Wednesday, July 30, 2008


Conference proceedings used to be dirt cheap. The ACM and IEEE would supply proceedings to a conference at a small loss because they could sell later copies of those proceedings to libraries and individuals at a large mark-up. Conferences would order extra proceedings because they could sell the excess at double the price during the conference to attendees who wanted copies for friends. I remember going to STOC or FOCS and buying an extra 7 or 8 proceedings and shipping them back to Chicago for those who didn't go.

But that's all changed. Libraries now subscribe to digital libraries—they don't buy paper proceedings anymore. Hardly anyone ever opens their proceedings anymore after the conference ends so no extra proceedings are sold. The per proceedings price goes up as the number printed go down and conferences try to order just to cover the number of expected attendees. For a typical medium-sized conference, the proceedings can add $50 to the average registration fee. That's expensive for a book that will only be used for a couple of days. So why should conferences stay with paper proceedings instead of going electronic (via CD or Internet)?

  • Tradition.
  • Some people like to look at a paper for a talk during the presentation. Sometimes I see a speaker say something that doesn't sound right and looking in the proceedings to figure out what they really meant. Some other people even take notes in the proceedings.
  • Status. Until STOC and FOCS move to electronic proceedings, other theory conferences might worry about their relative importance if they don't do paper.
  • Authors like to see their papers in print. And the vast majority of attendees at any theory conference are authors.
Eventually this point will be moot when electronic books become an acceptable reality. Even today the NSF could save money by buying each of their PIs a Kindle and refusing to fully reimburse conference fees that have paper proceedings. But computer scientists always do seem behind the curve in adapting new technologies.


  1. What kind of ego trip is that? "Oh! I see my name printed on paper."

    Also, is there anyone who still thinks that a conference with paper proceedings is more prestigious... in 2008?

    Paper proceedings are a negative point for me. Carrying a large book breaks my back.

    Don't even get me started about carrying a large book at the airport... these things never fit right in my luggage.

  2. "Until STOC and FOCS move to electronic proceedings, other theory conferences might worry about their relative importance if they don't do paper."

    Oh, please. Get over yourself.

    SIAM is supposed to distribute electronic copies of the SODA proceedings next year instead of paper, perhaps with a paper book of abstracts. I don't expect this to lower the SODA registration fee, of course, but it would save trees and back muscles.

  3. Authors like to see their papers in print.

    Perhaps more importantly: a dean (say the dean in a college of arts and sciences or in a college of engineering who is not a computer scientist) may only value papers that are published in print (for purposes of evaluating an assistant professor for tenure, evaluating a Ph.D. graduate for an assistant professor position, etc.).

    Since in many other fields conference papers tend to hold little or no value, having our conference papers published in print proceedings makes it easier to argue their value outside our field. Having the print volumes generally helps in explaining that publishing in (certain) computer science conferences is as good or better than publishing in journals.

    For whatever reason, many academics in other fields remain reluctant to accept the role of conferences in computer science. For the sake of our young reseachers' careers, the TCS community (or broader CS community) must keep this perception in mind and work to adjust/update this perception if we are going to move away from print and move to electronic-only publications.

  4. I still have my old copies of many FOCS proceedings, but have dumped all my copies of the STOC (and SODA, and SoCG) proceedings. However, this should not be taken as indicating any prestige advantage of FOCS over those other conferences: rather, it's my reaction to the relative ease of accessing ACM's digital library versus IEEE's.

  5. Let's go over these reasons again:

    1) Tradition. Always a good reason to do something, because we're too lazy and stupid to change.
    2) Looking at a paper during a talk: A potentially valid minor reason. But most of the benefit would be obviated by handing out a DVD or making papers available online (only to conference participants, if desire) since most everyone is toting a laptop these days.
    3) Status: You've got to be kidding. Who cares what STOC/FOCS is doing? And anonymous #3, I don't know a dean (or an applied scientist) who cares if your conference paper is in a paper proceedings. If they're the type that dismisses conferences out of hand, they don't care about the paper. If they aren't that type, they care about the reputation of the conference.
    4) Authors like to see their paper in print. Hunh? I want to see my paper publicly available on the web as soon as possible so anyone can access it -- I don't care about it being on paper! I reach many more people having the paper online than I do having conference attendees pick up tomes.

