Sunday, April 20, 2008

A new (to CS) model for publishing

One of the things I like about the database community is their willingess to play with new ideas in the space of conference publishing. SIGMOD and VLDB have been experimenting with the idea of semi-persistent reviews, where reviews from SIGMOD get passed on to VLDB for papers deemed on the border; SIGMOD went to double-blind mode, over some opposition, and there's been some interesting back-and-forth since on the effectiveness of this (read this, and then this). There's also a weak rebuttal mechanism (where authors can in a limited way respond to reviewer comments in the submission process).

An even more radical idea, which from the sound of it is nearing approval, is described in detail by Panos Ipeirotis. The main points are these:
  • A new journal called the Journal for Database Management Research will be created.
  • It will have a fast review process and turnaround time (akin to the biological journals - 3 months or so)
  • Submission deadlines will be rolling: i.e you submit a paper when it's ready.
  • SIGMOD, VLDB and other database conferences will convert to a by-invitation model, where the conferences choose a sample of the published works in the journal (over the last 12 months I imagine) to be "presented" at the conference.
  • To discourage frivolous submissions, papers rejected from the journal will have to undergo a year-long cooling off before they can be resubmitted.
It's a radical approach, and approximates to a degree the prevailing practice in journal-based publication environments. It does raise some questions (some raised by Panos in the original post):
  • The year-long cooling off seems excessive punishment for what will still by necessity be a less than perfect review process
  • How will this new journal interact with other database journals ?
  • Can one journal hope to keep up with the volume of papers being produced ? Just SIGMOD, VLDB and ICDE take in over 2000 submissions between the three of them. That's close to 6 submissions EACH DAY.
  • What happens when you get into areas that overlap with DB ? For example, what about KDD, and other data mining conferences ?
  • What about all the years and years spent arguing with tenure committees about the primacy of conferences in computer science ? "Oops ! we changed our mind" ?
  • What kind of prestige will now be attached to giving presentations at the "new" conferences ? More importantly, since it was JUST ESTABLISHED that double blind submission helps remove bias at conferences like SIGMOD, isn't this a step backward in terms of which papers are chosen for presentations at conferences ? I can't imagine the process of getting invited for a talk at such a conference getting easier with this process. Are we heading towards (again) the bio model of big plenary talks by bigwigs, and lots of posters, or the math model where anyone who wants to give a talk can ?
Separating the idea of publication and dissemination is dear to my heart (I have always felt that conferences in CS fail by needing to serve both these masters at the same time), and so I'm bullish on proposals like this. But I do see problems in the details, and am curious to see how things pan out over time.
Post a Comment

Disqus for The Geomblog