Principles for SIGCOMM

In an effort to encourage more discussion about how we might improve SIGCOMM-sponsored conferences, four of us – Fabián Bustamente, Nate Foster, Aurojit Panda, and Scott Shenker – posted a series of documents here, and announced them in this channel here, here, and here. Our last posting was on July 24th and much has happened since then, most notably: (i) an announcement by Ellen (in this channel) about the EC’s future plans and (ii) the Community Meeting at the SIGCOMM conference.

Now that this channel’s discussions have become less active, we wanted to share our thoughts on what has happened, what should happen, and what we fear will happen.

What has happened: The online and in-person exchanges about the future of our conferences have been admirably thoughtful, but they have focused more on specific mechanisms than on the goals we hope to achieve for our conferences. We would like to redirect the community’s attention to the two goals we see as paramount in helping our community best pursue its intellectual mission.¹

1. Accept papers that represent a broader range of contributions. This goal is not, as some people seem to think, to accept more papers. Instead, this is asking PCs to be more willing to accept different papers, those that fall outside of the typical SIGCOMM norms.

2. Expand conference programs to include a wider range of content-oriented activities, not just talks on papers. Our community has very few occasions to engage each other in discussions about the directions of the field. We should use more of our in-person time at conferences to do so, in a variety of forms (e.g., panels, keynotes, debates, etc.).

What should happen: First, we should debate these goals. If these are not the goals that the community favors, people should put forward other goals they think are more important to SIGCOMM’s intellectual mission. Whatever our goals are, we must find a way to implement changes that move us towards those goals. We strongly endorse the approach put forward in Ellen’s aforementioned announcement. To wit, we should establish an open process whereby: (i) anyone can propose changes, (ii) these changes are discussed by the community in an open process overseen by a committee devoted to ensuring that all perspectives can be heard, and (iii) in most cases, individual PC chairs or steering committees can choose to experiment with such proposals without obtaining the consensus of the community. See this essay for a justification for this decentralized and non-consensus approach.

What we fear will happen: To be more specific and concrete, we now focus our attention on the SIGCOMM conference itself, not on all SIGCOMM-sponsored conferences, What we fear will happen is that the discussions will continue to focus on individual mechanisms rather than on the desired goals, and as a result we make decisions that undermine goals that are widely supported. We offer two such examples. First, we think there is widespread agreement that SIGCOMM should help further the careers of our graduate students. For many this equates to preserving the practice of having the conference program dominated by paper presentations so that graduate students (who are the majority of speakers) get exposure and experience. However, if SIGCOMM, as a conference and a community, does not ensure the intellectual vitality of the field, the future prospects of our students will be harmed. There seems little question that a more varied conference program, with a wide range of activities, would do more to encourage that intellectual vitality than a program exclusively composed of talks on papers. Thus, we should not blindly assume that the short-term benefits to students of giving conference talks outweigh the long-term benefits of creating a more vibrant intellectual environment.

Another point of widespread agreement is that we want our conference program to retain its reputation for high quality. As a result, many worry that proposed changes in how SIGCOMM PCs decide to accept papers will lead to an unacceptable lowering of paper quality. For many such proposals, such as this one, we think this concern is misplaced, but such concerns are nonetheless widespread. These worries coexist with significant enthusiasm for having multiple submission deadlines for the annual SIGCOMM conference. The experience at NSDI and in the security community suggests that going to multiple deadlines leads to a jump in the number of accepted papers, which would almost certainly result in a significant decrease in average quality. To reject small reforms of our acceptance procedures yet welcome a massive increase in paper acceptances at our annual conference would be pure folly, leaving us with the worst of both worlds; many more papers similar to what we typically accept, with little attempt to broaden our horizons.

As argued in the essay referenced above, we do not think we as a community will come to consensus on a set of goals. However, when discussing a particular mechanism we must specify what goals it is intended to achieve, so its utility can be evaluated. This will help clarify if our community’s disagreements are about the intended goal, or about the mechanism’s ability to achieve that goal. To illustrate why this is so crucial, we note that the lack of clarity about goals has hampered the discussion about open reviewing. Open reviewing has been enthusiastically advocated by some, yet for many of us the anticipated end result remains ill-defined. To reasonably evaluate this proposal, and all other proposed mechanisms, we should insist that the goals be explicitly stated, so that the ensuing discussion can be clear and constructive.

Jul 24, 2023

Notes