    The only reason that I see any merit to is #2... and I don't see that much merit in it, actually.

    Now, on the other hand, there's the valid counterarguments -- cost, waste (of paper), having to lug an extra book through the airport, slowing down of the important task of getting the papers online, lack of innovation (get this -- some workshops/conferences I know -- like, say, the major networking conference SIGCOMM -- put papers up online, and allow people to comment/post reviews on papers right away, making the whole thing potentially more interactive!), etc.

    Seriously, this is a no-brainer. Most conferences I go too have already switched to electronic or DVD-based proceedings. Theory is way behind the times on this. Networking conferences have already moved on, and information theory conferences started even before that if I remember right.

  6. There is an inaccurate dollar cost for printed proceedings given in the post. It is at least important to get the facts right when we discuss the issue.

    Through IEEE publishing there is NO difference between the cost to the conference of paper proceedings and electronic proceedings published on CD (except for the small difference in the shipping cost).

    Through ACM about half the cost of "printed" proceedings for a good size conference like STOC is fixed editorial cost of compiling the proceedings. The proper incremental comparison is about $4 per CD versus about $20 per paper proceedings of 800 pages. For smaller conferences the fixed editorial cost is a larger percentage which is why the proceedings end up being so pricey.

    "Editorial cost" includes producing tables of contents, indices, cover art, and a lot of stuff that one could easily automate along with the organization imprimateur, ISBN, and distribution after the conference is over.

    It is true that conferences no longer can make money by selling printed proceedings at a hefty profit for those not attending (though all our extra printed copies for STOC 2006 sold out) but the remark about libraries does not seem to be correct for major ACM/IEEE/SIAM conferences like STOC/FOCS/SODA. (I just checked the UW library catalog.) What is true is that libraries don't get the Springer LNCS in hard copy (ICALP/CRYPTO etc). The major problem is not the cost of each proceedings volume but the fact that Springer will publish virtually ANYTHING in the LNCS series and there is a lot of crap that libraries would have to pay for and shelve that nobody ever would look at.

    As far as I can see the big win for going electronic will only occur if it allows us to cut significant time (more than a month, say) between submission of final articles and conference dates.

  7. Paul --

    Your argument seems to be that the negative of proceedings in terms of cost to attendees is exaggerated. Even conceding that (just for the sake of argument), you don't seem to be expressing any positive reasons in favor of having proceedings, or addressing the other potential negatives.

  8. Are you sure that "hardly anyone ever opens their proceedings anymore after the conference ends"? I do. (Although I print papers from the web when I want to read them fully, I often pull proceedings off the shelf and flip through them when I just want to check a result without reading the entire paper.)

    Even during the conference, I read the proceedings when I am not at the talks (I can't read on my laptop, and it's not always easy to find a printer at a conference) as well as while traveling afterwards.

    There is an argument to be made for not including proceedings as part of registration fees. But let's keep them available for people who want to purchase them. I would guess that at least half of conference attendees would buy a copy.

  9. The cost of electronic proceedings should be extremely low.

    If CDs are as expensive as paper, just have a laptop, and a dozen or so usb drives available at the conference. Use these to allow conference goers to copy the proceedings to their own laptop, or their own usb drive. Also allow conference goers to download to proceedings via wifi. Total Cost: ~$25.

    To organize the PDFs you can just use an html file, for example, the conference schedule webpage.

    Incidentally, can someone explain why PDFs aren't available on the conference website? It would be so easy to automate the creation of a blog with a post for each paper.

    The body of each post would be the abstract of the paper, include a link to the paper, and most importantly have COMMENTS, so folks can discuss the paper.

  10. SIGMOD/PODS-2008 gave proceedings on USB sticks. People who wanted paper-proceedings could buy them at an additional cost.

