Monday, August 17, 2009

Program Committee Duties

Since I'm thinking about SIGCOMM (going on this week), even though I'm not sure the official cfp is out, I assume it's OK to announce that I'll be on the SIGCOMM PC for 2010. This means I'll be repeating both NSDI and SIGCOMM PCs from last year.

My understanding is that this year they're not going to do the heavy-light split they've been doing in the recent past with SIGCOMM, where only a subset of the TPC is at the final PC meeting. Instead, they'll just be the one committee. With about 50 people.

I admit to having some concerns about how a PC meeting with around 50 people is going to run -- but that's not my job. I'd also like to point out the approach here: around 300 submission, 50 people on the PC, giving a review load of roughly 24 papers per person (assuming an average of 4 reviews), and no sub-refereeing (although experts outside the PC will be asked for reviews as needed). That's very different than the process on many other conferences I've been involved with, particularly the FOCS/STOC/SODA model -- about the same number of submissions, a PC about half the size, and lots of subreferees.

I'm not saying one approach is better or worse than another -- perhaps they're both good for their respective domains -- but the differences are worth commenting on, if anyone wants to comment.

Finally, it's at least somewhat unlikely I'll be motivated to "3-peat" for the NSDI or SIGCOMM PC for 2011. I imagine I'm filling an "network algorithmist" slot on these PCs. If anyone has any good ideas for other names I can pass on the PC chairs, feel free to comment (or send me mail directly).


Anonymous said...

The CFP is out. You are safe in revealing that you are in the PC :-)

Suresh Venkatasubramanian said...

SODA is actually significantly larger than FOCS/STOC in submissions: 450 vs 300. Which also means that even after assigning external reviews, one ends up with 20-25 reviews (unless you're REALLY good at finding external reviewers)

Anonymous said...

Some possibilities: muthukrishnan, moscibroda, kuhn, scheideler, srinivasan, goel.

Anonymous said...

I was on a recent soda PC in which some of the PC members sub-refereed EVERY SINGLE paper. Generally, this was not a sign that that particular PC member cared very much about the decisions.

Paul Beame said...

Such a large face-to-face PC presents problems both from a logistical point of view and from a financial one - a conference would have to be quite large to afford such a large PC. (Either that or PC members have to be pretty flush to pay their own way.)

I was on a recent soda PC in which some of the PC members sub-refereed EVERY SINGLE paper. Generally, this was not a sign that that particular PC member cared very much about the decisions.

Getting good sub-referee opinions is very worthwhile but one of the particular drawbacks of an all-electronic PC meeting - it is very easy for PC members to hide. I have been on such PCs where members have done this for large numbers of papers and then have not participated in any electronic discussion afterwards. It really doesn't work.

It has been a policy of SIGACT since STOC 1999 that PC members are supposed to choose subreferees based on specific expertise and are supposed to do more than simply pass on the work of their subreferees. They are also at least supposed to provide their own evaluations of those subreferee scores and opinions.

Michael Mitzenmacher said...

Paul -- for networking PCs, the standard seems to be that people pay their own way for the most part.

In fact, just to be clear, in my experience most PC members have to pay a substantial part of their way for theory conference PCs as well; there's just no budget to support paying everyone's travel. For STOC, we arranged to pay everyone's hotel and group meals; that pretty much took up the travel budget. There's often money left at the end (because of how the professional organizations make you budget), but you can't be sure that money will be there, so people generally pay out of their own funding for what's left.

Paul Beame said...


I disagree about theory conference PC expense coverage. For STOC 2006 we budgeted $625 per PC member for travel expenses. At least $500 per PC member has been typical for many years. The actual travel expenses reimbursed vary across the board - especially for those coming from overseas - but it all works out. Typically, those PC members in industry get their travel expenses covered by their company but those in academia do not.

FOCS budgets have often been much tighter with $5K to $8K total budgeted for the PC travel for at least the last 15 years though some PC chairs have argued successfully for more. A judicious choice of PC meeting location that is in the area with many PC members also can keep the costs down.

One other thing that often happens is that expenses only get reimbursed months later because the money only comes in around the time of the conference.

Michael Mitzenmacher said...

Paul --

I don't think we're disagreeing, so I'm not sure why you're saying we do.

For STOC we had budgeted, if I recall correctly (and I wasn't in charge of the budget, so I may be wrong) about $10000, for a 24 person committee. Since some were local, that averaged to close to $500 per committee member. 2 nights hotel and meals already takes up about $400 of this $500; it's clear you can't cover flights, much less much in the way of incidentals, with what's left. Where we might disagree is that while in the past PC members from industry generally covered their expenses, this year that seemed to be problematic for at least some.

Also, we agreed that sometimes you can only get reimbursed months after the fact -- after the conference is done. The local arrangements team often doesn't have the money to reimburse everything in advance; they have to wait until the end and see how much they have after all the expenses. As I said, usually they have enough (and even some extra) because the controlling organizations (IEEE/ACM) are very conservative in how they budget. But there's no guarantee you'll get the money (if the conference overspends, you're probably out of luck), and it's months away, so most people cover what's left themselves.

Where I disagree with you is I think, for anyone who might be new to serving on theory PCs (or who hasn't served for a number of years), is that you shouldn't expect full reimbursement, or if you do, only when the conference is over. If that's an issue, ask the PC Chair/Local Arrangements people to clarify ahead of time. I had one angry PC member who didn't realize this, even though I as PC chair I tried to make it clear (even though handling the money wasn't really my job -- and, obviously, I must not have made it clear enough).