Modular typestate checking of aliased objects

摘要

In conjunction with the program committee, it is my pleasure to present to you the research papers for the 2007 Conference on Object-Oriented Programming Systems, Languages, and Applications.After more than twenty years, ooPSLA remains a dynamic force for change and advance in the state of the art, as evidenced by the diverse program of 33 papers.I hope you find it as interesting and enjoyable as I do. >This year we accepted 33 out of 156 submissions, the highest number over the past ten years. This was a deliberate choice. In recent years there has been much debate in the field about whether our conference system (which is now widely accepted as a publication venue for tenure cases) has become overly selective to the point that authors tend to submit more conservative papers, as ACM president David Patterson cogently argued [1]. ooPSLA has been at the vanguard of addressing this problem with its Onward! And Essays programs, but I felt that we should apply some of the same ideas to the research program as well. >To implement this I first removed any particular limit or target for the number of papers accepted. This had the beneficial effect of allowing each paper to be considered independently, and avoiding considerations of whether one paper's acceptance would jeopardize another paper's chances. Secondly, I charged the committee to be "acceptance-positive," to forgive small faults (but correct them), and most importantly for the detractors of papers to give extra weight to the arguments of the proponents. However, continuity was also important. We continued the use of Oscar Nierstrasz's "Identify the Champion" paradigm, which tries to promote the selection of papers that are strongly advocated as opposed to those with good average scores [2]. Finally, I set a goal for the committee to decide all papers by consensus, and in the event that a vote was required to break a stalemate for the vote to only be among those committee members who had provided formal written reviews. >The committee met for two days, May 3rd and 4th , at IBM Research in Hawthorne, New York. In the event, all papers were decided by consensus and without recourse to voting. William Cook, the past chair, chaired the discussion of papers with which I had a conflict of interest. The modified system resulted in the acceptance of some papers that might otherwise have been rejected as too controversial or as more intriguing but less fully developed. The acceptance rate was 21% (up from 17%in 2006), so while we accommodated additional papers, ooPSLA remains a highly selective conference -- a healthy balance. >Two other issues regarding our community's conference system have recently been the subject of debate (and experimentation): double-blind review and submission of papers by members of the program committee. Allowing submissions by the committee increases the pool of submitters and increases the quality of the committee since it does not force them to choose between serving the community and publishing their own work, an especially difficult choice for academics who must consider not only their careers but those of their students. However, there is also the danger that such papers might receive preferential consideration. I chose the middle ground of allowing submissions by the committee but subjecting them to a quantitatively higher standard than other papers, and obtaining five reviews (rather than three for other submissions). Of the seven submissions by committee members, two were accepted. Of the rejected committee submissions, two had rankings that would otherwise likely have led to acceptance, but were not accepted according to the more stringent requirements (at least one A and no C's or D's, or else at least three A's and no more than one D). >Several SIGPLAN conferences have recently begun using double-blind review, a practice that is prevalent in some other subfields of computer science. The purpose of double-blind review is to increase fairness by eliminating bias (either conscious or unconscious) based on the identity of the authors. However, double-blind review can introduce other fairness issues:the required anonymization can make it more difficult to evaluate the work in the context of its infrastructure, and there is the potential for primary or secondary reviewers to be unknowingly assigned to review a paper with which they have a conflict of interest. >I chose to use non-blind submission for three reasons: first of all for continuity, since I had made other changes to the policies and processes, secondly, because of the fairness trade-offs mentioned above, and thirdly, in consultation with the chairs of other primary SIGPLAN conferences, to provide a basis for direct comparison of the two processes within a single year. I welcome your feedback on these and other issues regarding the review process.