  11. Michael is right: In my previous comment I did not try to argue for paper proceedings, just that I didn't yet see the big advantage of purely electronic ones in the current publishing environment. (Having them available online is a given as far as I am concerned - either not giving out proceedings at all or distributing them on CD/DVD or flash drives as opposed to distributing printed proceedings is the issue.) Once papers come up online I don't think anyone uses the CDs/DVDs or flash drives with the content. On the other hand, plenty of people use their printed proceedings even afterwards.

    I personally keep the last couple of conference proceedings at home and browse through them looking at random stuff for a period of time after the conference. However, I don't find my personal preferences a particularly compelling argument to foist on everyone (and I read most papers that I already know about online).

    Contrary to what Lance wrote, libraries still do want and buy printed copies. Once they have the printed copy they do not bear significant additional annual costs. This is in contrast to digital libraries where there are hefty annual fees just to maintain access to old content. These are unaffordable for poorer libraries. If we don't produce printed proceedings these libraries will have to do without our content.

    Having the content replicated in many places and not just on a few online servers is a good thing. There have been many instances where other recording media have become unusable but paper seems to have a good track record. It doesn't require that one can plug in or charge batteries either.

    The resources and environmental cost of recycling the waste for batteries and outdated electronic devices is not insignificant and quite comparable to or worse than the aggregate environmental costs of paper.

    Suppose that one could assign the true cost to proceedings as I described in my previous comment. I would argue for the following option at registration: Every attendee pays their share of the fixed costs of producing the proceedings - those who don't want any proceedings pay nothing extra - those who want the CD/DVD-ROM/flash option pay the incremental cost of that - and those who want printed proceedings pay for that incremental cost. Subsequent copies would be available at the the fixed+incremental cost. If one could predict the numbers in advance this seems like a pretty reasonable solution. (Students should probably be exempt from the fixed cost on their first copy.)

    As I said in my previous comment, there might be a compelling reason to go to all electronic if we got a significant structural advantage (that would have to outweigh the loss of paper proceedings). I had suggested a potential one: a significantly shorter turn-around time. In his first comment, which crossed mine in the posting, Michael suggested one that I had not thought of: comment threads on papers as at SIGCOMM. In contrast to SIGCOMM I would be surprised if many of those comments showed up for theory papers. At systems conferences, speakers will be challenged about their papers and there is often an amount of give and take. At major theory conferences it is relatively rare to have more than (or even) one polite question. Posted papers at online sites like ECCC that allow comments almost never get any. I think that this is a failing of theory culture and we would be better off if there were more give and take but I don't see online proceedings as doing anything to change it.

  12. BTW, EC took place 3 weeks ago and no proceedings are available at the ACM Digital Library; any idea when will the proceedings be available?

    If I knew about this delay earlier then I would order the (hardcopy of) proceedings to have it on my desk 3 weeks ago.

  13. I continue to be surprised by the fact that societies like ACM place such a drag on research by selling research literature back to the people who produce it. They make about 20% of their revenue from selling our publications to us, but in my opinion they should be serving their community better by disseminating research as widely as possible without placing friction on it by charging for it. Those of us who do not require paper proceedings are also subsidizing the cost of producing them for the few who want them. I think that paper publishing should exist as a business that does not depend on DRM to lock up scientific publications, and let the price settle to what is required to service that market. If they made it freely available on the web then they wouldn't have to produce paper or CD proceedings for anyone but those who wanted to buy them.

  14. If we had only electronic proceedings, would that prevent us from having to reformat every paper into the less readable two-column format (FOCS camera-ready deadline is fast approaching)? If so, no paper publishing would save the time of converting the formats and ensure that every paper was available in the more readable one-column format.


  15. One concrete disadvantage of switching to an electronic format is that you risk falling down the slippery slope of lower standards for the formatting of papers. The "slippery slope" rationale can be abused easily, but in this case Grant's post above is already one example, in my opinion, and next I expect someone will say that for an electronic proceedings there is no need for page limits. With some vigilance, though, this problem can be avoided, and I myself hate paper proceedings.

    (Parenthetically, I don't find that converting to two-column takes so much time at all, mostly just a little straightforward playing with equations. The bigger problem is that a ten-page two-column paper is longer than a ten-page one-column paper, so switching formats after submission makes authors go back to add more content.)