2012
10.19

From here:

Submission format: Authors are required to submit an extended abstract of at most 10 pages, including the bibliography, in (two-column) ACM-Proceedings format. All instructions, including latex templates, needed to produce a PDF file in the required format are available here. Note that, in a break from STOC tradition, STOC 2013 is using the same format and length restriction for submissions and proceedings contributions. Submissions that do not conform to these requirements will not be considered for STOC 2013.

However:

  1. Two columns with small font is harder to read than regular 11/12pt single column format. Poor referees that would have to read pile of papers in such a format. It also enables submitters to submit more stuff that the referees are supposed to read – bad.
  2. I write the full version of the paper first. Reformatting it to two columns if you have figures is a pain. It is a double pain if you use embedded figures (i.e., parpic).
  3. However, you can still submit, in addition, the full version of the paper. How the PC would use this full version, if submitted, is a mystery, but I assume it would involve human sacrifice and dancing near a lake on a full moon night.

To make it clear, submitting a short 10 page version, and a full version is a good idea (it was done in the last SODAs I think). However, I find the use of proceedings format for the abstract submission to be counter productive.

  1. If the two column tiny font version is so hard to read (and I tend to agree) why would you support continuing to use it in the actual proceedings? I think unifying submission and proceedings is a good idea — that way there’s no question of which parts the author intends to include or not include. And I also think having readable formats is a good idea, but I see no reason why we can’t have both at once.

  2. Yeh. I agree. We no longer should think about our papers as being formatted for proceedings paper format – anybody is going to look on them on a computer, and then maybe print them out. This has several implications:

    – Page limit does not really makes sense any more.
    – Dense unreadable formatting is not that crucial.
    – Use formatting that looks good on a computer (colors in figures, text, etc).
    –S

  3. Sariel,

    I, of course, completely disagree with you.

    First, you seem to be missing the point, which David makes clear — as it is, the final submission format is 2 column if your paper is accepted, so you’re going to have to put it in that format anyway (assuming your paper is accepted). Now, we can argue whether that should be the final submission format, but unfortunately we don’t get to decide that — ACM does. (And trust me, many of us have told them that they should change that, and supposedly they’re working on that.)

    So now take that as a given. Think from the authors’ point of view. Why do you want to create multiple different versions — your “submission version”, your “final version”, your “full version”? You’ve created more work for yourself. (Or, at least, for me, I hate doing this.) For most people (except those who have really long papers, and really can’t put everything in 10 2 column pages), they’ll be able to author one version and be done. Hooray!

    Referees will still be able to read exactly what they want to read — nobody is forcing them to read more. In fact, they shouldn’t have to do the silly work of reading the short version and then hopping over to the full version to find the details they need in another version of the paper. I find it’s less work as a reviewer when I have 1 version of the paper to worry about. Moreover, reviewers also get to see (roughly) what the paper will actually look like in the proceedings (as David pointed out), which I find to be a good thing. Hooray!

    Having written papers for and reviewed for multiple conferences doing things both ways (submit in the final format even if it’s ugly 2-column, and submit in single-column 10 page format which has nothing to do with what the final paper will look like), I can tell you my experience the first approach is dramatically better on both sides of the equation. Try it. You’ll like it.

  4. Sariel, thanks for posting about this. I found the requirement somewhat horrifying as well. First, I have never wanted to read a short version of a paper in 10-page 2-column small-font math-formula-squashed (or overflowing across columns) format if I had an alternative full paper. I don’t remember a single time. I find this to be largely (though not completely) true in CS fields outside of theory (finally I can see more data than just four figures that are supposed to summarize all your findings). I thought the format was a relic from the days when people were forced to carry around giant 1000-page proceedings, in fear that they might grow to 2000 pages and give all Computer Scientists chronic lower back pain.

    The 10-page 2-column format has often been a negative force. It has allowed “due to lack of space, you’ll never see this proof” to be (figuratively) uttered countless times. I also think it supports the mentality that a lot of youngsters have: “Once I prove the result, it’s dead to me; write it up and move on.” It emphasizes papers being result announcements and trophies as opposed to a way of carefully and thoughtfully expressing your ideas to the world. I really, really hope that most people don’t think of their 10-page 2-column abstract as “the definitive version” of their paper. Fortunately, these days it seems that most people just see it as a necessary hoop to jump through until someone can change the policy at ACM (or we abolish conferences altogether in favor of a better mechanism).

