The proceedings of this conference are being published by A K Peters, Ltd., in three hardcover volumes, totaling nearly 1400 pages, under the title Number Theory for the Millennium I, II, III. The proceedings contain 72 papers based on talks given at the conference, including contributions from nearly all plenary speakers and from most of the invited half-hour speakers. To put this in perspective, with 1400 pages these proceedings are about three times the size of the average conference proceedings, and they represent the equivalent of 14 issues of Acta Arithmetica, 7 issues of the Journal of Number Theory, and three years (12 issues) of The Ramanujan Journal. The Millennial Conference Proceedings are probably the largest proceedings of a number theory conference, and there exist very few proceedings or collection volumes of similar magnitude.
Producing these proceedings was thus a major undertaking with few precedents, and required dealing with a number of issues - both technical and non-technical. This page describes in some detail how this was accomplished.
Editorial processing. The editors of the Proceedings were M.A. Bennett, B.C. Berndt, N. Boston, H.G. Diamond, W. Phillip, and myself. Each paper that was received was passed on to one of these editors who would then take over the processing of the paper through either final acceptance or rejection. Once a paper has been accepted in final form, the author was asked to send the tex files of the final version to the submission email address. A simple Perl program was used to keep track of the status of each paper and to generate reports that were sent to editors on a regular basis during the processing stage. A total of 79 papers were received, of which 72 were accepted and seven rejected.
One reason for the process to take longer than planned was the "too many cooks" syndrome. In similar publication projects I had been involved with in the past, I had control over the production process through the delivery of the manuscript to the publisher, either as camera ready copy or as postscript files. These projects proceeded smoothly, were close to schedule, and everyone was pleased with the quality of the end product.
By contrast, this project involved at least six people - four at the publisher's end and two at the editors' end - all of whom had a say in the production process, but not all of whom were knowledgeable in the technical issues that had to be dealt with. However well-intended such a scheme was, instead of reducing individual workloads, it had the opposite effect, as communicating with all parties involved and trying to come to understandings on various technical issues introduced a significant overhead, that produced no tangible results, and a great deal of frustration and wasted time and energy on all sides. If it hadn't been for this overhead, the bulk of the production could have been completed in Summer 2001.
Another reason why the original schedule couldn't be adhered to is because people tend to miss deadlines more often than not, and sometimes by significant amounts. Authors and referees are usually busy with teaching and other obligations and underestimate the time it takes to complete a paper or a refereeing job. (I do sympathize here, as I often find myself under similar pressures and miss deadlines as a result.) The only way to ensure publication on the original schedule would have been through draconian enforcement of deadlines. However, this would have meant leaving out a good number of papers that came in late, some of whom turned out to be among the best papers in the proceedings. In addition, the quality of the papers would have suffered, as authors under time pressure would have had less time to proofread or polish a paper, and the refereeing would have been more perfunctory under tight deadlines.
Instead of rigorously enforcing deadlines, it was decided to give generous amounts of additional time to authors and referees, in order to ensure the highest scientific and production quality of the contributions to the proceedings. As a result, not a single paper had to be turned down because of a missed deadline, and all papers passed through a full refereeing process.
Here is how the timing of the three main stages in the processing of papers - submission, refereeing/revising, and proofing - turned out:
For most papers I did some amount of preprocessing and postprocessing, and I completely handled some of the TeXnically more complex papers, but the bulk of the TeXnical work was outsourced to a TeXnically experienced typist. Some TeXnical details on the typesetting of the proceedings are described here.
As a result, the quality of the editing of conference proceedings articles seems to depend mainly on the writing and editing skills of the authors, and the extent to which referees and editors are willing and able to polish up papers that are deficient in terms of language and style. There seems to be no correlation between the quality of the editing and the cost of a conference proceedings; I have come across articles published in some of the most expensive conference proceedings containing numerous, and blatant, grammatical errors.
