, nor will ever be interested in. A slavish devotion to statistically valid experiments has seemingly pushed system development and technology innovation to the sidelines. Even formal pressures within academia influence this, where shrinking limitations on the time one has to finish all degree requirements is slowly squeezing out room for substantive system building and meaningful partnerships with industry. Schools in the U.K. are probably in the worst shape for this, as there is a 4-year (post Bachelor’s degree) normative time to degree in many schools. Required post-docs anyone?
The difficult matter is determining what to do about this drift and, to the extent it is part of the problem, determining how to “fix” the conferences. The plague of inconsequential results is not confined to software engineering, and in that observation there is hope, for many fields are wrestling with this issue and some potential solutions are emerging. In the database community, for example, the VLDB conference accepts papers continuously throughout the year. Published versions appear in the PVLDB journal, and oral presentations corresponding to only those papers are given at the annual VLDB conference. Arguably this addresses some issues, such as quality and scalability of reviewing, but does not necessarily address the lack of wrestling with consequential issues. For that there must be a mechanism to encourage and reward system building and evaluation that may span several years. Let me know if you have any suggestions.
New Associate Director
Lastly, and on a decidedly happy note, I am very pleased to announce the appointment of Professor Crista Lopes as Associate Director for ISR. I look forward to her contributions and leadership as we work together to determine strategic emphases for ISR.