    Now the temporary solution that the field seem to have converged on is to write your full paper: the one you care about, the one you’ll post on the arxiv and your webpage, the one you’ve lovingly typeset and proofread page-by-page for weeks and months. Then (for the sake of antiquity) submit an “abstract” version which is basically the first 10 pages with the full version thrown in the appendix. Finally, in the event that your paper does get accepted, as a sort of induction ceremony, you’ll have to spend a couple hours chopping it into an ugly mess to fit it into the proceedings format. Fortunately, you can be assured that very few people will actually have to read this version if you’ve posted an appropriate full version online somewhere (which also has the added advantage of not being behind a pay wall).

    The *additional* problem with the current submission requirements is that (a) I have to present this ungodly 10 page beast to the PC and the referees after spending a great deal of time polishing and perfecting the real version of the paper and (b) I have to do it even if my paper is rejected. I don’t see how these requirements alleviate any problems. Michael is claiming that now people only have to write one version, but this seems philosophically flawed (and largely untrue in practice). The flaw being that somehow the paper is written “for a conference” as opposed to written for the advancement of science and then submitted to a conference for dissemination.

    -j

  5. May be a compromise between James’ and Michael’s positions would have been to ask for a submission in normal single-column format of length 15-17 pages (or whatever number those ten ugly 2-column pages typically amount to). This would serve the objective of the submission being closer in content and length to the final version, and the authors of rejected papers (who will be in the majority) would not have to spend their efforts on a 2-column formatting that doesn’t help anyone (and in particular the PC members who would have to read and evaluate it). As James says, authors of accepted papers might complain a bit about it, but it is not too bad as an “induction ceremony” to do the re-formatting for the camera ready version (though like almost everyone else I’d like to see this 2-column reformatting requirement eventually disappear at that stage too).

  6. Venkat —

    The compromise you suggested was discussed. In the end, the decision was that people should send in the paper that was as close to the final form as possible, hence in the actual 2-column format instead of a nominally equivalent length in 1-column format. I admit either approach would have been satisfactory to me; what I’ve always found ridiculous is the 10-page 1-column submission, which is like writing a whole another paper — and a bad one, as I have to randomly throw things into an “appendix” to meet this rather bizarre page-length requirement. I admit I don’t see the issue with reading 2-column pages, but maybe my eyes are still better than average. (Isn’t that what the zoom features are for on-screen?)

    On the other hand, while it’s perfectly fine to have people turn in “full versions” (and people may optionally do so), I think there are many good reasons to insist on a page limit for submissions. Besides sparing the reviewer, it’s a helpful writing tool; the well-known problem in writing is people write too much, not too little, and pushing people a little to edit things down is a net positive overall. (In my college writing class, the teacher liked to take our essays, cross words out, and give us notes like, “I reduced your word count from 1000 to 770 with no loss in meaning.” It was very useful.) I think the goal should be to choose a page limit that encourages people to self-edit to a reasonable amount, and simultaneously let them include their “full paper” or a good approximation a large fraction of the time. 10 pages should be a good approximation of that number — that’s what FOCS/STOC papers have been for a long time.

    In short, I think I hear from James some agreement that the old system was annoyingly bad, which I agree with. I’ve also heard that we as a community would like a different format for final versions — like, single column, with enough pages (say, maybe, 20 — not this annoying 12-page LNCS stuff either, which is about 3.5 real pages). I agree with that too, but that’s not currently available to us (and we’ve let the powers that be know that’s what we want). (To be clear, I disagree with any notion of unlimited pages for conferences.)

    I’m not clear I get the other points of disagreement, such as James’ assertion that it’s the 10-page 2-column format that somehow is a negative force pushing people to be what he considers to be poor scientists. I think the same problem occurs under other formats or the current system, and it’s an orthogonal issue.

  7. I agree with others in my dislike of the 2-column format; but I want specifically to add that the STOC’13 submission requirement could create a disincentive to properly revise accepted papers in the light of reviewer comments. Fitting it into 10 pages and avoiding equation overflows is always a tedious, ad hoc business, and I only want to do it once.

  8. If you want to mandate that a paper be well-written, do it directly. Instruct the referees to consider heavily the quality of the writing. But don’t make arguments like “requiring a short submission has the side effect of encouraging better writing.” People generally don’t shorten their papers under pressure by deciding to suddenly write more elegantly.

    A well-written paper, regardless of length, contains an introduction that explains the problem, the philosophy, the historical context, and the novelty of the content in a timely manner. This is even more important for 50-page papers. Here’s my proposed solution: Submissions can be of any length, but the PC is only required to read the first 10 pages. This does not require any additional jumping through hoops until the paper is accepted. Furthermore, it allows the PC and the referees to see the content presented in the manner that the author intended. (No more seeing a reference on page 42 of the appendix and trying to remember that the bibliography is on pages 23-26.) Whether we require a short abstract for accepted papers (which I actually could see as having merit) is a separate issue.

