The 2006 International Symposium on Memory Management (ISMM2006) was held June 10-11, 2006 in Ottawa, Canada. ISMM 2006 issponsored by the ACM Special Interest Group on ProgrammingLanguages (SIGPLAN), with additional financial support from IntelCorporation. ISMM is the leading forum for researchers, developers,practitioners, and students to present research on memorymanagement and garbage collection. Previous symposia in this seriesoccurred in 2004 (Vancouver), 2002 (Berlin), 2000 (Minneapolis),and 1998 (Vancouver). The earlier workshops (IWMM) occurred in 1995(Kinross, Scotland) and 1992 (St. Malo, France). Beginning thisyear, ISMM will be an annual symposium.We thank the authors of all submitted papers. Receiving a numberof high quality submissions is obviously necessary to continuing tooffer a high-quality program. We thank the program committee fortheir significant investment of time reading papers, writingdetailed reviews, attending the program committee meeting, andselecting the program. They worked hard to offer useful feedbackfor both selected and rejected papers. An additional 16 outsidereviewers also contributed expert reviews, for which many thanks.We especially thank IBM Research and, in particular David Bacon,for hosting the program committee meeting at their facility andproviding food and refreshments to keep us going. Thanks to RichGerber for his timely support of the START web-based conferencemanagement system.
In selecting the program committee, we followed SIGPLANguidelines, working to balance a combination of factors such asresearch specialty, seniority, gender, academia/industry,geographic location, and rotation over time. We allowed submissionsby PC members (but not the PC chair). We invited 14 individuals toserve on the PC to obtain a committee of 9 (plus the chair). Thelist of invitees was reviewed and approved by the ISMM SteeringCommittee. We had 45 submissions, from which we selected 17 toappear in the program. One submission was withdrawn. Therewere 3PC submissions; all were accepted. Each paper received at least 3reviews; a few received 4 or even 5. Thus, each PC member reviewedabout 14 papers.
We employed two review processes new to ISMM. First, we reviewedunder conditions of <i>author anonymity</i>, that is,with the reviewers not knowing the identity of the authors(sometimes called "double blind" reviewing). This worked fairlywell, but, as might be expected with something new, there were someglitches. A handful of authors did not obscure their identities.Most of those quickly submitted revised, anonymized, versions oftheir papers upon request. The 2 remaining papers we processedanyway because the requirement was new and not consistentlyadvertised in advance. We recommend better and consistent notice toauthors in the future, and that the PC chair check the submissionsimmediately after the deadline to give authors a short opportunityto correct papers that obviously reveal the authors. There wereonly a couple of cases where authorship was revealed by other means(e.g., a visitor giving a talk). We felt it adequate for thereviewers to indicate this in the private-to-the-committee sectionof their review.
Second, we included an author response opportunity. Most authorstook advantage of this. It is not clear whether it changed anyoutcomes, but it is possible, and we felt it a worthwhile practice.The committee suggests allowing more time between the authorresponse period and the PC meeting, to allow the committee to readand consider the author responses ahead of the meeting.
PC members declared conflicts of interest in reviewing papersfollowing the ACM guidelines. We note that this was unexpectedlycomplicated by our institution of author anonymity (How can onedeclare a conflict with an author if one does not know who theauthor is?). The provider of the START conference managementsoftware, Rich Gerber, worked with us (and some future SIGPLANconference PC chairs) to devise ways for (a) the PC chair to enterconflicts, and (b) forPC members to indicate institutions, etc.,that are conflicts for them, which the system can check moredirectly. (We could use only (a) for ISMM 2006.) At the PC meeting,before discussing a particular paper, we revealed its authors. Thiswas important since it could (and did, in one or two cases) revealpreviously unknown conflicts. Those PC members with conflicts leftthe room; if the PC chair had a conflict, he designated anon-conflicting member to chair the discussion in his absence.
We rated papers on a four step scale: A (will argue to accept),B (prefer accept but will not argue for it), C (prefer reject butwill not argue for it), D (will argue to reject). Reviewers alsoself-rated their expertise in a paper's topic area on a three stepscale. We did not discuss papers having no A or B scores, unless aPC member chose to add them to the list to discuss.
The committee would like to emphasize their displeasure withthose authors who did not revise their work rejected by previousconferences in line with previous reviewer comments. It is an abuseof reviewers to send them the same thing again.
We held the PC meeting at Hawthorne, NY on March 2, 2006, withall members present except one participating by speaker phone. Wediscussed about two thirds of the submitted papers, hearing fromall the reviewers, with the highest rating reviewer starting off.We did not return scores to the authors, either in the authorresponse period or after the meeting. A number of reviews wererefined as a result of the meeting, in hope of helping authorsunderstand our decision and to revise better for future submissionelsewhere. We received a few papers that the committee felt wereout of scope for the symposium, and we suggest the next committeecarefully review the call for papers and perhaps further clarify.For example, we felt that compiler optimizations and analyses notdirected at memory allocation, reclamation, or locality did notfit. We identified several papers for PC mentoring("shepherding").
In any group of papers,some are technically stronger thanothers. We felt it important to include in the program a number ofpapers that, while not ideal, we found interesting, intriguing, orof expected value to the community. Each selected paper wasallotted 25 minutes for presentation and 5 minutes for questions.We also included an invited talk (the invitee chosen by the generaland program chairs in consultation) and a Wild and Crazy Ideassession, organized by Chandra Krintz, whom we thank for undertakingthat effort.
收起