The standard set for the editing quality of the Millennial Conference Proceedings was significantly higher than that seen in typical conference proceedings, and comparable to that employed by traditional mathematical journals. Papers written by native English speakers were generally in good shape, and no detailed copy-editing was done on those papers other than a spell-check and some routine checks for consistency in style. However, of the remaining papers - about two dozen papers altogether - nearly all had some language problems, ranging from minor stylistic gaffes to very blatant grammatical blunders, and getting these papers into shape required a considerable investment of time. I read through, and copy-edited each of these papers, often going through several rounds of revisions until I found the result satisfactory. Most papers took several hours to copy-edit, and some took significantly longer. In most cases, the editing could not have been done without some knowledge of the underlying mathematics, and using professional copy-editors without adequate mathematical background might have resulted in a number of "mis-corrections" that may have been grammatically correct, but which distorted the mathematical meaning (aside from fact that the cost would have been prohibitive).
Although authors were given the option to receive Galley proofs as hard copies, none did request a hard copy. All but three of the corrections received were sent by email; two were sent by fax, and only one came by regular mail.
Doing the entire Galley proof cycle electronically is vastly more efficient than the traditional method of sending out paper copies of Gally proofs, and the experience here shows that this can be done successfully, to everybody's satisfaction. Some of the benefits include:
The only significant monetary expenditures in the production of the proceedings were $2,400 for TeXnical typesetting by a local freelancer, and about $300 for graduate student help in checking an early draft for possible problems. Incidental costs for mailing, copying, and laserprinting, were absorbed by the department, but since most of the communication between authors, editors, and referees was carried out electronically, those costs were relatively minor. Even with these costs added in, the total amount of money spent on the production was well below $3,000.
The production, however, did require a significant investment of my time. Between July 2001 and March 2002, when the bulk of the production work took place, I spent a total of about 400 hours working on these proceedings. A detailed breakdown of this work can be found under this link. The bulk of this time, about 260 hours, was devoted to working on individual papers - mostly copy-editing, but also some TeX work, and correcting of Galley proofs. Without the thorough copy-editing that these proceedings have undergone, the total production time would have been cut in half, and I suspect that most conference proceedings (in contrast to journals) are produced in this manner, with little, if any, copy-editing. However, I made a conscious decision not to cut corners here, in order to meet the quality standards set at the outset.
About half of the remaining time, some 60 hours, was spent on various global tasks - such as checking headers for consistency, or assembling a list of addresses for contributors - and on writing programs to automate a variety of tasks - such as conversions between different Tex formats, paginating papers, or creating indices and tables of contents. While writing these programs represented a significant time investment, it paid off by saving dozens of hours of manual work. Also, it was largely a one-time investment, as many of the programs can be adapted to other situations and used again in similar projects.
In contrast to the work on individual papers which is essentially proportional to the number of pages, most of these "global tasks" scale well, so that a project that is twice the size would require only a marginally higher time investment on such global tasks.
The remaining 65 hours was essentially nonproductive overhead - time that I had to spend answering correspondence from the various parties, both at the publisher's end and here, that had a say in the production process. This was something I did not expect; in past publication projects I had been in full control of the technical side of the production, and all the publisher expected was to receive the finished manuscript in postscript form. By contrast, here, the publisher insisted on being actively involved throughout the production process. However well-intentioned such an involvement was (it would have been quite appropriate, and probably welcome, in situations where authors lack the technical expertise to handle the task on their own), it only had the effect of slowing things down. Instead of spending time more productively, I ended up spending a substantial share of my time trying to communicate complex technical issues to people with little technical knowledge and justify the merits of certain technical choices I made - for instance, trying to explain, after being instructed to use method X, why X is not appropriate in this situation and that method Y should be used instead, only to be told again that X must be used and Y is not an option, and after a few more exchanges and similar "nyet's", and a dozen hours of my time wasted on this, finally getting the okay to use method Y. Having to deal with this sort of thing on an ongoing basis is frustrating and demoralizing, and it was one reason why the original time plan could not be adhered to.