  9. I agree completely with James. As for Michael comments – I think the main difference is that I consider the version to be published in the proceedings to be worthless. I find it hard to read (well, my eye sight sucks), ugly, and usually incomplete. (Page limit? for proceedings that are not printed – well that makes complete sense. Not.) I would be happy to replace current proceedings version with a one page abstract and a link to an arxiv version.

    What I find upsetting is the *requirement* to submit in two columns. Why would the PC forbid me from submitting in a more readable format???

  10. My opinion is that the submission format should be selected so as to best facilitate the difficult work of the program committee, which has to evaluate a huge number of submissions in a short time. I believe the traditional 10 pages plus appendices for additional details works very well for this purpose. I recognize that allowing the submission of a full version (with the proviso that the PC may only read the first 10 pages) or using the same format for submission and final version can save authors time, but my opinion is that neither is as good as the traditional format for the PC review process. The small font in the camera-ready format puts a strain on the reader’s eyes; while I read on a tablet and can zoom, many people still print papers to read them (at least that was the case for the STOC 2011 PC). And the first 10 pages of a long paper, even with a well-written intro, generally does not convey as much useful information as a 10-page submission where the authors have selected what details are most important for the PC to see (the same way an optimal 20min talk is not simply the first 20min of an hourlong talk). The point of the length limit for submissions is simply that the PC has limited time, which is a separate consideration from the page limits in proceedings (which used to be due to cost and size of the proceedings), though one could make a reasonable argument that the proceedings should not contain much more than what was actually reviewed.

    Personally, I’ve never found separating & distilling a paper into a 10-page body and an appendix to be a particularly time-consuming or difficult process. (On the other hand, getting a paper to compile properly in ACM format is more painful… and really is wasted work if the paper is not accepted and then gets resubmitted to a non-ACM conference.)

  11. Hi Salil,

    I agree with you on both counts. The most important thing is to facilitate the work of the PC. And the first 10 pages of a longer paper does not necessarily qualify as what you would suggest reading if you knew the referee only had 10 pages of time to spend on your submission. But a “we only promise to read the first 10 pages” standard would allow the authors to arrange those 10 pages however they wished (knowing full well that the PC is making acceptance decisions based on them, of course!)

    On the other hand, I have many times been frustrated by the 10 pages+appendix approach. For instance, one finds the statement of a lemma with a proof deferred to Appendix D. Then in Appendix D it says “Proof of Lemma …” But now since I’m in an “appendix” (which the PC probably won’t read anyway), the author has somehow forgotten that I enjoy having a friendly guide to all of these equations. I don’t think inequality (11) is correct, but I can’t really be sure I’m not confused about the notation which is back in Section 2. The proof relies on Lemma 2.2 which is not in the appendix, but stated in the main body. I flip back to the main text, where Lemma 2.2 is stated, with a reference to Appendix D.3 for the proof. Appendix D.3 says that the proof follows as in [23]. I try to flip to the bibliography–somewhere in the middle of the paper between the “extended abstract” and the appendix–while keeping my fingers lodged in Sections 1, 2, and Appendix D. At this point, the solution is clear: Do a google search and cross your fingers that the author has posted a full version on their web page. (Yes, it’s possible for these issues to exist anyway, but I find them far more prevalent in 10 page+appendix submission. YMMV.)

  12. BTW, James, a solution I started using lately is to have page numbers in the references. At least it makes this going back and forth a bit easier. I have things like “Lemma 2.1” and in subscript the page number. See for example:

  13. Hi James, I agree – the appendices that authors produce are often not so reader-friendly (typically due to a rushed cut & paste process happening in the hours before the deadline – I’m as guilty as anyone else about this). One solution, that has been explicitly allowed in some recent CFPs, is to allow the appendix to simply be the full paper. This forces authors to think about what material the PC should actually read in the process of extracting a 10-page main body, while also giving the reviewers access to a coherent full version. – Salil

  14. It seems strange to me that the community would collectively have to beg the ACM to format our papers better, when as James points out, reading the arxiv version is in most cases preferable anyway. Why not drop the printed proceedings altogether, and replace them with a list of links to the arxiv? Then we could require that conference submissions be in whatever format we thought optimal. If we need printed proceedings to build tenure cases, then we those printed proceedings could consist of author lists, titles and arXiv numbers. Maybe abstracts too.